3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server primary kernels built and tested during release cycle:
6 3.10.0-957.27.2.el7 (RHEL7.6)
7 3.10.0-1062.18.1.el7 (RHEL7.7)
8 3.10.0-1127.8.2.el7 (RHEL7.8)
9 * Server known to build on the following kernels (others may also work):
10 2.6.32-696.18.7.el6 (RHEL6.9)
11 3.10.0-862.14.4.el7 (RHEL7.5)
12 4.4.120-92.70 (SLES12 SP2)
13 4.4.180-94.97 (SLES12 SP3)
14 4.4.0-142 (Ubuntu 16.04)
15 4.15.0-45 (Ubuntu 18.04)
16 vanilla linux 4.6.7 (ZFS only)
17 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
18 * Client primary unpatched kernels built and tested during release cycle:
19 3.10.0-957.27.2.el7 (RHEL7.6)
20 3.10.0-1062.18.1.el7 (RHEL7.7)
21 3.10.0-1127.8.2.el7 (RHEL7.8)
22 4.18.0-80.11.2.el8 (RHEL8.0)
23 4.4.180-94.97 (SLES12 SP3)
24 4.12.14-95.45 (SLES12 SP4)
25 4.12.14-122.20 (SLES12 SP5)
26 4.15.0-45 (Ubuntu 18.04)
27 * Client known to build on unpatched kernels (others may also work):
28 2.6.32-696.18.7.el6 (RHEL6.9)
29 2.6.32-754.6.3.el6 (RHEL6.10)
30 3.10.0-862.14.4.el7 (RHEL7.5)
31 4.14.0-49.13.1.el7a (RHEL7.5)
32 4.14.0-115.2.2.el7a (RHEL7.6)
33 4.18.0-147.3.1.el8 (RHEL8.1)
34 4.18.0-193.6.3.el8 (RHEL8.2)
35 4.4.120-92.70 (SLES12 SP2)
36 4.4.0-142 (Ubuntu 16.04)
37 vanilla linux 4.15.0 (ZFS only)
38 * Recommended e2fsprogs version: 1.45.6.wc1 or newer
39 * Recommended ZFS / SPL version: 0.7.13
40 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
41 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
42 stack overflow. For more information, please refer to bugzilla 17630.
43 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
44 the re-exporting nodes to be mounted with "32bitapi" mount option
45 * Note for Ubuntu 14.04.5:
46 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
47 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
51 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
52 for currently supported client and server kernel versions.
53 * Server primary kernels built and tested during release cycle:
54 3.10.0-957.27.2.el7 (RHEL7.6)
55 3.10.0-1062.18.1.el7 (RHEL7.7)
56 3.10.0-1127.8.2.el7 (RHEL7.8)
57 * Server known to build on the following kernels (others may also work):
58 2.6.32-696.18.7.el6 (RHEL6.9)
59 3.10.0-862.14.4.el7 (RHEL7.5)
60 4.4.120-92.70 (SLES12 SP2)
61 4.4.180-94.97 (SLES12 SP3)
62 4.4.0-142 (Ubuntu 16.04)
63 4.15.0-45 (Ubuntu 18.04)
64 vanilla linux 4.6.7 (ZFS only)
65 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
66 * Client primary unpatched kernels built and tested during release cycle:
67 3.10.0-957.27.2.el7 (RHEL7.6)
68 3.10.0-1062.18.1.el7 (RHEL7.7)
69 3.10.0-1127.8.2.el7 (RHEL7.8)
70 4.18.0-80.11.2.el8 (RHEL8.0)
71 4.4.180-94.97 (SLES12 SP3)
72 4.12.14-95.45 (SLES12 SP4)
73 4.12.14-122.20 (SLES12 SP5)
74 4.15.0-45 (Ubuntu 18.04)
75 * Client known to build on unpatched kernels (others may also work):
76 2.6.32-696.18.7.el6 (RHEL6.9)
77 2.6.32-754.6.3.el6 (RHEL6.10)
78 3.10.0-862.14.4.el7 (RHEL7.5)
79 4.14.0-49.13.1.el7a (RHEL7.5)
80 4.14.0-115.2.2.el7a (RHEL7.6)
81 4.18.0-147.8.1.el8 (RHEL8.1)
82 4.18.0-193.1.2.el8 (RHEL8.2)
83 4.4.120-92.70 (SLES12 SP2)
84 4.4.0-142 (Ubuntu 16.04)
85 vanilla linux 4.15.0 (ZFS only)
86 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
87 * Recommended ZFS / SPL version: 0.7.13
88 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
89 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
90 stack overflow. For more information, please refer to bugzilla 17630.
91 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
92 the re-exporting nodes to be mounted with "32bitapi" mount option
93 * Note for Ubuntu 14.04.5:
94 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
95 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
99 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
100 for currently supported client and server kernel versions.
101 * Server primary kernels built and tested during release cycle:
102 3.10.0-957.27.2.el7 (RHEL7.6)
103 3.10.0-1062.9.1.el7 (RHEL7.7)
104 * Server known to build on the following kernels (others may also work):
105 2.6.32-696.18.7.el6 (RHEL6.9)
106 3.10.0-862.14.4.el7 (RHEL7.5)
107 4.4.120-92.70 (SLES12 SP2)
108 4.4.180-94.97 (SLES12 SP3)
109 4.4.0-142 (Ubuntu 16.04)
110 4.15.0-45 (Ubuntu 18.04)
111 vanilla linux 4.6.7 (ZFS only)
112 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
113 * Client primary unpatched kernels built and tested during release cycle:
114 3.10.0-957.27.2.el7 (RHEL7.6)
115 3.10.0-1062.9.1.el7 (RHEL7.7)
116 4.18.0-80.11.2.el8 (RHEL8.0)
117 4.4.180-94.97 (SLES12 SP3)
118 4.12.14-95.29 (SLES12 SP4)
119 4.15.0-45 (Ubuntu 18.04)
120 * Client known to build on unpatched kernels (others may also work):
121 2.6.32-696.18.7.el6 (RHEL6.9)
122 2.6.32-754.6.3.el6 (RHEL6.10)
123 3.10.0-862.14.4.el7 (RHEL7.5)
124 4.14.0-49.13.1.el7a (RHEL7.5)
125 4.14.0-115.2.2.el7a (RHEL7.6)
126 4.18.0-147.3.1.el8 (RHEL8.1)
127 4.4.120-92.70 (SLES12 SP2)
128 4.4.0-142 (Ubuntu 16.04)
129 vanilla linux 4.15.0 (ZFS only)
130 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
131 * Recommended ZFS / SPL version: 0.7.13
132 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
133 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
134 stack overflow. For more information, please refer to bugzilla 17630.
135 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
136 the re-exporting nodes to be mounted with "32bitapi" mount option
137 * Note for Ubuntu 14.04.5:
138 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
139 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
143 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
144 for currently supported client and server kernel versions.
145 * Server primary kernels built and tested during release cycle:
146 3.10.0-957.27.2.el7 (RHEL7.6)
147 3.10.0-1062.1.1.el7 (RHEL7.7)
148 * Server known to build on the following kernels (others may also work):
149 2.6.32-696.18.7.el6 (RHEL6.9)
150 3.10.0-862.14.4.el7 (RHEL7.5)
151 4.4.120-92.70 (SLES12 SP2)
152 4.4.180-94.97 (SLES12 SP3)
153 4.4.0-142 (Ubuntu 16.04)
154 4.15.0-45 (Ubuntu 18.04)
155 vanilla linux 4.6.7 (ZFS only)
156 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
157 * Client primary unpatched kernels built and tested during release cycle:
158 3.10.0-957.27.2.el7 (RHEL7.6)
159 3.10.0-1062.1.1.el7 (RHEL7.7)
160 4.18.0-80.7.1.el8 (RHEL8.0)
161 4.4.180-94.97 (SLES12 SP3)
162 4.12.14-95.29 (SLES12 SP4)
163 4.15.0-45 (Ubuntu 18.04)
164 * Client known to build on unpatched kernels (others may also work):
165 2.6.32-696.18.7.el6 (RHEL6.9)
166 2.6.32-754.6.3.el6 (RHEL6.10)
167 3.10.0-862.14.4.el7 (RHEL7.5)
168 4.14.0-49.13.1.el7a (RHEL7.5)
169 4.14.0-115.2.2.el7a (RHEL7.6)
170 4.4.120-92.70 (SLES12 SP2)
171 4.4.0-142 (Ubuntu 16.04)
172 vanilla linux 4.15.0 (ZFS only)
173 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
174 * Recommended ZFS / SPL version: 0.7.13
175 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
176 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
177 stack overflow. For more information, please refer to bugzilla 17630.
178 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
179 the re-exporting nodes to be mounted with "32bitapi" mount option
180 * Note for Ubuntu 14.04.5:
181 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
182 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
186 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
187 for currently supported client and server kernel versions.
188 * Server known to build on the following kernels (others may also work):
189 2.6.32-696.18.7.el6 (RHEL6.9)
190 3.10.0-862.14.4.el7 (RHEL7.5)
191 3.10.0-957.10.1.el7 (RHEL7.6)
192 4.4.120-92.70 (SLES12 SP2)
193 4.4.162-94.69 (SLES12 SP3)
194 4.4.0-142 (Ubuntu 16.04)
195 4.15.0-45 (Ubuntu 18.04)
196 vanilla linux 4.6.7 (ZFS only)
197 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
198 * Client known to build on unpatched kernels (others may also work):
199 2.6.32-696.18.7.el6 (RHEL6.9)
200 2.6.32-754.6.3.el6 (RHEL6.10)
201 3.10.0-862.14.4.el7 (RHEL7.5)
202 3.10.0-957.10.1.el7 (RHEL7.6)
203 4.14.0-49.13.1.el7a (RHEL7.5)
204 4.14.0-115.2.2.el7a (RHEL7.6)
205 4.4.120-92.70 (SLES12 SP2)
206 4.4.162-94.69 (SLES12 SP3)
207 4.12.14-95.13 (SLES12 SP4)
208 4.4.0-142 (Ubuntu 16.04)
209 4.15.0-45 (Ubuntu 18.04)
210 vanilla linux 4.15.0 (ZFS only)
211 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
212 * Recommended ZFS / SPL version: 0.7.13
213 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
214 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
215 stack overflow. For more information, please refer to bugzilla 17630.
216 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
217 the re-exporting nodes to be mounted with "32bitapi" mount option
218 * Note for Ubuntu 14.04.5:
219 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
220 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
225 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
226 for currently supported client and server kernel versions.
227 * Server known to build on the following kernels (others may also work):
228 2.6.32-696.18.7.el6 (RHEL6.9)
229 3.10.0-862.14.4.el7 (RHEL7.5)
230 3.10.0-957.10.1.el7 (RHEL7.6)
231 4.4.120-92.70 (SLES12 SP2)
232 4.4.162-94.69 (SLES12 SP3)
233 4.4.0-142 (Ubuntu 16.04)
234 4.15.0-45 (Ubuntu 18.04)
235 vanilla linux 4.6.7 (ZFS only)
236 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
237 * Client known to build on unpatched kernels (others may also work):
238 2.6.32-696.18.7.el6 (RHEL6.9)
239 2.6.32-754.6.3.el6 (RHEL6.10)
240 3.10.0-862.14.4.el7 (RHEL7.5)
241 3.10.0-957.10.1.el7 (RHEL7.6)
242 4.14.0-49.13.1.el7a (RHEL7.5)
243 4.14.0-115.2.2.el7a (RHEL7.6)
244 4.4.120-92.70 (SLES12 SP2)
245 4.4.162-94.69 (SLES12 SP3)
246 4.12.14-95.13 (SLES12 SP4)
247 4.4.0-142 (Ubuntu 16.04)
248 4.15.0-45 (Ubuntu 18.04)
249 vanilla linux 4.15.0 (ZFS only)
250 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
251 * Recommended ZFS / SPL version: 0.7.13
252 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
253 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
254 stack overflow. For more information, please refer to bugzilla 17630.
255 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
256 the re-exporting nodes to be mounted with "32bitapi" mount option
257 * Note for Ubuntu 14.04.5:
258 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
259 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
263 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
264 for currently supported client and server kernel versions.
265 * Server known to build on the following kernels (others may also work):
266 2.6.32-696.18.7.el6 (RHEL6.9)
267 3.10.0-862.14.4.el7 (RHEL7.5)
268 3.10.0-957.el7 (RHEL7.6)
269 4.4.120-92.70 (SLES12 SP2)
270 4.4.162-94.69 (SLES12 SP3)
271 4.4.0-131 (Ubuntu 16.04)
272 4.15.0-32 (Ubuntu 18.04)
273 vanilla linux 4.6.7 (ZFS only)
274 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
275 * Client known to build on unpatched kernels (others may also work):
276 2.6.32-696.18.7.el6 (RHEL6.9)
277 2.6.32-754.6.3.el6 (RHEL6.10)
278 3.10.0-862.14.4.el7 (RHEL7.5)
279 3.10.0-957.el7 (RHEL7.6)
280 4.14.0-49.13.1.el7a (RHEL7.5)
281 4.14.0-115.2.2.el7a (RHEL7.6)
282 4.4.120-92.70 (SLES12 SP2)
283 4.4.162-94.69 (SLES12 SP3)
284 4.4.0-131 (Ubuntu 16.04)
285 4.15.0-32 (Ubuntu 18.04)
286 vanilla linux 4.15.0 (ZFS only)
287 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
288 * Recommended ZFS / SPL version: 0.7.9
289 * Tested with ZFS / SPL version: 0.7.9
290 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
291 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
292 stack overflow. For more information, please refer to bugzilla 17630.
293 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
294 the re-exporting nodes to be mounted with "32bitapi" mount option
295 * Note for Ubuntu 14.04.5:
296 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
297 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
299 --------------------------------------------------------------------------------
302 2018-04-03 Intel Corporation
304 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
305 for currently supported client and server kernel versions.
306 * Server known to build on patched kernels:
307 2.6.32-431.29.2.el6 (RHEL6.5)
308 2.6.32-504.30.3.el6 (RHEL6.6)
309 2.6.32-573.26.1.el6 (RHEL6.7)
310 2.6.32-642.15.1.el6 (RHEL6.8)
311 2.6.32-696.18.7.el6 (RHEL6.9)
312 3.10.0-693.21.1.el7 (RHEL7.4)
313 3.0.101-0.47.71 (SLES11 SP3)
314 3.0.101-107 (SLES11 SP4)
315 3.12.74-60.64.40 (SLES12 SP1)
316 4.4.114-92.64 (SLES12 SP2)
317 4.4.114-94.11 (SLES12 SP3)
318 3.13.0-101 (Ubuntu 14.04, ZFS only)
319 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
320 4.4.0-116 (Ubuntu 16.04)
321 vanilla linux 4.6.7 (ZFS only)
322 * Client known to build on unpatched kernels:
323 2.6.32-431.29.2.el6 (RHEL6.5)
324 2.6.32-504.30.3.el6 (RHEL6.6)
325 2.6.32-573.26.1.el6 (RHEL6.7)
326 2.6.32-642.15.1.el6 (RHEL6.8)
327 2.6.32-696.18.7.el6 (RHEL6.9)
328 3.10.0-693.21.1.el7 (RHEL7.4)
329 3.0.101-0.47.71 (SLES11 SP3)
330 3.0.101-107 (SLES11 SP4)
331 3.12.74-60.64.40 (SLES12 SP1)
332 4.4.114-92.64 (SLES12 SP2)
333 4.4.114-94.11 (SLES12 SP3)
334 3.13.0-101 (Ubuntu 14.04)
335 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
336 4.4.0-116 (Ubuntu 16.04)
338 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
339 * Recommended ZFS / SPL version: 0.7.6
340 * Tested with ZFS / SPL version: 0.7.6
341 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
342 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
343 stack overflow. For more information, please refer to bugzilla 17630.
344 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
345 the re-exporting nodes to be mounted with "32bitapi" mount option
346 * Note for Ubuntu 14.04.5:
347 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
348 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
350 --------------------------------------------------------------------------------
352 2017-07-05 Intel Corporation
354 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
355 for currently supported client and server kernel versions.
356 * Server known to build on patched kernels:
357 2.6.32-431.29.2.el6 (RHEL6.5)
358 2.6.32-504.30.3.el6 (RHEL6.6)
359 2.6.32-573.26.1.el6 (RHEL6.7)
360 2.6.32-642.15.1.el6 (RHEL6.8)
361 2.6.32-696.3.1.el6 (RHEL6.9)
362 3.10.0-514.21.1.el7 (RHEL7.3)
363 3.0.101-0.47.71 (SLES11 SP3)
364 3.0.101-100 (SLES11 SP4)
365 3.12.74-60.64.40 (SLES12 SP1)
366 4.4.59-92.17 (SLES12 SP2)
367 vanilla linux 4.6.7 (ZFS only)
368 * Client known to build on unpatched kernels:
369 2.6.32-431.29.2.el6 (RHEL6.5)
370 2.6.32-504.30.3.el6 (RHEL6.6)
371 2.6.32-573.26.1.el6 (RHEL6.7)
372 2.6.32-642.15.1.el6 (RHEL6.8)
373 2.6.32-696.3.1.el6 (RHEL6.9)
374 3.10.0-514.21.1.el7 (RHEL7.3)
375 3.0.101-0.47.71 (SLES11 SP3)
376 3.0.101-100 (SLES11 SP4)
377 3.12.74-60.64.40 (SLES12 SP1)
378 4.4.59-92.17 (SLES12 SP2)
380 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
381 * Recommended ZFS / SPL version: 0.6.5.9
382 * Tested with ZFS / SPL version: 0.6.5.9
383 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
384 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
385 stack overflow. For more information, please refer to bugzilla 17630.
386 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
387 the re-exporting nodes to be mounted with "32bitapi" mount option
389 --------------------------------------------------------------------------------
391 12-07-2016 Intel Corporation
393 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
394 for currently supported client and server kernel versions.
395 * Server known to build on patched kernels:
396 2.6.32-431.29.2.el6 (RHEL6.5)
397 2.6.32-504.30.3.el6 (RHEL6.6)
398 2.6.32-573.26.1.el6 (RHEL6.7)
399 2.6.32-642.11.1.el6 (RHEL6.8)
400 3.10.0-514.el7 (RHEL7.3)
401 3.0.101-0.47.71 (SLES11 SP3)
402 3.0.101-80 (SLES11 SP4)
403 3.12.62-60.64.8 (SLES12 SP1)
404 vanilla linux 4.5.7 (ZFS only)
405 * Client known to build on unpatched kernels:
406 2.6.32-431.29.2.el6 (RHEL6.5)
407 2.6.32-504.30.3.el6 (RHEL6.6)
408 2.6.32-573.26.1.el6 (RHEL6.7)
409 2.6.32-642.11.1.el6 (RHEL6.8)
410 3.10.0-514.el7 (RHEL7.3)
411 3.0.101-0.47.71 (SLES11 SP3)
412 3.0.101-80 (SLES11 SP4)
413 3.12.62-60.64.8 (SLES12 SP1)
415 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
416 * Recommended ZFS / SPL version: 0.6.5.8
417 * Tested with ZFS / SPL version: 0.6.5.8
418 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
419 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
420 stack overflow. For more information, please refer to bugzilla 17630.
421 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
422 the re-exporting nodes to be mounted with "32bitapi" mount option
424 --------------------------------------------------------------------------------
426 02-29-2016 Intel Corporation
428 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
429 for currently supported client and server kernel versions.
430 * Server known to build on patched kernels:
431 2.6.32-431.29.2.el6 (RHEL6.5)
432 2.6.32-504.30.3.el6 (RHEL6.6)
433 2.6.32-573.12.1.el6 (RHEL6.7)
434 3.10.0-327.3.1.el7 (RHEL7.2)
435 3.0.101-0.47.71 (SLES11 SP3)
436 3.0.101-68 (SLES11 SP4)
437 vanilla linux 4.2.1 (ZFS only)
438 * Client known to build on unpatched kernels:
439 2.6.32-431.29.2.el6 (RHEL6.5)
440 2.6.32-504.30.3.el6 (RHEL6.6)
441 2.6.32-573.12.1.el6 (RHEL6.7)
442 3.10.0-327.3.1.el7 (RHEL7.2)
443 3.0.101-0.47.71 (SLES11 SP3)
444 3.0.101-68 (SLES11 SP4)
447 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
448 * Recommended ZFS / SPL version: 0.6.4.2
449 * Tested with ZFS / SPL version: 0.6.4.2
450 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
451 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
452 stack overflow. For more information, please refer to bugzilla 17630.
453 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
454 the re-exporting nodes to be mounted with "32bitapi" mount option
456 --------------------------------------------------------------------------------
458 03-10-2015 Intel Corporation
460 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
461 for currently supported client and server kernel versions.
462 * Server known to build on patched kernels:
463 2.6.32-431.29.2.el6 (RHEL6.5)
464 2.6.32-504.8.1.el6 (RHEL6.6)
465 3.0.101-0.46 (SLES11 SP3)
466 * Client known to build on unpatched kernels:
467 2.6.32-431.29.2.el6 (RHEL6.5)
468 2.6.32-504.8.1.el6 (RHEL6.6)
469 3.10.0-123.20.1.el7 (RHEL7)
470 3.0.101-0.46 (SLES11 SP3)
471 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
472 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
473 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
474 stack overflow. For more information, please refer to bugzilla 17630.
475 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
476 the re-exporting nodes to be mounted with "32bitapi" mount option
478 Severity : enhancement
480 Description: control OST-index in IDIF via ROCOMPAT flag.
481 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
482 last_rcvd file. For new formatted OST device, it will be auto set;
483 for the case of upgrading from old OST device, you can enable it
484 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
485 enabled, for new created OST-object, its IDIF-in-LMA will contain
486 the OST-index; for the existing OST-object, the OSD will convert
487 old format IDIF as new format IDIF with OST-index stored in the
488 LMA EA when accessing such OST-object or via OI scrub. Once such
489 flag is enabled, it cannot be reverted back, so the system cannot
490 be downgraded to the orignal incompatible version.
492 --------------------------------------------------------------------------------
494 07-30-2014 Intel Corporation
496 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
497 for currently supported client and server kernel versions.
498 * Server known to build on patched kernels:
499 2.6.32-431.20.3.el6 (RHEL6.5)
500 3.0.101-0.31 (SLES11 SP3)
501 * Client known to build on unpatched kernels:
502 2.6.32-431.20.3.el6 (RHEL6.5)
503 3.0.101-0.31 (SLES11 SP3)
506 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
507 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
508 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
509 stack overflow. For more information, please refer to bugzilla 17630.
510 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
511 the re-exporting nodes to be mounted with "32bitapi" mount option
513 --------------------------------------------------------------------------------
515 10-31-2013 Intel Corporation
517 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
518 * Server builds for kernels:
519 2.6.32-358.18.1.el6 (RHEL6.4)
520 * Client builds for unpatched kernels:
521 2.6.32-358.18.1.el6 (RHEL6.4)
522 3.0.92-0.8 (SLES11 SP3)
524 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
525 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
526 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
527 stack overflow. For more information, please refer to bugzilla 17630.
528 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
529 the re-exporting nodes to be mounted with "32bitapi" mount option
531 --------------------------------------------------------------------------------
533 05-30-2013 Intel Corporation
535 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
536 * Server builds for kernels:
537 2.6.32-358.6.2.el6 (RHEL6.4)
538 * Client builds for unpatched kernels:
539 2.6.32-358.6.2.el6 (RHEL6.4)
540 3.0.74-0.6.6 (SLES11 SP2)
542 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
543 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
544 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
545 stack overflow. For more information, please refer to bugzilla 17630.
546 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
547 the re-exporting nodes to be mounted with "32bitapi" mount option
549 --------------------------------------------------------------------------------
551 9-30-2012 Whamcloud, Inc.
553 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
554 * Server support for kernels:
555 2.6.32-279.5.1.el6 (RHEL6)
556 * Client support for unpatched kernels:
557 2.6.32-279.5.1.el6 (RHEL6)
559 2.6.18-238.19.1.el5 (RHEL5)
560 2.6.16.60-0.69.1 (SLES10)
561 * Recommended e2fsprogs version: 1.41.90.wc4
562 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
563 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
564 stack overflow. For more information, please refer to bugzilla 17630.
565 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
566 the re-exporting nodes to be mounted with "32bitapi" mount option
568 Severity : enhancement
570 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
572 Severity : enhancement
574 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
576 Severity : enhancement
578 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
580 Severity : enhancement
582 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
584 Severity : enhancement
586 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
588 --------------------------------------------------------------------------------
590 04-30-2012 Whamcloud, Inc.
592 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
593 * Server support for kernels:
594 2.6.32-220.4.2.el6 (RHEL6)
595 * Client support for unpatched kernels:
596 2.6.32-220.4.2.el6 (RHEL6)
598 2.6.18-238.19.1.el5 (RHEL5)
599 2.6.16.60-0.69.1 (SLES10)
600 * Recommended e2fsprogs version: 1.41.90.wc4
601 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
602 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
603 stack overflow. For more information, please refer to bugzilla 17630.
604 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
605 the re-exporting nodes to be mounted with "32bitapi" mount option
607 Severity : enhancement
609 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
611 Severity : enhancement
613 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
615 Severity : enhancement
617 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
619 Severity : enhancement
621 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
623 --------------------------------------------------------------------------------
625 09-30-2011 Whamcloud, Inc.
627 * Server support for kernels:
628 2.6.18-238.19.1.el5 (RHEL5)
629 2.6.32-131.6.1.el6 (RHEL6)
630 * Client support for unpatched kernels:
631 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
632 2.6.18-238.19.1.el5 (RHEL5)
633 2.6.32-131.6.1.el6 (RHEL6)
634 2.6.29.4-167.fc11 (FC11)
635 2.6.16.60-0.69.1 (SLES10)
637 * Recommended e2fsprogs version: 1.41.90.wc3
638 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
639 Lustre filesystem with 4K stack may cause a stack overflow. For more
640 information, please refer to bugzilla 17630.
641 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
642 mounted with "32bitapi" mount option
644 Severity : enhancement
645 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
647 Severity : enhancement
648 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
650 Severity : enhancement
651 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
654 Frequency : only with SLES10
656 Description: Use OFED "KMP" provided by Novell
657 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
658 Lustre is now built against this package. That means you need to
659 install the ofed-kmp package from Novell for the patchless client.
660 Note that the ofed-kmp that Novell ships may not exactly match the
661 kernel version but should still be compatible.
663 Severity : enhancement
665 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
666 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
668 --------------------------------------------------------------------------------
670 2010-07-15 Oracle, Inc.
672 * Server support for kernels:
673 2.6.18-164.11.1.el5 (RHEL 5)
674 2.6.18-164.11.1.0.1.el5 (OEL 5)
675 * Client support for unpatched kernels:
676 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
677 2.6.18-164.11.1.el5 (RHEL 5),
678 2.6.18-164.11.1.0.1.el5 (OEL 5)
679 2.6.16.60-0.42.8 (SLES 10),
681 2.6.29.4-167.fc11 (FC11)
682 * Recommended e2fsprogs version: 1.41.10-sun2
684 Severity : enhancement
686 Description: Distribute lfs_migrate script with lustre
687 Details : The lfs_migrate mentioned in the manual is now included in the
688 lustre tree in lustre/scripts and will distributed in lustre RPMs.
690 Severity : enhancement
692 Description: Failover nid list added to import proc file.
694 Severity : enhancement
696 Description: "som_preview" mount option required to use SOM
697 Details : The clients will need to use som_preview mount option
701 Description: Lustre does not do 1MB IOs to HW RAID
702 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
704 Severity : enhancement
706 Description: lfs quota output is not very convenient for awk/sed-parsing
707 Details : Some positions in lfs quota output table could be empty or
708 non-empty which made it hard to parse it with scripts, now
709 a dash is put instead of space where there is not supposed
714 Description: Update to OFED 1.5.1
715 Details : OFED 1.5.1 will be used in environments that don't use
716 vendor supplied OFED.
720 Description: Remove lnet. prefix from lctl params display
721 Details : Also renamed lnet.memused to lnet_memused to
726 Description: Use conf_param -d to remove permanant settings.
730 Description: Add list_param -R
731 Details : lctl list_param now supports recursive mode.
734 Bugzilla : 21829, 22312
735 Description: Fix llobdstat.
736 Details : 1. Now updating the write_bytes counter
737 2. Port script fixes from 1.8.
741 Description: decrease the usage of memory on clients.
742 Details : 1. On clients, recycle dentries and inodes unused.
743 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
748 Description: llobdstat fix and enhancement
749 Details : add a counter to set a limit to how many samples will be returned
750 fix a wildcard in the path to limit to obdfilter stats only
754 Description: lfs check only allowed for root
755 Details : Code cleanups and added a regresion test of non-root use of
760 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
761 Details : Fix hr->hr_index overload.
765 Description: Fix FIEMAP functionality.
769 Description: FC11 patchless client support added
773 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
774 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
778 Description: New acc-sm script and various test-framework enhancments.
782 Description: Previously landed patch was reverted.
786 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
790 Description: Add ldiskfs maxdirsize mount option.
794 Description: Fix mkfs.lustre support for 16TB LUNs.
798 Description: Print mmp_check_interval
799 Details : print mmp_check_interval and make it possible to abort
800 mount operation in case it takes too long.
804 Description: Add cascading_rw to lustre mpi tests.
808 Description: Store SOM data in EA.
809 Details : Also several SOM related cleanups.
813 Description: Add refcount to md_open_data
817 Description: Use refcount on lov_request_set to prevent memory corruption.
821 Description: Statahead: use parent dir's i_mutex for synchronization.
825 Description: loadgen improvements.
829 Description: Enable peer health detection in ksocklnd and ko2iblnd.
833 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
837 Description: Fix race on obd_devcie:md_stats
841 Description: Fix SLES 11 client builds.
842 Details : Resolved uninitalized pointer gcc warnings.
846 Description: Allow non-root access for "lfs check".
847 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
851 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
853 Severity : enhancement
855 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
856 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
860 Description: Sanity test_27p fails.
861 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
862 Sanity tests 27 were cleaned up and improved.
866 Description: Unify req format on client/servers
867 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
868 buffers and swabbers approach..
873 Description: drop join file stuff.
875 Severity : enhancement
877 Description: Add 2.6.27 and SLES11 (patchless client) support
881 Description: Fix in ptlrpc_expire_one_request() to print the signed time
884 Severity : enhancement
886 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
889 Severity : enhancement
891 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
893 Severity : enhancement
895 Description: correctly handle big reply message.
896 Details : send LNet event if reply is bigger than buffer and adjust this buffer
901 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
903 Severity : enhancement
905 Description: Add support for OEL5.
909 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
911 Severity : enhancement
913 Description: "weak-modules" support
914 Details : Implement "weak-modules" support which enables kernel modules
915 to be used with any kernel that implements the same kABI. In
916 order to achieve this modules are now installed in
917 /lib/modules/$(uname -r)/updates/kernel on all distributions.
921 Description: client could not reconnect to OST because of an active request.
922 Details : The client now retries to reconnect to the same server, if a
923 connect request failed with EBUSY or -EAGAIN. Test case 35b
924 conf-sanity.sh has been created.
928 Description: Don't allow make backward step on assignin osc next id.
929 Details : race between allocation next id and ll_sync thread can be cause
930 of set wrong osc next id and can be kill valid ost objects.
934 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
938 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
942 Description: Error handling in osc_statfs_interpret() has been improved.
943 Details : Check in osc_statfs_interpret() for EBADR.
947 Description: Do not send statfs() requests to OSTs disabled by administrator.
948 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
950 Severity : enhancement
952 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
954 Severity : enhancement
956 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
958 Severity : enhancement
960 Description: Enhancement for lfs(1) command to use numeric uid/gid.
964 Description: Improved cleanup procedure for conf-sanity test_46a.
968 Description: Do not start bulk transfer, if client tried to reconnect.
969 Details : Do not start bulk transfer, if client tried to reconnect,
970 but failed due to active rpcs. Check for exp_abort_active_req
971 has been implemented.
975 Description: Prevent inconsistences between linux and lustre mount structures.
976 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
977 Make the sleep interruptible.
979 Severity : enhancement
981 Description: provide server to client comms path
982 Details : server to client communications path via new LDLM_SET_INFO rpc
986 Description: truncate starts GFP_FS allocation under transaction causing deadlock
987 Details : ldiskfs_truncate calls grab_cache_page which may start page
988 allocation under an open transaction. This may lead to
989 calling prune_icache with consequent lustre reentrance.
993 Description: send correctly lsm on open replay
994 Details : MDS is trust to LSM size on replay open, but client can set wrong size
999 Description: enable client interface failover
1000 Details : When a child reconnects from another NID, properly update export
1001 nid hash position and ldlm reverse import.
1003 Severity : enhancement
1005 Description: Communicate OST degraded/readonly state via statfs to MDS
1006 Details : Flags in the statfs returned from OSTs indicate whether the
1007 OST is in a degraded RAID state, or if the filesystem has
1008 turned read-only after a filesystem error is detected.
1013 Description: don't panic if EPROTO was hit when reading symlink
1014 Details : correctly handling request reference in error cases.
1016 Severity : enhancement
1018 Description: Add LustreNetLink, a kernel-userspace communcation path.
1020 Severity : enhancement
1022 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1027 Description: access to llog context before init.
1028 Details : move handling CATALOGS file at osc layer and forbid access to llog
1029 context before init.
1032 Frequency : rare, only when client is writing to OST on local node
1034 Description: Avoid deadlock for local client writes
1035 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1036 memory freeing context. This allows OST threads to set the
1037 PF_MEMALLOC flag on task structures in order to allocate memory
1038 from reserved pools and complete IO.
1039 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1040 so that the OST threads generate memory pressure and allow
1041 inactive pages to be reclaimed.
1043 Severity : enhancement
1045 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1050 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1051 Details : move ldlm namespace creation in setup phase to avoid grab
1052 _lprocfs_lock with cli_sem held.
1056 Description: Temporarily disable grant shrink.
1057 Details : Disable the feature for debugging.
1060 Frequency : only when formatting test filesystems
1062 Description: Unable to run several mkfs.lustre on loop devices at the same time
1063 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1064 formatting. The solution is to proper handle the error.
1066 Severity : enhancement
1068 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1070 Severity : enhancement
1072 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1077 Description: resolve race between obd_disconnect and class_disconnect_exports
1078 Details : if obd_disconnect will be called to already disconnected export he
1079 forget release one reference and osc module can't unloaded.
1081 Severity : enhancement
1083 Description: Allow tuning service thread via /proc
1084 Details : For each service a new
1085 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1086 created that can be used to set min/max thread counts, and get the
1087 current number of running threads.
1092 Descriptoin: don't return error if have partially created objects for file.
1093 Details : lov_update_create_set uses set->set_success as index for created
1094 objects, so if some requests will be failed, they will have hole
1095 at end of array and qos_shrink_lsm for allocate correct lsm.
1097 Severity : enhancement
1099 Description: Update OFED support to 1.4.1
1104 Description: Reduce small size read RPC
1105 Details : Set read-ahead limit for every file and only do read-ahead when
1106 available read-ahead pages are bigger than 1M to avoid small size
1109 Severity : enhancement
1111 Description: move AT tunable parameters for more consistent usage
1112 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1114 Severity : enhancement
1116 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1117 Details : allow skip disconnected ost for send statfs request and hide error
1123 Description: fix racy locking of mballoc block bitmaps causing BUG
1124 Details : The locking of the mballoc buddy bitmap and the in-memory
1125 block bitmap was using two different spin locks in some
1126 cases. This made it possible to incorrectly access the
1127 mballoc bitmap while another process was modifying it,
1128 causing a sanity assertion to fail. While no on-disk corruption
1129 was reported, there was some risk of this happening.
1132 Frequency : rare, on llog test 6
1134 Descriptoin: don't allow connect to already connected import
1135 Details : allowing connect to already connected import is hide connecting problem.
1140 Descriptoin: allow kill process which wait statahead result
1141 Details : for some reasons 'ls' can stick in waiting result from statahead,
1142 in this case need way for kill this process.
1144 Severity : enhancement
1146 Description: Add state history info file, enhance import info file
1147 Details : Track import connection state changes in a new osc/mdc proc file;
1148 add overview-type data to the osc/mdc import proc file.
1150 Severity : enhancement
1152 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1154 Severity : enhancement
1156 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1161 Descriptoin: OST grant too much space to client even there are not enough space.
1162 Details : Client will shrink its grant cache to OST if there are no write
1163 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1164 this grant cache if there are already not enough avaible space
1165 (left_space < total_clients * 32M).
1168 Frequency : start MDS on uncleanly shutdowned MDS device
1170 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1171 Details : stay in waiting mds<>ost recovery finished produce random bugs
1172 due race between two ll_sync thread for one lov target. send
1173 ACTIVATE event only if connect realy finished and import have
1177 Frequency : rare, connect and disconnect target at same time
1179 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1180 Details : don't call obd_disconnect under lov_lock. this long time
1181 operation and can block ptlrpcd which answer to connect request.
1186 Descriptoin: don't lose wakeup for imp_recovery_waitq
1187 Details : recover_import_no_retry or invalidate_import and import_close can
1188 both sleep on imp_recovery_waitq, but we was send only one wakeup
1192 Frequency : always with long access acl
1194 Descriptoin: mds can't pack reply with long acl.
1195 Details : mds don't control size of acl but they limited by reint/getattr
1198 Severity : enhancement
1200 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1202 Severity : enhancement
1204 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1207 Frequency : start MDS on uncleanly shutdowned MDS device
1209 Descriptoin: aborting recovery hang on MDS
1210 Details : don't throttle destroy RPCs for the MDT.
1213 Frequency : on remount
1215 Description: external journal device not working after the remount
1216 Details : clear dev_rdonly flag for external journal devices in
1222 Description: shutdown vs evict race
1223 Details : client_disconnect_export vs connect request race.
1224 if client will evicted at this time - we start invalidate
1225 thread without referece to import and import can be freed
1229 Frequency : rare, need acl's on inode.
1231 Description: client can't handle ost additional correctly
1232 Details : if ost was added after client connected to mds client can have
1233 hit lnet_try_match_md ... to big messages to wide striped files.
1234 in this case need teach client to handle config events about add
1235 lov target and update client max ea size at that event.
1237 Severity : enhancement
1239 Description: Changelogs
1240 Details : Changelogs are a lightweight mechanism to track filesystem
1241 metadata and namespace changes. The changelog is recorded
1242 permanently on the MDTs, and is periodically "consumed" / purged
1243 when records are no longer needed.
1245 Severity : enhancement
1247 Description: compact fld format with extents
1248 Details : Store range of seq rather than every seq in FLD. Seq
1249 controller update FLD rather than clients. In Case of CMD, mdt0
1250 has FLD, all other metadata server act as non persistent proxy
1251 for FLD queries and cache fld entries in fld cache.
1256 Description: don't skip ost target if they assigned to file
1257 Details : Drop slow OSCs if we can, but not for requested start idx.
1258 This means "if OSC is slow and it is not the requested
1259 start OST, then it can be skipped, otherwise skip it only
1260 if it is inactive/recovering/out-of-space.
1264 Description: more cleanup in mds_lov
1265 Details : not send LOV EA under replay, we can't know about they size at this
1266 time. Don't allow client connect to mds before any ost connected,
1267 for avoid problems with LOV EA size and returning EIO to client.
1269 Severity : enhancement
1271 Description: Interoperability at server side (Disk interoperability)
1273 Severity : enhancement
1275 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1277 Severity : enhancement
1279 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1281 Severity : enhancement
1283 Description: New client IO stack (CLIO).
1285 Severity : enhancement
1287 Description: Commit on sharing. Eliminate inter-client dependencies between
1288 uncommitted transactions by doing transaction commits.
1289 Thereby clients may recovery independently.
1292 Frequency : Create a symlink file with a very long name
1294 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1295 Details : If there is no extra space in the request for early cancels,
1296 ldlm_req_handles_avail() returns 0 instead of a negative value.
1298 Severity : enhancement
1300 Description: Add /proc entry for import status
1301 Details : The mdc, osc, and mgc import directories now have
1302 an import directory that contains useful import data for debugging
1303 connection problems.
1305 Severity : enhancement
1307 Description: Re-disable certain /proc logging
1308 Details : Enable and disable client's offset_stats, extents_stats and
1309 extents_stats_per_process stats logging on the fly.
1312 Frequency : Only on FC kernels 2.6.22+
1314 Description: oops in statahead
1315 Details : Do not drop reference count for the dentry from VFS when lookup,
1316 VFS will do that by itself.
1318 Severity : enhancement
1320 Description: Generic /proc file permissions
1321 Details : Set /Proc file permissions in a more generic way to enable non-
1322 root users operate on some /proc files.
1326 Description: Hitting mdc_commit_close() ASSERTION
1327 Details : Properly handle request reference release in
1328 ll_release_openhandle().
1332 Description: quota recovery deadlock during mds failover
1333 Details : This patch includes att18982, att18236, att18237 in bz14840.
1335 1. fix osts hang when mds does failover with quotaon
1336 2. prevent watchdog storm when osts threads wait for the
1341 Frequency : only patchless client
1342 Description: add workaround for race between add/remove dentry from hash
1344 Severity : enhancement
1346 Description: Allow OST glimpses to return PW locks
1350 Description: LBUG when llog conf file is full
1351 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1355 Description: Prevent import from entering FULL state when server in recovery
1359 Description: service mount cannot take device name with ":"
1360 Details : Only when device name contains ":/" will mount treat it as
1366 Description: replace ptlrpcd with the statahead thread to interpret the async
1367 statahead RPC callback
1371 Frequency : on recovery
1372 Description: I/O failures after umount during fail back
1373 Details : if client reconnected to restarted server we need join to recovery
1374 instead of find server handler is changed and process self
1375 eviction with cancel all locks.
1377 Severity : enhancement
1379 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1381 Severity : enhancement
1383 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1385 Severity : enhancement
1387 Description: Upcall on Lustre log has been dumped
1388 Details : Allow for a user mode script to be called once a Lustre log has
1389 been dumped. It passes the filename of the dumped log to the
1390 script, the location of the script can be specified via
1391 /proc/sys/lnet/debug_log_upcall.
1396 Description: avoid idr_remove called for id which is not allocated.
1397 Details : Move assigment s_dev for clustered nfs to end of initialization,
1398 to avoid problem with error handling.
1403 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1404 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1405 into CONN_USED_HASH and this prodice warning when put connection
1406 again in unused hash.
1409 Severity : enhancement
1411 Description: File striping can now be set to use an arbitrary pool of OSTs.
1413 Severity : enhancement
1415 Description: Export bytes_read/bytes_write count on OSC/OST.
1419 Description: Early reply size mismatch, MGC loses connection
1420 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1421 the connect flags are properly negotiated.
1426 Description: quotas are not honored with O_DIRECT
1427 Details : all writes with the flag O_DIRECT will use grants which leads to
1428 this problem. Now using OBD_BRW_SYNC to guard this.
1432 Description: add quota statistics
1433 Details : 1. sort out quota proc entries and proc code.
1434 2. add quota statistics
1436 Severity : enhancement
1438 Description: enable quota support for HEAD.
1442 Description: Properly propagate oinfo flags from lov to osc for statfs
1443 Details : restore missing copy oi_flags to lov requests.
1445 Severity : enhancement
1447 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1448 : l_getgroups(8), lst(8), routerstat(8)
1449 Details : included man pages for llobdstat(8), llstat(8),
1450 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1452 Severity : enhancement
1454 Description: configure's --enable-quota should check the
1455 : kernel .config for CONFIG_QUOTA
1456 Details : configure is terminated if --enable-quota is passed but
1457 : no quota support is in kernel
1461 Description: Remove portals compatibility
1462 Details : Remove portals compatibility, not interoperable with releases
1467 Description: Resolve device initialization race
1468 Details : Prevent proc handler from accessing devices added to the
1469 obd_devs array but yet be intialized.
1471 Severity : enhancement
1473 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1475 Severity : enhancement
1477 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1482 Description: Using wrong pointer in osc_brw_prep_request
1483 Details : Access to array[-1] can produce panic if kernel compiled with
1484 CONFIG_PAGE_ALLOC enabled
1488 Description: Client runs out of low memory
1489 Details : Consider only lowmem when counting initial number of llap pages
1493 Description: *optional* service tags registration
1494 Details : if the "service tags" package is installed on a Lustre node
1495 When the filesystem is mounted, a local-node service tag will
1496 be created. See http://inventory.sun.com/ for more information
1497 about the Service Tags asset management system.
1501 Description: Kernel BUG tries to release flock
1502 Details : Lustre does not destroy flock lock before last reference goes
1503 away. So always drop flock locks when client is evicted and
1504 perform unlock regardless of successfulness of speaking to MDS.
1508 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1512 Description: sanity test 65a fails if stripecount of -1 is set
1513 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1518 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1519 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1522 Severity : enhancement
1524 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1528 Description: do not process already freed flock
1529 Details : flock can possibly be freed by another thread before it reaches
1530 to ldlm_flock_completion_ast.
1534 Description: LBUG during stress test
1535 Details : Need properly lock accesses the flock deadlock detection list.
1539 Description: oops in page fault handler
1540 Details : kernel page fault handler can return two special 'pages' in error
1541 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1545 Description: timeout with invalidate import.
1546 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1547 should be handled by ptlrpcd. This produce long age waiting and
1548 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1550 Severity : enhancement
1552 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1556 Description: enable MGS and MDT services start separately
1557 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1558 a co-located MGS without starting the MGS, which is a complement
1559 to 'nosvc' mount option.
1563 Frequency : after recovery
1564 Description: precreate to many object's after del orphan.
1565 Details : del orphan st in oscc last_id == next_id and this triger growing
1566 count of precreated objects. Set flag LOW to skip increase count
1567 of precreated objects.
1571 Frequency : rare, on clear nid stats
1572 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1573 Details : when clean nid stats sometimes try destroy live entry,
1574 and this produce panic in free.
1578 Description: Stack overflow during MDS log replay
1579 ease stack pressure by using a thread dealing llog_process.
1583 Description: wait until IO finished before start new when do lock cancel.
1584 Details : VM protocol want old IO finished before start new, in this case
1585 need wait until PG_writeback is cleared until check dirty flag and
1586 call writepages in lock cancel callback.
1588 Severity : enhancement
1590 Description: using special macro for print time and cleanup in includes.
1594 Description: mds_mfd_close() ASSERTION(rc == 0)
1595 Details : In mds_mfd_close(), we need protect inode's writecount change
1596 within its orphan write semaphore to prevent possible races.
1600 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1604 Frequency : rare, on shutdown ost
1605 Description: don't hit live lock with umount ost.
1606 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1607 use shrink_dcache_sb instead.
1611 Description: don't panic with use echo client
1612 Details : echo client pass NULL as client nid pointer and this produce null
1613 pointer dereference.
1617 Description: fix build on ppc32
1618 Details : compile code with -m64 flag produce wrong object file for ppc32.
1622 Description: add message levels for liblustreapi
1626 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1627 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1628 will result in an -EAGAIN and in the exit call path we call
1629 alloc_rr() which will with increasing aggressiveness attempt to
1630 aquire precreated objects on the minimum number of required OSCs.
1634 Description: Use old size assignment to avoid deadlock
1635 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1636 the scheduling while holding a spinlock. We do not need locking
1637 for size in ll_update_inode() because size is only updated from
1638 the MDS for directories or files without objects, so there is no
1639 other place to do the update, and concurrent access to such inodes
1640 are protected by the inode lock.
1644 Description: resolve "_IOWR redefined" build error on SLES10
1648 Description: dump the memory debugging after all modules are unloaded to
1649 suppress false negative in conf_sanity test 39
1651 Severity : enhancement
1653 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1658 Description: fix SLES kernel versioning
1659 Details : the kernel version for our SLES 10 kernel did not include a "-"
1660 before the "smp" at the end. while this was not a problem in
1661 general, it did mean that software trying to use the kernel
1662 version to try to detect a vendor specific kernel would fail.
1663 this was most evident by the OFED build scripts.
1667 Description: Don't update lov_desc members until making sure they are valid
1668 Details : When updating lov_desc members via proc fs, need fix their
1669 validities before doing the real update.
1673 Description: don't put request into delay list while invalidate in flight.
1674 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1675 import in flight. this produce timeout for invalidate and sometimes
1676 can cause stale data.
1680 Frequency : on ppc only
1681 Description: not convert ost objects for directory because it's not exist.
1682 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1686 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1688 Severity : enhancement
1690 Description: Update to support 2.6.22.14 vanilla kernel.
1694 Frequency : rare, on recovery
1695 Description: read procfs can produce deadlock in some situation
1696 Details : Holding lprocfs lock which send rpc can produce block for destroy
1697 obd objects and this also block reconnect with -EALREADY. This
1698 isn't fix all lprocfs bugs - but make it rare.
1700 Severity : enhancement
1702 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1705 Frequency : frequent on X2 node
1707 Description: mdc_set_open_replay_data LBUG
1708 Details : Set replay data for requests that are eligible for replay.
1712 Description: lustre_mgs: operation 101 on unconnected MGS
1713 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1714 MGC, and late on MGC cannot successfully connect to MGS and a lot
1715 of the error messages complaining that MGS is not connected.
1719 Frequency : on network error
1720 Description: panic with double free request if network error
1721 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1722 but it's true only for synchronus enqueue, for async enqueue
1723 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1726 Severity : enhancement
1728 Description: client-side metadata stat-ahead during readdir(directory readahead)
1729 Details : perform client-side metadata stat-ahead when the client detects
1730 readdir and sequential stat of dir entries therein
1733 Frequency : on start mds
1735 Description: Implement get_info(last_id) in obdfilter.
1738 Frequency : occasional
1740 Description: Correctly check stale fid, not start epoch if ost not support SOM
1741 Details : open with flag O_CREATE need set old fid in op_fid3 because
1742 op_fid2 was overwritten with new generated fid, but mds can answer
1743 with one of these two fids and both is not stale. Setattr
1744 incorrectly started an epoch and assume will be called
1745 done_writing, but without SOM done_writing ever being called.
1748 Frequency : rare, depends on device drivers and load
1750 Description: MDS or OSS nodes crash due to stack overflow
1751 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1752 In some cases, in conjunction with device drivers that use a lot
1753 of stack the MDS (or possibly OSS) service threads could overflow
1754 the stack. One change which was identified to consume additional
1755 stack has been reworked to avoid the extra stack usage.
1758 Frequency : occasional
1760 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1761 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1762 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1763 oscc_has_objects and see OSC has no objests and call
1764 oscc_internal_create to resend create request.
1766 Severity : enhancement
1768 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1770 Severity : enhancement
1772 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1775 Frequency : very rare
1777 Description: Fix replay if there is an un-replied request and open
1778 Details : In some cases, older replay request will revert the
1779 mcd->mcd_last_xid on MDS which is used to record the client's
1780 latest sent request.
1782 Severity : enhancement
1784 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1786 Severity : enhancement
1788 Description: Add rhel5 support to HEAD.
1790 Serverity : enhancement
1792 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1797 Description: Don't allow skipping OSTs if index has been specified.
1798 Details : Don't allow skipping OSTs if index has been specified, make
1799 locking in internal create lots better.
1803 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1804 Details : access to bitfield in structure is always rounded to long
1805 and this produce problem with not atomic change any bit.
1809 Description: Lustre make rpms failed.
1810 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1811 builds Lustre rpms from tarball.
1814 Frequency : rare on shutdown ost
1816 Description: If llog cancel was not send before clean_exports phase, this can
1817 produce deadlock in llog code.
1818 Details : If llog thread has last reference to obd and call class_import_put
1819 this produce deadlock because llog_cleanup_commit_master wait when
1820 last llog_commit_thread exited, but this never success because was
1821 called from llog_commit_thread.
1825 Description: allow userland application know is lost one of stripes.
1826 Details : fill lvb_blocks with error code on ost and return it to
1827 application if error flag found.
1831 Description: NULL lov_tgts causing MDS oops
1832 Details : more safe checks for NULL lov_tgts for avoid oops.
1834 Severity : enhancement
1836 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1840 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1841 Details : If an OST is down the MDS will hang indefinitely in
1842 obd_statfs_rqset() waiting for the statfs data. While for
1843 MDS QOS usage of statfs, it should not stuck in waiting.
1845 Severity : enhancement
1847 Description: remote_acl support
1848 Details : Support ACL-based permission check for remote user.
1849 Support setfacl/getfacl for remote user with the utils
1850 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1851 the system "{s,g}etfacl" utils.
1853 Severity : enhancement
1855 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1857 Severity : enhancement
1859 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1863 Description: make mgs_setparam() handle fsname containing dash
1864 Details : fsname containing a dash does not work with lctl conf_param
1866 Severity : enhancement
1868 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1870 Severity : enhancement
1872 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1876 Frequency : rare, at shutdown
1877 Description: access already free / zero obd_namespace.
1878 Details : if client_disconnect_export was called without force flag set,
1879 and exist connect request in flight, this can produce access to
1880 NULL pointer (or already free pointer) when connect_interpret
1881 store ocd flags in obd_namespace.
1885 Frequency : only at startup
1886 Description: not alloc memory with spinlock held.
1887 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1888 if any spinlock held.
1890 Severity : enhancement
1892 Description: make lustre randomly fail allocating memory
1893 Details : Make lustre randomly failed allocating memory for testing purpose.
1895 Severity : enhancement
1897 Description: lost problems with lov objid file
1898 Details : Fixes some scability and access to not inited memory problems
1899 in work with lov objdid file.
1904 Description: lfs find does not continue on file error
1905 Details : Continue other files processing when a file/dir is absent.
1909 Description: Inconsistent usage of lustre_pack_reply()
1910 Details : Standardize the usage of lustre_pack_reply() such that it
1911 always generate a CERROR on failure.
1914 Frequency : occasional
1916 Description: MDS hang or stay in waiting lock
1917 Details : If client receive lock with CBPENDING flag ldlm need send lock
1918 cancel as separate rpc, to avoid situation when cancel request
1919 can't processed due all i/o threads stay in wait lock.
1923 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1927 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1931 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1935 Description: Update to RHEL4 latest kernel.
1937 Severity : enhancement
1939 Description: Build SLES10 patchless client fails
1940 Details : The configure was broken by run ./configure with
1941 --with-linux-obj=.... argument for patchless client. When the
1942 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1943 can't search header adequately. Use absolute path such as
1944 -I($LINUX)/include instead.
1948 Description: interrupt oig_wait produce painc on resend.
1949 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1950 private set, and this produce situation when rq_ptlrpcd_data not
1951 copyed to new allocated request and triggered LBUG on assert
1952 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1953 ptlrpcd set if request is changed and can be safety changed to use
1958 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1959 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1960 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1961 these macros and replace them with their difinitions instead.
1965 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1966 Details : Update which_patch & target file for SLES10 latest kernel.
1969 Description: Remove client patches from SLES 10 kernel.
1970 Details : This causes SLES 10 clients to behave as patchless clients
1971 even on a Lustre-patched (server) kernel.
1973 Severity : enhancement
1975 Description: self-adjustable client's lru lists
1976 Details : use adaptive algorithm for managing client cached locks lru
1977 lists according to current server load, other client's work
1978 pattern, memory activities, etc. Both, server and client
1979 side namespaces provide number of proc tunables for controlling
1982 Severity : enhancement
1984 Description: light-weight GSS support
1985 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1986 and reduce performance overhead.
1988 Severity : enhancement
1990 Description: Linux keyring support
1991 Details : Support using service of Linux keyring for Lustre GSS internal
1992 context refresh/cache mechanism.
1996 Description: Fix errors in lfs documentation
1997 Details : Fixes man pages
2001 Description: don't use GFP_* in generic Lustre code.
2002 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2006 Description: obdclass is limited by single OBD_ALLOC(idarray)
2007 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2011 Description: warnings with build patchless client with vanila 2.6.19 and up
2012 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2013 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2017 Frequency : Only for SLES
2018 Description: sanity_quota fail test_1
2019 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2020 file, which makes TSTID[2] inunique.
2024 Frequency : Only for SLES9
2025 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2026 Details : Update target/ChangeLog/which_patch .
2030 Frequency : Only for test_10 in sanity-quota.sh
2031 Description: a bug in quota test code
2032 Details : checking a test flag in a wrong place
2036 Frequency : Only for test_13 in sanity-quota.sh
2037 Description: a bug in quota test script
2038 Details : shouldn't check overfull allocated quota
2042 Description: jbd statistics
2043 Details : Port older jbd statistics patch for sles10
2045 Severity : enhancement
2047 Description: Add printing inode info into message about error in writepage.
2052 Description: reply_lock_interpret crash due to race with it and lock cancel.
2053 Details : Do not replay locks that are being cancelled. Do not reference
2054 locks by their address during replay, just by their handle.
2058 Frequency : When flocks are used.
2059 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2060 Details : Flock locks might destroy just granted lock if it could be merged
2061 with another existing flock, this is done in completion handler,
2062 so teach ldlm_cli_enquque_fini that this is a valid case for
2068 Description: Oops in read and write path when failing to allocate lock.
2069 Details : Check if lock allocation failed and return error back.
2073 Description: lstripe command fails for valid OST index
2074 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2075 instead of lov->desc.ld_active_tgt_count.
2079 Description: sanity.sh failed test 103
2080 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2085 Description: df doesn't work properly if diskfs blocksize != 4K
2086 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2090 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2091 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2093 Severity : enhancement
2095 Description: Adaptive timeouts
2096 Details : RPC timeouts adapt to changing server load and network
2097 conditions to reduce resend attempts and improve recovery time.
2101 Description: llapi_file_create() does not allow some changes
2102 Details : add llapi_file_open() that allows specifying the mode and
2103 open flags, and also returns an open file handle.
2107 Description: merge and cleanup kernel patches.
2108 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2112 Description: Add journal checksum support.(Kernel part)
2113 Details : The journal checksum feature adds two new flags i.e
2114 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2115 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2116 indicates that the commit block contains the checksum for
2117 the blocks described by the descriptor blocks. Now commit
2118 record can be sent to disk without waiting for descriptor
2119 blocks to be written to disk. This behavior is controlled
2120 using JBD2_FEATURE_ASYNC_COMMIT flag.
2124 Description: OSS needs mutliple precreate threads
2125 Details : Add ability to start more than one create thread per OSS.
2129 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2130 Details : execute lfs setstripe on client
2134 Description: mds_obd_create error creating tmp object
2135 Details : When the user sets quota on root, llog will be affected and can't
2136 create files and write files.
2139 Frequency : Always on ia64 patchless client, and possibly others.
2141 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2142 Details : This allows the patchless client to be loaded on architectures
2143 without this export.
2147 Description: RedHat Update kernel for RHEL5
2148 Details : Add the kernel config file for RHEL5.
2152 Description: RedHat Update kernel for RHEL5
2153 Details : Modify the kernel config file more closer RHEL5.
2157 Description: Build failure against Centos5 (RHEL5)
2158 Details : Define PAGE_SIZE when it isn't present.
2162 Description: test_7 on sanity-quota.sh can't be used on separate machines
2163 Details : add facet to handle it
2167 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2168 Details : with NFS, the anon dentry's parent was set to itself in
2169 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2170 corresponding dentry other than use rec->ur_name.
2172 Severity : enhancement
2174 Description: lfs setstripe enhancement
2175 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2179 Description: enable data checksumming by default
2180 Details : enable checksum by default, allow --disable-checksum
2181 configure option and "-o nochecksum" mount option. Checksums
2182 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2186 Description: lustre support for RHEL5
2187 Details : Add support for RHEL5.
2191 Description: Client eviction due to failover config
2192 Details : after a connection loss, the lustre client should attempt to
2193 reconnect to the last active server first before trying the
2194 other potential connections.
2198 Description: when mds and osts use different quota unit(32bit and 64bit),
2199 quota will be released repeatly.
2200 Details : void sending multiple quota reqs to mds, which will keep the
2201 status between the reqs.
2205 Description: osts not allocated evenly to files
2206 Details : change the condition to increase offset_idx
2209 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2211 Description: Data corruption for OSTs that are formatted larger than 2TB
2213 Details : When generating the bio request for lustre file writes the
2214 sector number would overflow a temporary variable before being
2215 used for the IO. The data reads correctly from Lustre (which
2216 will overflow in a similar manner) but other file data or
2217 filesystem metadata may be corrupted in some cases.
2221 Description: Tune the kernel for good SCSI performance.
2222 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2223 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2228 Description: rewrite ext2-derived code in obdclass/uuid.c
2229 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2230 from scratch preserving functionality.
2234 Description: "lfs find -obd UUID" prints directories
2235 Details : "lfs find -obd UUID" will return all directory names instead
2236 of just file names. It is incorrect because the directories
2237 do not reside on the OSTs.
2241 Description: use i_size_read and i_size_write in 2.6 port
2242 Details : replace inode->i_size access with i_size_read/write()
2245 Frequency : when removing large files
2247 Description: scheduling issue during removal of large Lustre files
2248 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2249 It causes scheduling issues when removing large files (17TB in the
2253 Frequency : only with liblustre clients on XT3
2255 Description: evictions taking too long
2256 Details : allow llrd to evict clients directly on OSTs
2259 Frequency : only on ppc
2261 Description: /proc/fs/lustre/devices broken on ppc
2262 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2263 We should make sure the type of 'index' is loff_t and then cast
2264 explicitly as needed below. Do not assign an explicitly cast
2268 Frequency : only for rhel5
2270 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2271 Details : Modify the target file & which_kernel.
2273 Severity : enhancement
2275 Description: omit set fsid for export NFS
2276 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2277 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2281 Frequency : Always, for kernels after 2.6.16
2282 Description: Fix warning idr_remove called for id=.. which is not allocated.
2283 Details : Last kernels save old s_dev before kill super and not allow
2284 to restore from callback - restore it before call kill_anon_super.
2288 Description: buffer overruns could theoretically occur
2289 Details : llapi_semantic_traverse() modifies the "path" argument by
2290 appending values to the end of the origin string, and a
2291 overrun may occur. Adding buffer overrun check in liblustreapi.
2295 Description: Fix error on 'ls .' at the top of the Lustre mount.
2296 Details : Don't revalidate dentry if it is a root dentry.
2300 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2301 Details : Modify vm-tunables-rhel4.patch.
2305 Description: Kernel config for 2.6.18-vanilla.
2306 Details : Modify targets/2.6-vanilla.target.in.
2307 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2308 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2309 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2310 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2314 Description: improve handling recoverable errors
2315 Details : if request processig with error which can be recoverable on server
2316 request should be resend, otherwise page released from cache and
2321 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2322 Details : Modify target file & which_patch.
2323 A flaw was found in the IA32 system call emulation provided
2324 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2325 value could be stored in the %RAX register, which could trigger an
2326 out-of-bounds system call table access. An untrusted local user
2327 could exploit this flaw to run code in the kernel
2328 (ie a root privilege escalation). (CVE-2007-4573).
2332 Description: O_DIRECT bypasses client statistics.
2333 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2334 read_ahead_stats were updated. copied the stats section from
2335 osc_send_oap_rpc() into async_internal().
2339 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2343 Description: change order of libsysio includes
2344 Details : '#include sysio.h' should always come before '#include xtio.h'
2348 Description: handle "serious error: objid * already exists" more gracefully
2349 Details : If LAST_ID value on disk is smaller than the objects existing in
2350 the O/0/d* directories, it indicates disk corruption and causes an
2351 LBUG(). If the object is 0-length, then we should use the existing
2352 object. This will help to avoid a full fsck in most cases.
2354 Severity : enhancement
2356 Description: adapt the lustre_config script to support the upgrade case
2357 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2358 server targets to 1.6 in parallel.
2362 Description: To avoid grant space > avaible space when the disk is almost
2363 full. Without this patch you might see the error "grant XXXX >
2364 available" or some LBUG about grant, when the disk is almost
2366 Details : In filter_check_grant, for non_grant cache write, we should
2367 check the left space by if (*left > ungranted + bytes), instead
2368 of (*left > ungranted), because only we are sure the left space
2369 is enough for another "bytes", then the ungrant space should be
2370 increase. In client, we should update cl_avail_grant only there
2371 is OBD_MD_FLGRANT in the reply.
2375 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2376 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2377 described in CVE-2007-4573. This problem could allow untrusted
2378 local users to gain root access.
2381 Frequency : when using O_DIRECT and quotas
2383 Description: Incorrect file ownership on O_DIRECT output files
2384 Details : block usage reported by 'lfs quota' does not take into account
2385 files that have been written with O_DIRECT.
2390 Description: touch file failed when fs is not full
2391 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2392 otherwise we can get ENOSP while fs is not full.
2396 Description: parallel lock callbacks
2397 Details : Instead of sending blocking and completion callbacks as separated
2398 requests, adding them to a set and sending in parallel.
2401 Frequency : only for Cray XT3
2402 Bugzilla : 12829/13455
2403 Description: Changing primary group doesn't change the group lustre assigns to
2405 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2406 overridden with the primary group provided by the group upcall,
2407 whereas the supplied fsgid can be trusted if it is in the list of
2408 supplementary groups returned by the group upcall.
2410 Severity : enhancement
2412 Description: Allow masking D_WARNING, D_ERROR messages from console
2413 Details : Console messages can now be disabled via lnet.printk.
2417 Description: User code with malformed file open parameter crashes client node
2418 Details : Before packing join_file req, all the related reference should be
2419 checked carefully in case some malformed flags cause fake
2420 join_file req on client.
2424 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2425 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2426 mode is equal to the granted mode. However, it should also check
2427 that LDLM_FL_CP_REQD is not set.
2431 Description: Heavy nfs access might result in deadlocks
2432 Details : After ELC code landed, it is now improper to enqueue any mds
2433 locks under och_sem, because enqueue might want to decide to
2434 cancel open locks for same inode we are holding och_sem for.
2438 Description: Client eviction while running blogbench
2439 Details : A lot of unlink operations with concurrent I/O can lead to a
2440 deadlock causing evictions. To address the problem, the number of
2441 oustanding OST_DESTROY requests is now throttled to
2442 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2443 callbacks are processed in priority.
2447 Description: enable ACLs on MDS by default
2448 Details : ACLs must be enabled on MDS by default.
2451 Frequency : PPC/PPC64 only
2453 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2454 Details : fix duplicated definitions between asm-ppc64/types.h and
2455 lustre_types.h on PPC.
2458 Frequency : PPC/PPC64 only
2460 Description: asm-ppc/segment.h does not exist
2461 Details : fix compile issue on PPC.
2465 Description: better handle error messages in extents code
2468 Frequency : RHEL4 only
2470 Description: mkfs is very slow on IA64/RHEL4
2471 Details : A performance regression has been discovered in the MPT Fusion
2472 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2473 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2474 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2476 Severity : enhancement
2478 Description: SNMP support enhancement
2479 Details : Adding total number of sampled request for an MDS node in snmp
2482 Severity : enhancement
2484 Description: Optimize ldlm waiting list processing for PR extent locks
2485 Details : When processing waiting list for read extent lock and meeting read
2486 lock that is same or wider to it that is not contended, skip
2487 processing rest of the list and immediatelly return current
2488 status of conflictness, since we are guaranteed there are no
2489 conflicting locks in the rest of the list.
2493 Description: Time out and refuse to reconnect
2494 Details : When the failover node is the primary node, it is possible
2495 to have two identical connections in imp_conn_list. We must
2496 compare not conn's pointers but NIDs, otherwise we can defeat
2497 connection throttling.
2501 Description: port llog fixes from b1_6 into HEAD
2502 Details : Port llog reference couting and some llog cleanups from b1_6
2503 (bug 10800) into HEAD, for protect from panic and access to already
2504 free llog structures.
2508 Description: Detect stride IO mode in read-ahead
2509 Details : When a client does stride read, read-ahead should detect that and
2510 read-ahead pages according to the detected stride pattern.
2514 Description: data checksumming impacts single node performance
2515 Details : add support for several checksum algorithm. Currently, only CRC32
2516 and Adler-32 are supported. The checksum type can be changed on
2517 the fly via /proc/fs/lustre/osc/*/checksum_type.
2521 Description: use adler32 for page checksums
2522 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2527 Description: build for x2 fails
2528 Details : fix compile issue on Cray systems.
2532 Description: Properly match for duplicate locks
2533 Details : Due to different lock order from skiplists code, we need to
2534 traverse entire list for now
2537 Frequency : only on PPC/SLES10
2539 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2540 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2541 be defined. Add a hack in mkfs_lustre.c to work around this compile
2546 Description: LASSERT on MDS when client holding flock lock dies
2547 Details : ldlm pool logic depends on number of granted locks equal to
2548 number of released locks which is not true for flock locks, so
2549 just exclude such locks from consideration.
2553 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2554 Details : Use fsfilt_sync() for both the whole filesystem sync and
2555 individual file sync to eliminate dangerous inode locking
2556 with I_LOCK that can lead to a deadlock.
2560 Description: performance in 1.6.3
2561 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2562 sure that 1MB requests are not fragmented by the block layer.
2564 Severity : enhancement
2566 Description: organize the server-side client stats on per-nid basis
2567 Details : Change the structure of stats under obdfilter and mds to
2575 The "uuid"s file would list the uuids of _active_ exports.
2576 And the clear entry is to clear all stats and stale nids.
2578 Severity : enhancement
2580 Description: eliminate client locks in face of contention
2581 Details : file contention detection and lockless i/o implementation
2582 for contended files.
2586 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2589 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2591 Description: MDS may lose file striping (and hence file data) in some cases
2592 Details : If there are additional extended attributes stored on the MDS,
2593 in particular ACLs, SELinux, or user attributes (if user_xattr
2594 is specified for the client mount options) then there is a risk
2595 of attribute loss. Additionally, the Lustre file striping
2596 needs to be larger than default (e.g. striped over all OSTs),
2597 and an additional attribute must be stored initially in the
2598 inode and then increase in size enough to be moved to the
2599 external attribute block (e.g. ACL growing in size) for file
2604 Description: skiplist implementation simplification
2605 Details : skiplists are used to group compatible locks on granted list
2606 that was implemented as tracking first and last lock of each
2607 lock group the patch changes that to using doubly linked lists
2611 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2612 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2613 because LASSERT on both the data supplied by a client, and the
2614 data on disk is dangerous and incorrect.
2616 Severity : enhancement
2618 Description: Slow truncate/writes to huge files at high offsets.
2619 Details : Directly associate cached pages to lock that protect those pages,
2620 this allows us to quickly find what pages to write and remove
2621 once lock callback is received.
2625 Description: more ldlm soft lockups
2626 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2627 starve other threads from the resource lock for a long time in
2628 case of long waiting queue, so change the debug level from
2629 D_OTHER to the less frequently used D_INFO.
2631 Severity : enhancement
2633 Description: add -gid, -group, -uid, -user options to lfs find
2637 Description: Hung threads in invalidate_inode_pages2_range
2638 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2639 once one is completed. As a result, some RPCs are stuck in the
2640 queue and are never sent.
2644 Description: filter threads hungs on waiting journal commit
2645 Details : Cleanup filter group llog code, then only filter group llog will
2646 be only created in the MDS/OST syncing process.
2650 Description: Procfs and llog threads access destoryed import sometimes.
2651 Details : Sync the import destoryed process with procfs and llog threads by
2652 the import refcount and semaphore.
2654 Severity : enhancement
2656 Description: openlock cache of b1_6 port to HEAD
2661 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2662 Details : If insertion of an extent fails, then discard the inode
2663 preallocation and free data blocks else it can lead to duplicate
2668 Description: don't always update ctime in ext3_xattr_set_handle()
2669 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2670 In some cases the ctime should not be updated, for example for
2671 2.0->1.8 compatibility it is necessary to delete an xattr and it
2672 should not update the ctime.
2676 Bugzilla : 15713/16362
2677 Description: Assertion in iopen_connect_dentry in 1.6.3
2678 Details : looking up an inode via iopen with the wrong generation number can
2679 populate the dcache with a disconneced dentry while the inode
2680 number is in the process of being reallocated. This causes an
2681 assertion failure in iopen since the inode's dentry list contains
2682 both a connected and disconnected dentry.
2686 Description: assertion failure in ldlm_handle2lock()
2687 Details : fix a race between class_handle_unhash() and class_handle2object()
2688 introduced in lustre 1.6.5 by bug 13622.
2693 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2694 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2699 Description: oops in sd_iostats_seq_show()
2700 Details : unloading/reloading the scsi low level driver triggers a kernel
2701 bug when trying to access the sd iostat file.
2706 Description: Kernel panics during QLogic driver reload
2707 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2708 patch, causing memory corruption.
2713 Description: journal_dev option does not work in b1_6
2714 Details : pass mount option during pre-mount.
2716 Severity : enhancement
2718 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2719 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2720 extent information of a file. It can be used to map logical blocks
2721 in a file to physical blocks in the block device.
2725 Description: LDLM soft lockups - improvement
2726 Details : It is be possible to send the lock handle along with each read
2727 or write request because the client is already doing a lock match
2728 itself so there isn't any reason the OST should have to re-do that
2732 Frequency : only on Cray X2
2734 Description: X2 build failures
2735 Details : fix build failures on Cray X2.
2739 Description: xid & resent requests
2740 Details : Initialize RPC XID from clock at startup (randomly if clock is
2743 Severity : enhancement
2745 Description: Add lustre_start utility to start or stop multiple Lustre servers
2750 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2751 Details : In case of memory pressure, list_del() can be called twice on
2752 req->rq_history_list, causing a kernel oops.
2756 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2757 Details : fix stack overflow in the distributed lock manager by defering
2758 export eviction after a failed AST to the elt thread instead of
2759 handling it in the dlm interpret routine.
2763 Description: Convert some comments to new format.
2764 Details : Update documenting comments to match doxygen conventions.
2768 Description: Grammar fixes.
2769 Details : A couple of trivial spelling fixes.
2773 Description: OSD_COUNTERS-mandatory
2774 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2775 to simplify the code.
2779 Description: simplify cmm_device freeing logic.
2780 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2784 Description: Add lockdep support to dt_object_operations locking interface.
2785 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2786 parameter indicating lock ordering. Update mdd code to use new
2791 Description: Introduce failloc constants for lockless IO tests.
2792 Details : Add two new failloc constants to test lockless IO. Only one of
2793 them in implemented---another is checked in yet to be landed
2798 Description: Add lockdep support for inode mutex.
2799 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2800 in the situations where Lustre has to lock more than one inode
2805 Description: Add optional invariants checking support.
2806 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2807 configure switch. This macro is to be used for consistency and
2808 sanity checks that are too expensive to be left in `production'
2813 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2814 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2819 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2820 Details : Use spin_lock_nested() in (the only) situation where more than
2821 one ldlm_lock is locked simultaneously. Also, fix possible
2822 dead-lock in ldlm_lock_change_resource() by enforcing particular
2827 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2828 Details : Instead of specifying each ldlm_lock call-back through separate
2829 parameter, wrap them into struct ldlm_callback_suite.
2833 Description: Kill join_lru obd method and its callers.
2834 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2835 in memory---a requirement implemented through ->o_join_lru() obd
2836 method in HEAD. Remove this method and its users.
2840 Description: Add asynchronous ldlm ENQUEUE completion handler.
2841 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2842 case that stock ldlm_completion_ast() cannot handle as it waits
2843 until lock is granted. Introduce new ldlm_completion_ast_async()
2844 for this. Also comment ldlm_completion_ast().
2848 Description: ldlm_error <-> errno conversion.
2849 Details : Add functions to map (rather arbitrary) between LDLM error codes
2850 and standard errno values. CLIO needs this to prevent LDLM specific
2851 constants from escaping ldlm and osc.
2855 Description: Kill unused ldlm_handle2lock_ns() function.
2856 Details : Kill unused ldlm_handle2lock_ns() function.
2860 Description: Add lu_ref support to ldlm_lock
2861 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2862 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2863 and ldlm_resource. LDLM interface has to be changed, because code that
2864 releases a reference on a lock, has to "know" what reference this is.
2865 In the most frequent case
2867 lock = ldlm_handle2lock(handle);
2869 LDLM_LOCK_PUT(lock);
2871 no changes are required. When any other reference (received _not_ from
2872 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2873 instead of LDLM_LOCK_PUT().
2875 Arguably, changes are pervasive, and interface requires some discipline
2876 for proper use. On the other hand, it was very instrumental in finding
2877 a few leaked lock references.
2881 Description: Add ldlm_lock_addref_try().
2882 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2883 attempts to addref a lock that might be being canceled
2888 Description: Add ldlm_weigh_callback().
2889 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2890 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2891 hard-coded `number of pages' logic.
2895 Description: Add lockdep annotations to llog code.
2896 Details : Use appropriately tagged _nested() locking calls in the places
2897 where llog takes more than one ->lgh_lock lock.
2901 Description: Add loi_kms_set().
2902 Details : Wrap kms updates into a helper function.
2906 Description: Constify instances of struct lsm_operations.
2907 Details : Constify instances of struct lsm_operations.
2911 Description: lu_conf support.
2912 Details : On a server, a file system object is uniquely identified
2913 by a fid, which is sufficient to locate and load all object
2914 state (inode). On a client, on the other hand, more data are
2915 necessary instantiate an object. Change lu_object_find() and
2916 friends to take additional `lu_conf' argument describing object.
2917 Typically this includes layout information.
2921 Description: lu_context fixes.
2922 Details : Introduce new lu_context functions that are needed on the client
2923 side, where some system threads (ptlrpcd) are shared by multiple
2924 modules, and so cannot be stopped during module shutdown.
2928 Description: Add start and stop methods to lu_device_type_operations.
2929 Details : Introduce two new methods in lu_device_type_operations, that are
2930 invoked when first instance of a given type is created and last one
2931 is destroyed respectively. This is need by CLIO.
2935 Description: Add lu_ref support to struct lu_device.
2936 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2937 track leaked references.
2941 Description: Introduce lu_kmem_descr.
2942 Details : lu_kmem_descr and its companion interface allow to create
2943 and destroy a number of kmem caches at once.
2947 Description: Fix lu_object finalization race.
2948 Details : Fix a race between lu_object_find() finding an object and its
2949 concurrent finalization. This race is (most likely) not possible
2950 on the server, but might happen on the client.
2954 Description: Introduce lu_ref interface.
2955 Details : lu_ref is a debugging module allowing to track references to
2956 a given object. It is quite cpu expensive, and has to be
2957 explicitly enabled with --enable-lu_ref. See usage description
2962 Description: Factor lu_site procfs stats into a separate function.
2963 Details : Separate lu_site stats printing code into a separate function
2964 that can be reused on a client.
2968 Description: Constify instances of struct {lu,dt,md}_device_operations.
2969 Details : Constify instances of struct {lu,dt,md}_device_operations.
2973 Description: Introduce struct md_site and move meta-data specific parts of
2974 struct lu_site here.
2975 Details : Move md-specific fields out of struct lu_site into special struct
2976 md_site, so that lu_site can be used on a client.
2980 Description: Kill mdd_lov_destroy().
2981 Details : Remove unused mdd code.
2985 Description: Add st_block checking to multistat.c.
2986 Details : Add st_block checking to multistat.c.
2990 Description: Add lu_ref support to struct obd_device.
2991 Details : Add lu_ref tracking to obd_device.
2995 Description: Kill obd_set_fail_loc().
2996 Details : Remove unused code.
3000 Description: Add special type for ptlrpc_request interpret functions.
3001 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
3002 there. Actual usage will be in CLIO.
3006 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
3007 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
3008 doesn't work with lockdep.
3012 Description: Add rwv.c test program.
3013 Details : New testing program exercising readv(2) and writev(2) (Qian).
3017 Description: Add sendfile.c test program.
3018 Details : New testing program exercising sendfile(2) (Jay).
3022 Description: Ratelimit a message that can be very frequent.
3023 Details : Ratelimit a memory allocation failure message that can
3028 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3029 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3033 Description: Liblustre build fixes.
3034 Details : Liblustre build fixes.
3038 Description: libcfs: add cfs_{need,cond}_resched() interface.
3039 Details : libcfs: add cfs_{need,cond}_resched() definition and
3040 implementations for Linux, NT, and liblustre.
3042 Severity : enhancement
3044 Description: More exported tunables for mballoc
3045 Details : Add support for tunable preallocation window and new tunables for
3046 large/small requests
3050 Description: Detect corruption of block bitmap and checking for preallocations
3051 Details : Checks validity of on-disk block bitmap. Also it does better
3052 checking of number of applied preallocations. When corruption is
3053 found, it turns filesystem readonly to prevent further corruptions.
3057 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3058 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3059 window to 0. If it does hit cache miss, and read-pattern is still
3060 stride-io mode, does not reset the stride window, but also does not
3061 increase the stride window length in this case.
3065 Frequency : only for big-endian servers
3066 Description: Check if system is big-endian while mounting fs with extents feature
3067 Details : Mounting a filesystem with extents feature will fail on big-endian
3068 systems since ext3-based ldiskfs is not supported on big-endian
3069 systems. This can be over-riden with "bigendian_extents" mount option.
3071 Severity : enhancement
3073 Description: The root squash functionality
3074 Details : A security feature, which is to prevent users from being able
3075 to mount lustre on their desktop, run as root, and delete
3076 all of the files in the filesystem. The goal is accomplished by
3077 remapping user id (UID) and group id (GID) of the root user to
3078 a UID and GID specified by the system administartor via Lustre
3079 configuration management server (MGS). The functionality also
3080 allows to specify sets of clients for which the remapping does
3085 Description: Excessive recovery window
3086 Details : With AT enabled, the recovery window can be excessively long (6000+
3087 seconds). To address this problem, we no longer use
3088 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3089 timeout no longer depends on the service time, it is set to
3090 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3091 time via pb_service_time.
3095 Description: Watchdog triggered on MDS failover
3096 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3097 the OSTs know that the MDS "UUID" can be reused for the same export
3098 from a different NID, so we do not need to wait for the export to be
3102 Frequency : rare, only if using MMP with Linux RAID
3104 Description: MMP doesn't work with Linux RAID
3105 Details : While using HA for Lustre servers with Linux RAID, it is possible
3106 that MMP will not detect multiple mounts. To make this work we
3107 need to unplug the device queue in RAID when the MMP block is being
3108 written. Also while reading the MMP block, we should read it from
3109 disk and not the cached one.
3111 Severity : enhancement
3113 Description: open file using fid
3114 Details : A file can be opened using just its fid, like
3115 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3118 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3121 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3122 Details : add proper locking for creation of jbd_4k slab cache
3126 Description: MMP check in ext3_remount() fails without displaying any error
3127 Details : When multiple mount protection fails during remount, proper error
3130 Severity : enhancement
3132 Description: Allow stripe size to be up to 4G-64k
3133 Details : Fix math logic to allow large stripe sizes.
3137 Description: add check for >8TB ldiskfs filesystems
3138 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3139 Don't allow >8TB ldiskfs filesystems to be mounted without
3140 force_over_8tb mount option
3145 Descriptoin: OSS complains "too many blocks" and crashes
3146 Details : In filter_commitrw_write():
3147 1. lquota_chkquota();
3148 2. obdo_from_inode();
3149 3. lquota_pending_commit();
3150 "2." has changed uid and gid in this case, then caused
3151 lquota_chkquota() and lquota_pending_commit() to use
3152 different uid and gid. That is the root of the bug.
3156 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3157 Details : Detect and print out the 'stripe count = -1' case properly.
3161 Description: Out or order replies might be lost on replay
3162 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3163 request with tid smaller than one currently being added, add it
3164 to the start, not end of the list.
3166 --------------------------------------------------------------------------------
3168 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3170 * Support for kernels:
3171 2.6.5-7.283 (SLES 9),
3172 2.6.9-55.EL (RHEL 4),
3173 2.6.16.46-0.14 (SLES 10),
3174 2.6.18.8 vanilla (kernel.org)
3175 * Client support for unpatched kernels:
3176 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3177 2.6.16 - 2.6.22 vanilla (kernel.org)
3178 * Due to recently discovered recovery problems, we do not recommend
3179 using patchless RHEL 4 clients with this or any earlier release.
3180 * Recommended e2fsprogs version: 1.39.cfs8
3181 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3182 * Starting with this release, the ldiskfs backing filesystem required
3183 by Lustre is now in its own package, lustre-ldiskfs. This package
3184 should be installed. It is versioned separately from Lustre and
3185 may be released separately in future.
3189 Description: block reactivating mgc import until all deactivates complete
3190 Details : Fix race when failing back MDT/MGS to itself (testing)
3193 Frequency : at statup only
3195 Description: mds_lov_synchronize race leads to various problems
3196 Details : simultaneous MDT->OST connections at startup can cause the
3197 sync to abort, leaving the OSC in a bad state.
3199 Severity : enhancement
3201 Description: add optional extra BUILD_VERSION info
3202 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3203 to override the lustre version.
3206 Frequency : 2.6.18 servers only
3208 Description: ll_kern_mount() doesn't release the module reference
3209 Details : The ldiskfs module reference count never drops down to 0
3210 because ll_kern_mount() doesn't release the module reference.
3215 Description: server LBUG when using old ost_num_threads parameter
3216 Details : Accept the old ost_num_threads parameter but warn that it
3217 is deprecated, and fix an off-by-one error that caused an LBUG.
3222 Description: Transient SCSI error results in persistent IO issue
3223 Details : iobuf->dr_error is not reinitialized to 0 between two
3227 Frequency : sometimes when underlying device returns I/O errors
3229 Description: OSTs not going read-only during write failures
3230 Details : OSTs are not remounted read-only when the journal commit threads
3231 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3232 instead of the ext3 wrappers.
3236 Description: poor connect scaling with increasing client count
3237 Details : Don't run filter_grant_sanity_check for more than 100 exports
3238 to improve scaling for large numbers of clients.
3241 Frequency : SLES10 only
3243 Description: sanity-quota.sh quotacheck failed: rc = -22
3244 Details : Quotas cannot be enabled on SLES10.
3247 Frequency : liblustre clients only
3249 Description: getdirentries does not give error when run on compute nodes
3250 Details : getdirentries does not fail when the size specified as an argument
3251 is too small to contain at least one entry
3253 Severity : enhancement
3255 Description: Add LNET router traceability for debug purposes
3256 Details : If a checksum failure occurs with a router as part of the
3257 IO path, the NID of the last router that forwarded the bulk data
3258 is printed so it can be identified.
3263 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3264 Details : Due to a race condition, liblustre clients were occasionally
3265 evicted incorrectly.
3267 Severity : enhancement
3269 Description: lfs setstripe use optional parameters instead of postional
3272 Severity : enhancement
3274 Description: Nanosecond timestamp support for ldiskfs
3275 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3276 resolution timestamps. There is not yet support for this at
3277 the Lustre filesystem level.
3280 Frequency : during server recovery
3282 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3283 Details : request with rq_no_resend flag not awake l_wait_event if they get
3287 Frequency : nfs export on patchless client
3289 Description: connectathon hang when test nfs export over patchless client
3290 Details : Disconnected dentry cannot be found with lookup, so we do not need
3291 to unhash it or make it invalid
3294 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3295 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3296 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3301 Description: open req refcounting wrong on reconnect
3302 Details : If reconnect happened between getting open reply from server and
3303 call to mdc_set_replay_data in ll_file_open, we will schedule
3304 replay for unreferenced request that we are about to free.
3305 Subsequent close will crash in variety of ways.
3306 Check that request is still eligible for replay in
3307 mdc_set_replay_data().
3312 Description: disable writes to filesystem when reading health_check file
3313 Details : the default for reading the health_check proc file has changed
3314 to NOT do a journal transaction and write to disk, because this
3315 can cause reads of the /proc file to hang and block HA state
3316 checking on a healthy but otherwise heavily loaded system. It
3317 is possible to return to the previous behaviour during configure
3318 with --enable-health-write.
3320 Severity : enhancement
3322 Description: 64-bit inode version
3323 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3324 to the inode. This will be required for version-based recovery.
3329 Description: MDS fails to start if a duplicate client export is detected
3330 Details : in some rare cases it was possible for a client to connect to
3331 an MDS multiple times. Upon recovery the MDS would detect this
3332 and fail during startup. Handle this more gracefully.
3334 Severity : enhancement
3336 Description: Add -o localflock option to simulate old noflock
3338 Details : This will achieve local-only flock/fcntl locks
3344 Description: log_commit_thread vs filter_destroy race leads to crash
3345 Details : Take import reference before releasing llog record semaphore
3350 Description: Wrong request locking in request set processing
3351 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3352 imp_delayed_list, in this place should be used imp_lock.
3355 Frequency : when reconnection
3357 Description: Grant Leak when osc reconnect to OST
3358 Details : When osc reconnect ost, OST(filter) should check whether it
3359 should grant more space to client by comparing fed_grant and
3360 cl_avail_grant, and return the granted space to client instead
3361 of "new granted" space, because client will call osc_init_grant
3362 to update the client grant space info.
3365 Frequency : when client reconnect to OST
3367 Description: Grant Leak when osc do resend and replay bulk write
3368 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3369 bulk write request, because the grant info while be sync between
3370 OSC and OST when reconnect, and we should ignore the grant info
3371 these of resend/replay write req.
3376 Description: Grant space more than available left space sometimes.
3377 Details : When then OST is about to be full, if two bulk writing from
3378 different clients came to OST. Accord the avaliable space of the
3379 OST, the first req should be permitted, and the second one
3380 should be denied by ENOSPC. But if the seconde arrived before
3381 the first one is committed. The OST might wrongly permit second
3382 writing, which will cause grant space > available space.
3385 Frequency : when client is evicted
3387 Description: Grant might be wrongly erased when osc is evicted by OST
3388 Details : when the import is evicted by server, it will fork another
3389 thread ptlrpc_invalidate_import_thread to invalidate the
3390 import, where the grant will be set to 0. While the original
3391 thread will update the grant it got when connecting. So if
3392 the former happened latter, the grant will be wrongly errased
3393 because of this race.
3398 Description: Checking Stale with correct fid
3399 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3400 to check whether it is stale, because sometimes, we want the
3401 enqueue happened anyway, and op_data.fid2 will not be initialized.
3403 Severity : enhancement
3405 Description: update patchless client
3406 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3409 Frequency : only with 2.4 kernel
3411 Description: random memory corruption
3412 Details : size of struct ll_inode_info is to big for union inode.u and this
3413 can be cause of random memory corruption.
3418 Description: Memory leak in recovery
3419 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3420 It should also check obd_fail before fsfilt_start, otherwise if
3421 fsfilt_start return -EROFS,(failover mds during mds recovery).
3422 then the req will return with repmsg->transno = 0 and rc = EROFS.
3423 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3424 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3425 be freed no matter whether fsfilt_commit success or not.
3428 Frequency : only with huge count clients
3430 Description: Prevents from taking the superblock lock in llap_from_page for
3432 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3433 for llap_from_page call in ll_removepage prevents from taking the
3434 superblock lock for a soon died page.
3439 Description: Not check open intent error before release open handle
3440 Details : in some rare cases, the open intent error is not checked before
3441 release open handle, which may cause
3442 ASSERTION(open_req->rq_transno != 0), because it tries to release
3443 the failed open handle.
3448 Description: Set cat log bitmap only after create log success.
3449 Details : in some rare cases, the cat log bitmap is set too early. and it
3450 should be set only after create log success.
3454 Description: Accessing a block bevice can re-enable I/O when Lustre is
3455 tearing down a device.
3456 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3461 Description: service threads may hog cpus when there are a lot of requests
3462 Details : Insert cond_resched to give other threads a chance to use some CPU
3467 Description: the cat log was not initialized in recovery
3468 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3469 unlink log on mds will not be initialized until mds post
3470 recovery. And also in mds post recovery, the unlink log will
3471 initialization will be done asynchronausly, so there will be race
3472 between add unlink log and unlink log initialization.
3476 Description: brw_stats were being printed incorrectly
3477 Details : brw_stats were being printed as log2 but all of them were not
3478 recorded as log2. Also remove some code duplication arising from
3479 filter_tally_{read,write}.
3483 Frequency : rare, only in recovery.
3484 Description: ASSERTION(req->rq_type != LI_POISON) failed
3485 Details : imp_lock should be held while iterating over imp_sending_list for
3486 prevent destroy request after get timeout in ptlrpc_queue_wait.
3490 Description: replay-single.sh test 52 fails
3491 Details : A lock's skiplist need to be cleanup when it being unlinked
3492 from its resource list.
3496 Description: Short directio read returns full requested size rather than
3498 Details : Direct I/O operations should return actual amount of bytes
3499 transferred rather than requested size.
3501 Severity : enhancement
3503 Description: metadata RPC reduction (e.g. for rm performance)
3504 Details : decrease the amount of synchronous RPC between clients and servers
3505 by canceling conflicing lock before the operation on the client
3506 and packing thier handles into the main operation RPC to server.
3508 Severity : enhancement
3510 Description: Async OSC create to avoid the blocking unnecessarily.
3511 Details : If an OST has no remaining object, system will block on the
3512 creation when it needs to create a new object on this OST. Now,
3513 ways use pre-created objects when available, instead of blocking on
3514 an empty osc while others are not empty. If we must block, we
3515 block for the shortest possible period of time.
3519 Description: improve handling recoverable errors
3520 Details : if request processig with error which can be recoverable on server
3521 request should be resend, otherwise page released from cache and
3524 Severity : enhancement
3526 Description: refine locking for avoid write wrong info into lov_objid file
3527 Details : fix possible races with add new target and write/update data in
3530 --------------------------------------------------------------------------------
3532 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3537 Frequency : on some architectures
3539 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3540 Details : Uninitialized flags sometimes cause configuration commands to
3544 Frequency : patchless clients only
3546 Description: missing __iget() symbol export
3547 Details : The __iget() symbol export is missing. To avoid the need for
3548 this on patchless clients the deathrow inode reaper is turned
3549 off, and we depend on the VM to clean up old inodes. This
3550 dependency was during via the fix for bug 12181.
3552 --------------------------------------------------------------------------------
3554 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3556 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3557 INTEROPERATE with older versions automatically. In many cases a
3558 special upgrade step is needed. Please read the
3559 user documentation before upgrading any part of a live system.
3560 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3561 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3562 * WARNING: Lustre configuration and startup changes are required with
3563 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3567 Severity : enhancement
3569 Description: MountConf
3570 Details : Lustre configuration is now managed via mkfs and mount
3571 commands instead of lmc and lconf. New obd types (MGS, MGC)
3572 are added for dynamic configuration management. See
3573 https://mail.clusterfs.com/wikis/lustre/MountConf for
3576 Severity : enhancement
3578 Description: dynamic OST addition
3579 Details : OSTs can now be added to a live filesystem
3581 Severity : enhancement
3583 Description: startup order invariance
3584 Details : MDTs and OSTs can be started in any order. Clients only
3585 require the MDT to complete startup.
3587 Severity : enhancement
3589 Description: parallel, asynchronous orphan cleanup
3590 Details : orphan cleanup is now performed in separate threads for each
3591 OST, allowing parallel non-blocking operation.
3593 Severity : enhancement
3595 Description: optimized stripe assignment
3596 Details : stripe assignments are now made based on ost space available,
3597 ost previous usage, and OSS previous usage, in order to try
3598 to optimize storage space and networking resources.
3600 Severity : enhancement
3602 Description: Permanently set tunables
3603 Details : All writable /proc/fs/lustre tunables can now be permanently
3604 set on a per-server basis, at mkfs time or on a live system.
3606 Severity : enhancement
3608 Description: Lustre message v2
3609 Details : Add lustre message format v2.
3611 Severity : enhancement
3613 Description: client OST exclusion list
3614 Details : Clients can be started with a list of OSTs that should be
3615 declared "inactive" for known non-responsive OSTs.
3619 Description: SPEC SFS validation failure on NFS v2 over lustre.
3620 Details : Changes the blocksize for regular files to be 2x RPC size,
3621 and not depend on stripe size.
3623 Severity : enhancement
3625 Description: Multiple MD RPCs in flight.
3626 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3627 To avoid overly-overloading MDS, introduce a limit on number of
3628 MDS RPCs in flight for a single client and add /proc controls
3629 to adjust this limit.
3631 Severity : enhancement
3633 Description: client read/write statistics
3634 Details : Add client read/write call usage stats for performance
3635 analysis of user processes.
3636 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3637 file access. extents_stats shows chunk size distribution.
3638 extents_stats_per_process show chunk size distribution per
3641 Severity : enhancement
3643 Description: mds statistics
3644 Details : Add detailed mds operations statistics in
3645 /proc/fs/lustre/mds/*/stats.
3649 Description: Failure of copying files with lustre special EAs.
3650 Details : Client side always return success for setxattr call for lustre
3651 special xattr (currently only "trusted.lov").
3656 Description: Refcount LNET uuids
3657 Details : The global LNET uuid list grew linearly with every startup;
3658 refcount repeated list entries instead of always adding to
3661 Severity : enhancement
3663 Description: Dynamic service threads
3664 Details : Within a small range, start extra service threads
3665 automatically when the request queue builds up.
3668 Frequency : mixed-endian client/server environments
3670 Description: mixed-endian crashes
3671 Details : The new msg_v2 system had some failures in mixed-endian
3674 Severity : enhancement
3676 Description: Easy OST removal
3677 Details : OSTs can be permanently deactivated with e.g. 'lctl
3678 conf_param lustre-OST0001.osc.active=0'
3680 Severity : enhancement
3682 Description: MGS proc entries
3683 Details : Added basic proc entries for the MGS showing what filesystems
3686 Severity : enhancement
3688 Description: provide MGS failover
3689 Details : Added config lock reacquisition after MGS server failover.
3691 Severity : enhancement
3693 Description: add Linux 2.4 support
3694 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3698 Description: a large application tries to do I/O to the same resource and dies
3699 in the middle of it.
3700 Details : Check the req->rq_arrival time after the call to
3701 ost_brw_lock_get(), but before we do anything about
3702 processing it & sending the BULK transfer request. This
3703 should help move old stale pending locks off the queue as
3704 quickly as obd_timeout.
3707 Frequency : when an incorrect nid is specified during startup
3709 Description: ptlrpc connect to non-existant node causes kernel crash
3710 Details : LNET can't be re-entered from an event callback, which
3711 happened when we expire a message after the export has been
3712 cleaned up. Instead, hand the zombie cleanup off to another
3715 Severity : enhancement
3717 Description: plain/inodebits lock performance improvement
3718 Details : Grouping plain/inodebits in granted list by their request modes
3719 and bits policy, thus improving the performance of search through
3723 Frequency : only if OST filesystem is corrupted
3725 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3726 Details : for a short time RPCs with bulk IO are in the replay list,
3727 but replay of bulk IOs is unimplemented. If the OST filesystem
3728 is corrupted due to disk cache incoherency and then replay is
3729 started it is possible to trip an assertion. Avoid putting
3730 committed RPCs into the replay list at all to avoid this issue.
3733 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3735 Description: System hang on startup
3736 Details : This bug allowed the liblustre (e.g. catamount) client to
3737 return to the app before handling all startup RPCs. This
3738 could leave the node unresponsive to lustre network traffic
3739 and manifested as a server ptllnd timeout.
3741 Severity : enhancement
3743 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3744 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3745 boolean to control whether to print peer debug info when a
3746 client's RPC times out.
3749 Frequency : only for kernels with patches from Lustre below 1.4.3
3751 Description: Remove old rdonly API
3752 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3755 Frequency : only for devices with external journals
3757 Description: Set external device read-only also
3758 Details : During a commanded failover stop, we set the disk device
3759 read-only while the server shuts down. We now also set any
3760 external journal device read-only at the same time.
3763 Frequency : when upgrading from 1.4 while trying to change parameters
3765 Description: The wrong (new) MDC name was used when setting parameters for
3766 upgraded MDT's. Also allows changing of OSC (and MDC)
3767 parameters if --writeconf is specified at tunefs upgrade time.
3770 Frequency : when setting specific ost indicies
3772 Description: QOS code breaks on skipped indicies
3773 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3774 created with --index need not be sequential.
3776 Severity : enhancement
3778 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3779 Details : The uninit_groups feature works in conjunction with the kernel
3780 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3781 up the pass1 processing of e2fsck. This is a read-only feature
3782 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3783 mount filesystems that have had this feature enabled.
3785 Severity : enhancement
3787 Description: Improve multi-block allocation algorithm to avoid fragmentation
3788 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3789 allocation locality and avoid filesystem fragmentation.
3791 ------------------------------------------------------------------------------
3793 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3795 * Support for kernels:
3796 2.4.21-47.0.1.EL (RHEL 3)
3797 2.6.5-7.283 (SLES 9)
3798 2.6.9-42.0.10.EL (RHEL 4)
3799 2.6.12.6 vanilla (kernel.org)
3800 2.6.16.27-0.9 (SLES 10)
3801 * Recommended e2fsprogs version: 1.39.cfs5
3803 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3807 Frequency : occasional, depends on client load and configuration
3808 Bugzilla : 12181, 12203
3809 Description: data loss for recently-modified files
3811 Details : In some cases it is possible that recently written or created
3812 files may not be written to disk in a timely manner (this should
3813 normally be within 30s unless client IO load is very high).
3814 The problem appears as zero-length files or files that are a
3815 multiple of 1MB in size after a client crash or client eviction
3816 that are missing data at the end of the file.
3818 This problem is more likely to be hit on clients where files are
3819 repeatedly created and unlinked in the same directory, clients
3820 have a large amount of RAM, have many CPUs, the filesystem has
3821 many OSTs, the clients are rebooted frequently, and/or the files
3822 are not accessed by other nodes after being written.
3824 The presence of the problem can be detected by looking at
3825 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3826 smaller than the second (nr_unused) then dirty files will not
3827 be flushed automatically to disk. "sync; sleep 10" should be
3828 run several times on the node before unmounting it to update
3829 Lustre (this is also safe to run on nodes without this problem).
3831 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3832 that can cause this same problem, so customers using that kernel
3833 also need to update the kernel in addition to Lustre. In order
3834 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3836 It is normal that files written just before a client crash (less
3837 than 30s) may not yet have been flushed to disk, even for local
3841 Frequency : frequent on thin XT3 nodes
3843 Description: UUID collision on thin XT3 Linux nodes
3844 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3845 randomly, since we relied on an insufficiently-seeded PRNG.
3850 Description: OSS hangs after "All ost request buffers busy"
3851 Details : A deadlock between quota and journal operations caused OSS
3852 hangs after printing "All ost request buffers busy."
3855 Frequency : always on liblustre builds
3857 Description: Cleanup compiler warnings on liblustre
3860 Frequency : always on liblustre builds on XT3
3862 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3866 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3867 > 32000 subdirectories
3868 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3869 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3870 subdirectory count crosses 32000. This will aid e2fsck to
3871 correctly handle more than 32000 subdirectories.
3874 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3876 Description: System hang on startup
3877 Details : This bug allowed the liblustre (e.g. catamount) client to
3878 return to the app before handling all startup RPCs. This
3879 could leave the node unresponsive to lustre network traffic
3880 and manifested as a server ptllnd timeout.
3882 Severity : enhancement
3884 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3885 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3886 boolean to control whether to print peer debug info when a
3887 client's RPC times out.
3892 Description: make O_SYNC working on 2.6 kernels
3893 Details : 2.6 kernels use different method for mark pages for write,
3894 so need add a code to lustre for O_SYNC work.
3899 Description: Failure to close file and release space on NFS
3900 Details : Put inode details into lock acquired in ll_intent_file_open.
3901 Use mdc_intent_lock in ll_intent_open to properly
3902 detect all kind of errors unhandled by mdc_enqueue.
3907 Description: proc file read during shutdown sometimes raced obd removal,
3909 Details : Add lock to prevent obd access after proc file removal.
3912 Frequency : Only for files larger than 4GB on 32-bit clients.
3914 Description: improperly doing page alignment of locks
3915 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3916 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3921 Description: RPCs being resent when they shouldn't be
3922 Details : Some RPCs that should not be resent are being resent. This
3923 can cause inconsistencies in the RPC state machine. Do not
3924 resend such requests.
3927 Frequency : rare, only with NFS export
3929 Description: Crash on NFS re-export node
3930 Details : under very unusual load conditions an assertion is hit in
3931 ll_intent_file_open()
3934 Frequency : only if OST filesystem is corrupted
3936 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3937 Details : for a short time RPCs with bulk IO are in the replay list,
3938 but replay of bulk IOs is unimplemented. If the OST filesystem
3939 is corrupted due to disk cache incoherency and then replay is
3940 started it is possible to trip an assertion. Avoid putting
3941 committed RPCs into the replay list at all to avoid this issue.
3946 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3947 Details : Large single O_DIRECT read and write calls can fail to allocate
3948 a sufficiently large buffer to process the request. In case of
3949 allocation failure the allocation is retried with a smaller
3950 buffer and broken into smaller requests.
3952 Severity : enhancement
3954 Description: Add -o localflock option to simulate old noflock behaviour.
3955 Details : This will achieve local-only flock/fcntl locks coherentness.
3960 Description: versioning check is incomplete
3961 Details : Checking the version difference of client vs. server, report
3962 error if the gap is too big.
3967 Description: add support PG_writeback bit
3968 Details : add support for PG_writeback bit for Lustre, for more carefull
3969 work with page cache in 2.6 kernel. This also fix some deadlocks
3970 and remove hack for work O_SYNC with 2.6 kernel.
3972 Severity : enhancement
3974 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3975 Details : The uninit_groups feature works in conjunction with the kernel
3976 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3977 up the pass1 processing of e2fsck. This is a read-only feature
3978 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3979 mount filesystems that have had this feature enabled.
3981 Severity : enhancement
3983 Description: Improve multi-block allocation algorithm to avoid fragmentation
3984 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3985 allocation locality and avoid filesystem fragmentation.
3987 ------------------------------------------------------------------------------
3989 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3991 * Support for kernels:
3992 2.6.9-42.0.3.EL (RHEL 4)
3993 2.6.5-7.276 (SLES 9)
3994 2.4.21-47.0.1.EL (RHEL 3)
3995 2.6.12.6 vanilla (kernel.org)
3996 2.6.16.21-0.8 (SLES10)
3997 * Recommended e2fsprogs version: 1.39.cfs2-0
3999 * The backwards-compatible /proc/sys/portals symlink has been removed
4000 in this release. Before upgrading, please ensure that you change
4001 any configuration scripts or /etc/sysctl.conf files that access
4002 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4003 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4004 in advance of the upgrade on any system running Lustre 1.4.6 or
4005 newer, since /proc/sys/lnet was added in that version.
4006 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4010 Frequency : only when quota is used
4012 Description: avoid scanning export list for quota master
4013 Details : Change the algorithms to avoid scanning export list in order
4014 to improve the efficiency.
4017 Frequency : MDS failover only, very rarely
4019 Description: "went back in time" messages on mds failover
4020 Details : The greatest transno may be lost when the current operation
4021 finishes with an error (transno==0) and the client's last_rcvd
4022 record is over-written. Save the greatest transno in the
4023 mds_last_transno for this case.
4026 Frequency : always for specific kernels and striping counts
4028 Description: client may get "Matching packet too big" without ACL support
4029 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4030 when trying to access files in certain configurations. The
4031 clients should in fact be denied when mounting because they do
4032 not understand ACLs.
4035 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4037 Description: many clients connecting with IO in progress causes connect timeouts
4038 Details : Avoid synchronous journal commits to avoid delays caused by many
4039 clients connecting/disconnecting when bulk IO is in progress.
4040 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4041 OST_IO_PORTAL to avoid delays behind potentially many pending
4045 Frequency : occasionally with multiple writers to a single file
4047 Description: shared writes to file may result in wrong size reported by stat()
4048 Details : Allow growing of kms when extent lock is cancelled
4051 Frequency : always with random mmap IO to multi-striped file
4053 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4054 Details : Only if the hole is at the end of OST object so that kms is too
4055 small. Fix is to increase kms accordingly in ll_nopage.
4058 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4060 Description: writes to a missing object would leak memory on the OST
4061 Details : If there is an inconsistency between the MDS and OST filesystems,
4062 such that the MDS references an object that doesn't exist, writes
4063 to that object will leak memory due to incorrect cleanup in the
4064 error handling path, eventually running out of memory on the OST.
4069 Description: Creating too long symlink causes lustre errors
4070 Details : Check symlink and name lengths before sending requests to MDS.
4073 Frequency : only if flock is enabled (not on by default)
4075 Description: posix locks not released on fd closure on 2.6.9+
4076 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4077 kernels, this caused such locks not to be released on fd close and
4078 then assertions on fs unmount about still used locks.
4081 Frequency : MDS failover only, very rarely
4083 Description: clients may get ASSERTION(granted_lock != NULL)
4084 Details : When request was taking a long time, and a client was resending
4085 a getattr by name lock request. The were multiple lock requests
4086 with the same client lock handle and
4087 mds_getattr_name->fixup_handle_for_resent_request found one of the
4088 lock handles but later failed with ASSERTION(granted_lock != NULL).
4093 Description: handle->h_buffer_credits > 0, assertion failure
4094 Details : h_buffer_credits is zero after truncate, causing assertion
4095 failure. This patch extends the transaction or creates a new
4099 Frequency : NFS re-export or patchless client
4100 Bugzilla : 11179, 10796
4101 Description: Crash on NFS re-export node (__d_move)
4102 Details : We do not want to hash the dentry if we don't have a lock.
4103 But if this dentry is later used in d_move, we'd hit uninitialised
4104 list head d_hash, so we just do this to init d_hash field but
4105 leave dentry unhashed.
4108 Frequency : NFS re-export or patchless client
4110 Description: NFS exports has problem with symbolic link
4111 Details : lustre client didn't properly install dentry when re-exported
4112 to NFS or running patchless client.
4115 Frequency : NFS re-export or patchless client
4117 Description: Various nfs/patchless fixes.
4118 Details : fixes reuse disconected alias for lookup process - this fixes
4119 warning "find_exported_dentry: npd != pd",
4120 fix permission error with open files at nfs.
4121 fix apply umask when do revalidate.
4124 Frequency : occasional
4126 Description: Crash on NFS re-export node
4127 Details : calling clear_page() on the wrong pointer triggered oops in
4128 generic_mapping_read().
4131 Frequency : rarely, using O_DIRECT IO
4133 Description: unaligned directio crashes client with LASSERT
4134 Details : check for unaligned buffers before trying any requests.
4137 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4139 Description: stale data returned from RAID cache
4140 Details : If only a small amount of IO is done to the RAID device before
4141 reading it again it is possible to get stale data from the RAID
4142 cache instead of reading it from disk.
4145 Frequency : always for sles10 kernel
4147 Description: sles10 support
4148 Details : ll_follow_link: compile fixes and using of nd_set_link
4149 under newer kernels.
4152 Frequency : depends on arch, kernel and compiler version, always on sles10
4155 Description: recursive or deep enough symlinks cause stack overflow
4156 Details : getting rid of large stack-allocated variable in
4160 Frequency : depends on hardware
4162 Description: lustre write performance loss in the SLES10 kernel
4163 Details : the performance loss is caused by using of write barriers in the
4164 ext3 code. The SLES10 kernel turns barrier support on by
4165 default. The fix is to undo that change for ldiskfs.
4167 ------------------------------------------------------------------------------
4169 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4171 * Support for kernels:
4172 2.6.9-42.0.3EL (RHEL 4)
4173 2.6.5-7.276 (SLES 9)
4174 2.4.21-47.0.1.EL (RHEL 3)
4175 2.6.12.6 vanilla (kernel.org)
4179 Frequency : quota enabled and large files being deleted
4181 Description: releasing more than 4GB of quota at once hangs OST
4182 Details : If a user deletes more than 4GB of files on a single OST it
4183 will cause the OST to spin in an infinite loop. Release
4184 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4189 Description: statfs data retrieved from /proc may be stale or zero
4190 Details : When reading per-device statfs data from /proc, in the
4191 {kbytes,files}_{total,free,avail} files, it may appear
4192 as zero or be out of date.
4195 Frequency : systems with MD RAID1 external journal devices
4197 Description: lconf's call to blkid is confused by RAID1 journal devices
4198 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4199 of returning all block devices that match the journal UUID.
4202 Frequency : always, for aggregate stripe size over 4GB
4204 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4205 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4206 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4209 Frequency : always if "lfs find" used on a local file/directory
4211 Description: "lfs find" segfaults if used on a local file/directory
4212 Details : The case where a directory component was not specified wasn't
4213 handled correctly. Handle this properly.
4216 Frequency : always on ppc64
4218 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4219 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4220 perform endianness conversion.
4223 Frequency : rarely (truncate to non-zero file size after write under load)
4224 Bugzilla : 10730, 10687
4225 Description: Files padded with zeros to next 4K multiple
4226 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4227 it is possible that files that are truncated to a non-zero size
4228 immediately after being written are filled with zero bytes beyond
4229 the truncated size. No file data is lost.
4231 Severity : enhancement
4233 Description: Allow recovery/failover for liblustre clients.
4234 Details : liblustre clients were unaware of failover configurations until
4237 Severity : enhancement
4239 Description: user file locks should fail when not mounting with flock option
4240 Details : Set up an error-returning stub in ll_file_operations.lock field
4241 to prevent incorrect behaviour when client is mounted without
4242 flock option. Also, set up properly f_op->flock field for
4246 Frequency : always on ia64
4248 Description: "lfs df" loops on printing out MDS statfs information
4249 Details : The obd_ioctl_data was not initialized and in some systems
4250 this caused a failure during the ioctl that did not return
4251 an error. Initialize the struct and return an error on failure.
4254 Frequency : SLES 9 only
4256 Description: Error of copying files with lustre special EAs as root
4257 Details : Client side always return success for setxattr call for lustre
4258 special xattr (currently only "trusted.lov").
4261 Frequency : rarely on clusters with both ia64+i386 clients
4263 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4264 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4265 clients are doing concurrent writes to the same file it is
4266 possible that the ia64 clients may LASSERT because the OST
4267 extent locks are not PAGE_SIZE aligned. Ensure that grown
4268 locks are always aligned on the request boundary.
4271 Frequency : specific use, occasional
4273 Description: Overwriting in use executable truncates on-disk binary image
4274 Details : If one node attempts to overwrite an executable in use by
4275 another node, we now correctly return ETXTBSY instead of
4276 truncating the file.
4278 Severity : enhancement
4280 Description: Async OSC create to avoid the blocking unnecessarily.
4281 Details : If an OST has no remaining object, system will block on the creating
4282 when need to create a new object on this OST. Now, ways use
4283 pre-created objects when available, instead of blocking on an
4284 empty osc while others are not empty. If we must block, we block
4285 for the shortest possible period of time.
4290 Description: chmod on Lustre root is propagated to other clients
4291 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4292 invalid by the follow_mount time.
4295 Frequency : liblustre clients only
4297 Description: Race in 'instant cancel' lock handling could lead to such locks
4298 never to be granted in case of SMP MDS
4299 Details : Do not destroy not yet granted but cbpending locks in
4303 Frequency : replay/resend of open
4305 Description: non null lock assetion failure in mds_intent_policy
4306 Details : Trying to replay/resend lockless open requests resulted in
4307 mds_open() returning 0 with no lock. Now it sets a flag if
4308 a lock is going to be returned.
4310 Severity : enhancement
4312 Description: Checksum enhancements
4313 Details : New checksum enhancements allow for resending RPCs that failed
4316 Severity : enhancement
4318 Description: Tunables on number of dirty pages in cacche
4319 Details : Allow to set limit on number of dirty pages cached.
4324 Description: client crash on unmount - lock still has references
4325 Details : In some error handling cases it was possible to leak a lock
4326 reference on a client while accessing a file. This was not
4327 harmful to the client during operation, but would cause the
4328 client to crash when the filesystem is unmounted.
4331 Frequency : specific case, rare
4333 Description: ETXTBSY on mds though file not in use by client
4334 Details : ETXTBSY is no longer incorrectly returned when attempting to
4335 chmod or chown a directory that the user previously tried to
4336 execute or a currently-executing binary.
4339 Frequency : extremely rare except on liblustre-based clients
4341 Description: Lustre space not freed when files are deleted
4342 Details : Clean up open-unlinked files after client eviction. Previously
4343 the unlink was skipped and the files remained as orphans.
4348 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4349 Details : In some cases it was possible to send a blocking callback to
4350 a client doing a glimpse, even though that client didn't get
4351 a lock granted. When the glimpse lock is cancelled on the OST
4352 the freed lock is left on the waiting list and corrupted the list.
4355 Frequency : all core dumps
4357 Description: Broke core dumps to lustre
4358 Details : Negative dentry may be unhashed if parent does not have UPDATE
4359 lock, but some callers, e.g. do_coredump, expect dentry to be
4360 hashed after successful create, hash it in ll_create_it.
4362 ------------------------------------------------------------------------------
4364 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4366 * Support for kernels:
4367 2.6.9-42.0.2.EL (RHEL 4)
4368 2.6.5-7.276 (SLES 9)
4369 2.4.21-40.EL (RHEL 3)
4370 2.6.12.6 vanilla (kernel.org)
4374 Frequency : always on RHEL 3
4376 Description: Number of open files grows over time
4377 Details : The number of open files grows over time, whether or not
4378 Lustre is started. This was due to a filp leak introduced
4379 by one of our kernel patches.
4381 ------------------------------------------------------------------------------
4383 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4385 * Support for kernels:
4386 2.6.9-42.EL (RHEL 4)
4387 2.6.5-7.276 (SLES 9)
4388 2.4.21-40.EL (RHEL 3)
4389 2.6.12.6 vanilla (kernel.org)
4394 Bugzilla : 5719, 9635, 9792, 9684
4395 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4396 Details : If a server is under heavy load and cannot reply to new
4397 connection requests before the client resends the (re)connect,
4398 the connection handling code can behave badly if two service
4399 threads are concurrently handing separate (re)connections from
4400 the same client. Add better locking to the connection handling
4401 code, and ensure that only a single connection will be processed
4402 for a given client UUID, even if the lock is dropped.
4404 Severity : enhancement
4406 Description: add TCP zero-copy support to kernel
4407 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4408 sends if the hardware supports scatter-gather and checksumming.
4409 This allows socklnd to do client-write and server-read more
4410 efficiently and reduce CPU utilization from skbuf copying.
4413 Frequency : only if NFS exporting from client
4415 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4416 Details : If setting attributes on a file created under NFS that had
4417 never been opened it would be possible to oops the client
4418 if the file had no objects.
4421 Frequency : always for liblustre
4423 Description: liblustre client does MDS+OSTs setattr RPC for each write
4424 Details : When doing a write from a liblustre client, the client
4425 incorrectly issued an RPC to the MDS and each OST the file was
4426 striped over in order to update the timestamps. When writing
4427 with small chunks and many clients this could overwhelm the MDS
4428 with RPCs. In all cases it would slow down the write because
4429 these RPCs are unnecessary.
4431 Severity : enhancement
4433 Description: allow number of MDS service threads to be changed at module load
4434 Details : It is now possible to change the number of MDS service threads
4435 running. Adding "options mds mds_num_threads={N}" to the MDS's
4436 /etc/modprobe.conf will set the number of threads for the next
4437 time Lustre is restarted (assuming the "mds" module is also
4438 reloaded at that time). The default number of threads will
4439 stay the same, 32 for most systems.
4444 Description: OST crash if filesystem is unformatted or corrupt
4445 Details : If an OST is started on a device that has never been formatted
4446 or if the filesystem is corrupt and cannot even mount then the
4447 error handling cleanup routines would dereference a NULL pointer.
4452 Description: NULL pointer deref in llap_from_page.
4453 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4454 incorrect) mapping if the page was truncated/reclaimed while it was
4455 searched for. Check for this condition and skip such pages when
4456 doing readahead. Introduce extra check to llap_from_page() to
4457 verify page->mapping->host is non-NULL (so page is not anonymous).
4460 Frequency : Sometimes when using sys_sendfile
4462 Description: "page not covered by a lock" warnings from ll_readpage
4463 Details : sendfile called ll_readpage without right page locks present.
4464 Now we introduced ll_file_sendfile that does necessary locking
4465 around call to generic_file_sendfile() much like we do in
4469 Frequency : with certain MDS communication failures at client mount time
4471 Description: NULL pointer deref after failed client mount
4472 Details : a client connection request may delayed by the network layer
4473 and not be sent until after the PTLRPC layer has timed out the
4474 request. If the client fails the mount immediately it will try
4475 to clean up before the network times out the request. Add a
4476 reference from the request import to the obd device and delay
4477 the cleanup until the network drops the request.
4480 Frequency : occasionally during client (re)connect
4482 Description: assertion failure during client (re)connect
4483 Details : processing a client connection request may be delayed by the
4484 client or server longer than the client connect timeout. This
4485 causes the client to resend the connection request. If the
4486 original connection request is replied in this interval, the
4487 client may trip an assertion failure in ptlrpc_connect_interpret()
4488 which thought it would be the only running connect process.
4491 Frequency : only with obd_echo servers and clients that are rebooted
4493 Description: kernel BUG accessing uninitialized data structure
4494 Details : When running an obd_echo server it did not start the ping_evictor
4495 thread, and when a client was evicted an uninitialized data
4496 structure was accessed. Start the ping_evictor in the RPC
4497 service startup instead of the OBD startup.
4499 Severity : enhancement
4500 Bugzilla : 10193 (patchless)
4501 Description: Remove dependency on various unexported kernel interfaces.
4502 Details : No longer need reparent_to_init, exit_mm, exit_files,
4503 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4506 Frequency : rare (only users of deprecated and unsupported LDAP config)
4508 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4509 Details : LDAP apparently contains 'lustreName' attributes instead of
4510 'name'. A simple remapping of the name is sufficient.
4513 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4515 Description: waiting_locks_callback trips kernel BUG if client is evicted
4516 Details : Running with the dump_on_timeout debug flag turned on makes
4517 it possible that the waiting_locks_callback() can try to dump
4518 the Lustre kernel debug logs from an interrupt handler. Defer
4519 this log dumping to the expired_lock_main() thread.
4521 Severity : enhancement
4523 Description: Support NFS exporting on 2.6 kernels.
4524 Details : Implement non-rawops metadata methods for NFS server to use without
4525 changing NFS server code.
4528 Frequency : very rare (synthetic metadata workload only)
4530 Description: two racing renames might cause an MDS thread to deadlock
4531 Details : Running the "racer" program may cause one MDS thread to rename
4532 a file from being the source of a rename to being the target of
4533 a rename at exactly the same time that another thread is doing
4534 so, and the second thread has already enqueued these locks after
4535 doing a lookup of the target and is trying to relock them in
4536 order. Ensure that we don't try to re-lock the same resource.
4539 Frequency : only very large systems with liblustre clients
4541 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4542 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4543 all OSTs in parallel. This allows the eviction of stale liblustre
4544 clients to proceed much faster than if they were done in series,
4545 and also offers similar improvements for other set_info RPCs.
4550 Description: excessive CPU usage during initial read phase on client
4551 Details : During the initial read phase on a client, it would agressively
4552 retry readahead on the file, consuming too much CPU and impacting
4553 performance (since 1.4.5.8). Improve the readahead algorithm
4554 to avoid this, and also improve some other common cases (read
4555 of small files in particular, where "small" is files smaller than
4556 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4561 Description: MDS crash when receiving packet with unknown intent.
4562 Details : Do not LBUG in unknown intent case, just return -EFAULT
4564 Severity : enhancement
4565 Bugzilla : 9293, 9385
4566 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4567 Details : Do not serialize getattr (non-intent version) and statfs.
4570 Frequency : occasional, when OST network is overloaded/intermittent
4572 Description: client evicted by OST after bulk IO timeout
4573 Details : If a client sends a bulk IO request (read or write) the OST
4574 may evict the client if it is unresposive to its data GET/PUT
4575 request. This is incorrect if the network is overloaded (takes
4576 too long to transfer the RPC data) or dropped the OST GET/PUT
4577 request. There is no need to evict the client at all, since
4578 the pinger and/or lock callbacks will handle this, and the
4579 client can restart the bulk request.
4582 Frequency : Always when mmapping file with no objects
4584 Description: client crashes when mmapping file with no objects
4585 Details : Check that we actually have objects in a file before doing any
4586 operations on objects in ll_vm_open, ll_vm_close and
4592 Description: Request leak when working with deleted CWD
4593 Details : Introduce advanced request refcount tracking for requests
4594 referenced from lustre intent.
4596 Severity : Enhancement
4598 Description: Cache open file handles on client.
4599 Details : MDS now will return special lock along with openhandle, if
4600 requested and client is allowed to hold openhandle, even if unused,
4601 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4602 closing files for every read/write openration.
4604 Severity : Enhancement
4606 Description: Cache open negative dentries on client when possible.
4607 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4608 negative dentries on lock revocation.
4613 Description: Remounting a client read-only wasn't possible with a zconf mount
4614 Details : It wasn't possible to remount a client read-only with llmount.
4616 Severity : enhancement
4617 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4618 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4619 MPICH is included to add Lustre-specific ADIO interfaces.
4620 This is based closely on the UFS ADIO layer and only differs
4621 in file creation, in order to allow the OST striping to be set.
4622 This is user-contributed code and not supported by CFS.
4627 Description: extended inode attributes (immutable, append-only) work improperly
4628 when 2.4 and 2.6 kernels are used on client/server or vice versa
4629 Details : Introduce kernel-independent values for these flags.
4631 Severity : enhancement
4634 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4635 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4636 granularity. Now, Fractional MB granularity is be supported,
4637 this is very useful for low memory system.
4639 Severity : enhancement
4641 Description: Getattr by fid
4642 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4643 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4644 have locks and avoids getattr rpc requests for it on every stat.
4647 Frequency : Always, for filesystems larger than 2TB
4649 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4650 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4651 filesystems larger than 2TB to be mounted with mballoc enabled.
4654 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4656 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4657 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4658 filesystems larger than 2TB is mounted then the kernel will
4659 silently corrupt the start of the filesystem. CONFIG_LBD is
4660 enabled for all CFS-supported kernels, but the possibility of
4661 this happening with a modified kernel config exists.
4663 Severity : enhancement
4665 Description: add client O_DIRECT support for 2.6 kernels
4666 Details : It is now possible to do O_DIRECT reads and writes to files
4667 in the Lustre client mountpoint on 2.6 kernel clients.
4669 Severity : enhancement
4671 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4672 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4673 parallel, not waiting for response from every OST before sending
4674 a rpc to the next OST.
4679 Description: setattr vs write race when updating file timestamps
4680 Details : Client processes that update a file timestamp into the past
4681 right after writing to the file (e.g. tar) it is possible that
4682 the updated file modification time can be reset to the current
4683 time due to a race between processing the setattr and write RPC.
4685 Severity : enhancement
4687 Description: Bring 'lfs find' closer in line with regular Linux find.
4688 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4689 -print0 options and obtains all the needed info through the lustre
4692 Severity : enhancement
4694 Description: support up to 1024 configured devices on one node
4695 Details : change obd_dev array from statically allocated to dynamically
4696 allocated structs as they are first used to reduce memory usage
4701 Description: Flush dirty partially truncated pages during truncate
4702 Details : Immediatelly flush partially truncated pages in filter_setattr,
4703 this way we completely avoid having any pages in page cache on OST
4704 and can retire ugly workarounds during writes to flush such pages.
4709 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4710 Details : protect inode from truncation within vfs_unlink() context
4711 just take a reference before calling vfs_unlink() and release it
4712 when parent's i_sem is free.
4717 Description: last_id value checked outside lock on OST caused LASSERT failure
4718 Details : If there were multiple MDS->OST object precreate requests in
4719 flight, it was possible that the OST's last object id was checked
4720 outside a lock and incorrectly tripped an assertion. Move checks
4721 inside locks, and discard old precreate requests.
4724 Frequency : always, if extents are used on OSTs
4726 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4727 Details : OSTs using the extents format would not zero the high 16 bits of
4728 the index physical block number. This is not a problem for any
4729 OST filesystems smaller than 16TB, and no kernels support ext3
4730 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4731 new/modified files) and can be fixed for existing filesystems
4732 with e2fsprogs-1.39-cfs1.
4737 Description: import connection selection may be incorrect if timer wraps
4738 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4739 import connections to be ignored if the 32-bit jiffies counter
4740 wraps. Use a 64-bit jiffies counter.
4743 Frequency : very large clusters immediately after boot
4745 Description: LNET request buffers exhausted under heavy short-term load
4746 Details : If a large number of client requests are generated on a service
4747 that has previously never seen so many requests it is possible
4748 that the request buffer growth cannot keep up with the spike in
4749 demand. Instead of dropping incoming requests, they are held in
4750 the LND until the RPC service can accept more requests.
4753 Frequency : Sometimes during replay
4755 Description: Assertion failure in ll_local_open after replay.
4756 Details : If replay happened on an open request reply before we were able
4757 to set replay handler, reply will become not swabbed tripping the
4758 assertion in ll_local_open. Now we set the handler right after
4759 recognising of open request
4762 Frequency : very rare
4764 Description: kernel reports "badness in vsnprintf"
4765 Details : Reading from the "recovery_status" /proc file in small chunks
4766 may cause a negative length in lprocfs_obd_rd_recovery_status()
4767 call to vsnprintf() (which is otherwise harmless). Exit early
4768 if there is no more space in the output buffer.
4770 Severity : enhancement
4772 Description: clear OBD RPC statistics by writing to them
4773 Details : It is now possible to clear the OBD RPC statistics by writing
4774 to the "stats" file.
4779 Description: Client mtime is not the same on different clients after utimes
4780 Details : In some cases, the client was using the utimes() syscall on
4781 a file cached on another node. The clients now validate the
4782 ctime from the MDS + OSTs to determine which one is right.
4787 Description: Inability to activate failout mode
4788 Details : lconf script incorrectly assumed that in pythong string's numeric
4789 value is used in comparisons.
4792 Frequency : always with multiple stripes per file
4794 Description: Inefficient object allocation for mutli-stripe files
4795 Details : When selecting which OSTs to stripe files over, for files with
4796 a stripe count that divides evenly into the number of OSTs,
4797 the MDS is always picking the same starting OST for each file.
4798 Return the OST selection heuristic to the original design.
4803 Description: mount failures may take full timeout to return an error
4804 Details : Under some heavy load conditions it is possible that a
4805 failed mount can wait for the full obd_timeout interval,
4806 possibly several minutes, before reporting an error.
4807 Instead return an error as soon as the status is known.
4809 Frequency : quota enabled and large files being deleted
4811 Description: releasing more than 4GB of quota at once hangs OST
4812 Details : If a user deletes more than 4GB of files on a single OST it
4813 will cause the OST to spin in an infinite loop. Release
4814 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4819 Description: statfs data retrieved from /proc may be stale or zero
4820 Details : When reading per-device statfs data from /proc, in the
4821 {kbytes,files}_{total,free,avail} files, it may appear
4822 as zero or be out of date.
4825 Frequency : systems with MD RAID1 external journal devices
4827 Description: lconf's call to blkid is confused by RAID1 journal devices
4828 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4829 of returning all block devices that match the journal UUID.
4832 Frequency : always, for aggregate stripe size over 4GB
4834 Description: assertion fails when trying to use 4GB stripe size
4835 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4836 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4839 Frequency : always on ppc64
4841 Description: the first write on an ext3 filesystem with mballoc got stuck
4842 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4843 perform endianness conversion.
4845 ------------------------------------------------------------------------------
4847 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4849 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4850 INTEROPERATE with older versions automatically. Please read the
4851 user documentation before upgrading any part of a live system.
4852 * WARNING: Lustre networking configuration changes are required with
4853 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4856 * Support for kernels:
4857 2.6.9-22.0.2.EL (RHEL 4)
4858 2.6.5-7.244 (SLES 9)
4859 2.6.12.6 vanilla (kernel.org)
4862 Severity : enhancement
4863 Bugzilla : 7981/8208
4864 Description: Introduced Lustre Networking (LNET)
4865 Details : LNET is new networking infrastructure for Lustre, it includes
4866 a reorganized network configuration mode (see the user
4867 documentation for full details) as well as support for routing
4868 between different network fabrics. Lustre Networking Devices
4869 (LNDs) for the supported network fabrics have also been
4870 created for this new infrastructure.
4872 Severity : enhancement
4873 Description: Introduced Access control lists
4874 Details : clients can set ACLs on files and directories in order to have
4875 more fine-grained permissions than the standard Unix UGO+RWX.
4876 The MDS must be started with the "-o acl" mount option.
4878 Severity : enhancement
4879 Description: Introduced filesystem quotas
4880 Details : Administrators may now establish per-user quotas on the
4883 Severity : enhancement
4885 Description: Configuration change for the XT3
4886 The PTLLND is now used to run Lustre over Portals on the XT3
4887 The configure option(s) --with-cray-portals are no longer used.
4888 Rather --with-portals=<path-to-portals-includes> is used to
4889 enable building on the XT3. In addition to enable XT3 specific
4890 features the option --enable-cray-xt3 must be used.
4895 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4896 Details : A race between allocating a new llcd and re-getting the llcd_lock
4897 allowed another thread to grab newly-allocated llcd.
4899 Severity : enhancement
4901 Description: 2.6 OST async journal commit and locking fix to improve performance
4902 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4903 2.6 kernels are now async as they already were in 2.4 kernels so
4904 that they can commit concurrently with the network bulk transfer.
4905 For block-allocated files the filter allocation semaphore is held
4906 to avoid filesystem fragmentation during allocation. BKL lock
4907 removed for 2.6 xattr operations where it is no longer needed.
4912 Description: lconf incorrectly determined whether two IP networks could talk
4913 Details : In some more complicated routing and multiple-network
4914 configurations, lconf will avoid trying to make a network
4915 connection to a disjoint part of the IP space. It was doing the
4916 math incorrectly for one set of cases.
4921 Description: Fix for potential infinite loop processing records in an llog.
4922 Details : If an llog record is corrupted/zeroed, it is possible to loop
4923 forever in llog_process(). Validate the llog record length
4924 and skip the remainder of the block on error.
4927 Frequency : occasional (liblustre only)
4929 Description: liblustre could not open files whose last component is a symlink
4930 Details : sysio_path_walk() would incorrectly pass the open intent to
4931 intermediate path components.
4934 Frequency : rare (liblustre only with non-standard tuning)
4935 Bugzilla : 7201 (7350)
4936 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4937 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4938 after it has been released, only do this for referenced locks
4940 Severity : enhancement
4942 Description: specifying an (invalid) directory default stripe_size of -1
4943 would reset the directory default striping
4944 Details : stripe_size -1 was used internally to signal directory stripe
4945 removal, now use "all default" to signal dir stripe removal
4946 as a directory striping of "all default" is not useful
4949 Frequency : common for large clusters running liblustre clients
4951 Description: doing an ls when liblustre clients are running is slow
4952 Details : sending a glimpse AST to a liblustre client waits for every AST
4953 to time out, as liblustre clients will not respond. Since they
4954 cannot cache data we refresh the OST lock LVB from disk instead.
4956 Severity : enhancement
4958 Description: doing an ls at the same time as file IO can be slow
4959 Details : enqueue and other "small" requests can be blocked behind many
4960 large IO requests. Create a new OST IO portal for non-IO
4961 requests so they can be processed faster.
4964 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4966 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4967 created file with mode 000
4968 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4969 allow an ftruncate() as a non-root user to a file with mode 000.
4970 Lustre now respects this flag to disable mode checks when
4971 truncating a file owned by the user
4974 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4977 Description: Revoking locks from clients that went dead or catatonic might take
4979 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4980 cancellation of such locks instant on servers without waiting for
4981 any reply from clients. Clients drops these locks when cancel
4982 notification from server is received without replying.
4985 Frequency : liblustre-only, when liblustre client dies or becomes busy
4987 Description: Doing ls on Linux clients can take a long time with active
4989 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4990 granting such locks to it in the first place if possible. Locks
4991 are taken by proxy on the OST during the read or write and
4992 dropped immediately afterward. Add connect flags handling, do
4993 not grant locks to liblustre clients for glimpse ASTs.
4995 Severity : enhancement
4997 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4998 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4999 sequential reads and behaves badly with applications doing random
5000 reads. Improve it by reading ahead at least read region, and
5001 avoiding excessive large RPC for small reads.
5003 Severity : enhancement
5005 Description: Creating more than 1000 files for a single job may cause a load
5006 imbalance on the OSTs if there are also a large number of OSTs.
5007 Details : qos_prep_create() uses an OST index reseed value that is an
5008 even multiple of the number of available OSTs so that if the
5009 reseed happens in the middle of the object allocation it will
5010 still utilize the OSTs as uniformly as possible.
5015 Description: OST or MDS may oops in ping_evictor_main()
5016 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5017 but doesn't restart at beginning of obd_exports_timed list
5020 Severity : enhancement
5022 Description: improve by-nid export eviction on the MDS and OST
5023 Details : allow multiple exports with the same NID to be evicted at one
5024 time without re-searching the exports list.
5027 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5029 Description: MDS may oops in groups_free()
5030 Details : in rare race conditions a newly allocated group_info struct is
5031 freed again, and this can be NULL. The 2.4 compatibility code
5032 for groups_free() checked for a NULL pointer, but 2.6 did not.
5035 Frequency : common for liblustre clients doing little filesystem IO
5036 Bugzilla : 9352, 7313
5037 Description: server may evict liblustre clients accessing contended locks
5038 Details : if a client is granted a lock or receives a completion AST
5039 with a blocking AST already set it would not reply to the AST
5040 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5043 Frequency : lfs setstripe, only systems with more than 160 OSTs
5045 Description: unable to set striping with a starting offset beyond OST 160
5046 Details : llapi_create_file() incorrectly limited the starting stripe
5047 index to the maximum single-file stripe count.
5050 Frequency : LDAP users only
5052 Description: lconf did not handle in-kernel recovery with LDAP properly
5053 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5055 Severity : enhancement
5057 Description: bind OST threads to NUMA nodes to improve performance
5058 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5059 node and do their allocations there to localize memory access
5061 Severity : enhancement
5063 Description: llmount can determine client NID directly from Myrinet (GM)
5064 Details : the client NID code from gmnalnid was moved directly into
5065 llmount, removing the need to use this or specifying the
5066 client NID explicitly when mounting GM clients with zeroconf
5069 Frequency : if client is started with down MDS
5071 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5072 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5073 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5074 ptlrpc_import_delayed_req() because we didn't check if the
5075 request was interrupted, and we also didn't break out of the
5076 event loop if there was no timeout
5081 Description: data loss during non-page-aligned writes to a single file from
5082 both multiple nodes and multiple threads on one node at same time
5083 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5084 inconsistency led to false short-reads, that were cached and later
5085 used by ->prepare_write() to fill in partially written page,
5086 leading to data loss.
5089 Frequency : always, if lconf --abort_recovery used
5091 Description: lconf --abort_recovery fails with 'Operation not supported'
5092 Details : lconf was attempting to abort recovery on the MDT device and not
5095 Severity : enhancement
5097 Description: remove cleanup logs
5098 Details : replace lconf-generated cleanup logs with lustre internal
5099 cleanup routines. Eliminates the need for client-cleanup and
5102 Severity : enhancement
5104 Description: add support for EAs (user and system) on lustre filesystems
5105 Details : it is now possible to store extended attributes in the Lustre
5106 client filesystem, and with the user_xattr mount option it
5107 is possible to allow users to store EAs on their files also
5109 Severity : enhancement
5111 Description: Add possibility (config option) to show minimal available OST free
5113 Details : When compiled with --enable-mindf configure option, statfs(2)
5114 (and so, df) will return least minimal free space available from
5115 all OSTs as amount of free space on FS, instead of summary of
5116 free spaces of all OSTs.
5118 Severity : enhancement
5120 Description: do not expand extent locks acquired on OST-side
5121 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5122 by server: they are not cached anyway.
5125 Frequency : when mmap is used/binaries executed from Lustre
5127 Description: Unmmap pages before throwing them away from read cache.
5128 Details : llap_shrink cache now attempts to unmap pages before discarding
5129 them (if unmapping failed - do not discard). SLES9 kernel has
5130 extra checks that trigger if this unmapping is not done first.
5135 Description: lconf didn't resolve symlinks before checking to see whether a
5136 given mountpoint was already in use
5139 Frequency : when migrating failover services
5140 Bugzilla : 6395, 9514
5141 Description: When migrating a subset of services from a node (e.g. failback
5142 from a failover service node) the remaining services would
5143 time out and evict clients.
5144 Details : lconf --force (implied by --failover) sets the global obd_timeout
5145 to 5 seconds in order to quickly disconnect, but this caused
5146 other RPCs to time out too quickly. Do not change the global
5147 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5149 Severity : enhancement
5150 Frequency : if MDS is started with down OST
5151 Bugzilla : 9439,5706
5152 Description: Allow startup/shutdown of an MDS without depending on the
5153 availability of the OSTs.
5154 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5155 Add appropriate locking and lov-osc refcounts for safe
5156 cleaning. Add osc abort_inflight calls in case the
5157 synchronize never started.
5160 Frequency : occasional (Cray XT3 only)
5162 Description: root not authorized to access files in CRAY_PORTALS environment
5163 Details : The client process capabilities were not honoured on the MDS in
5164 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5165 been accessed by an authorized user then root was able to access
5166 the file on the local client also. The root user capabilities
5167 are now allowed on the MDS, as this environment has secure UID.
5170 Frequency : occasional
5172 Description: ldiskfs "too long searching" message happens too often
5173 Details : A debugging message (otherwise harmless) prints too often on
5174 the OST console. This has been reduced to only happen when
5175 there are fragmentation problems on the filesystem.
5180 Description: Division by zero in statfs when all OSCs are inactive
5181 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5182 lov_statfs divides by value returned by lov_get_stripecnt().
5186 Bugzilla : 9489, 3273
5187 Description: First write from each client to each OST was only 4kB in size,
5188 to initialize client writeback cache, which caused sub-optimal
5189 RPCs and poor layout on disk for the first written file.
5190 Details : Clients now request an initial cache grant at (re)connect time
5191 and so that they can start streaming writes to the cache right
5192 away and always do full-sized RPCs if there is enough data.
5193 If the OST is rebooted the client also re-establishes its grant
5194 so that client cached writes will be honoured under the grant.
5199 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5200 Details : Now I/O RPCs go to different portal number and (presumably) fast
5201 lock requests (and glimses) and other RPCs get their own service
5202 threads pool that should be able to service those RPCs
5205 Severity : enhancement
5207 Description: Ability to exchange lustre version between client and servers and
5208 issue warnings at client side if client is too old. Also for
5209 liblustre clients there is ability to refuse connection of too old
5211 Details : New 'version' field is added to connect data structure that is
5212 filled with version info. That info is later checked by server and
5216 Frequency : rare, liblustre only.
5217 Bugzilla : 9296, 9581
5218 Description: Two simultaneous writes from liblustre at offset within same page
5219 might proceed at the same time overwriting eachother with stale
5221 Details : I/O lock withing llu_file_prwv was released too early, before data
5222 actually was hitting the wire. Extended lock-holding time until
5223 server acknowledges receiving data.
5226 Frequency : extremely rare. Never observed in practice.
5228 Description: avoid generating lustre_handle cookie of 0.
5229 Details : class_handle_hash() generates handle cookies by incrementing
5230 global counter, and can hit 0 occasionaly (this is unlikely, but
5231 not impossible, because initial value of cookie counter is
5232 selected randonly). Value of 0 is used as a sentinel meaning
5233 "unassigned handle" --- avoid it. Also coalesce two critical
5234 sections in this function into one.
5236 Severity : enhancement
5238 Description: allow liblustre clients to delegate truncate locking to OST
5239 Details : To avoid overhead of locking, liblustre client instructs OST to
5240 take extent lock in ost_punch() on client's behalf. New connection
5241 flag is added to handle backward compatibility.
5243 Severity : enhancement
5244 Bugzilla : 4928, 7341, 9758
5245 Description: allow number of OST service threads to be specified
5246 Details : a module parameter allows the number of OST service threads
5247 to be specified via "options ost ost_num_threads={N}" in the
5248 OSS's /etc/modules.conf or /etc/modprobe.conf.
5252 Bugzilla : 6146, 9635, 9895
5253 Description: servers crash with bad pointer in target_handle_connect()
5254 Details : In rare cases when a client is reconnecting it was possible that
5255 the connection request was the last reference for that export.
5256 We would temporarily drop the export reference and get a new
5257 one, but this may have been the last reference and the export
5258 was just destroyed. Get new reference before dropping old one.
5260 Severity : enhancement
5261 Frequency : if client is started with failover MDS
5263 Description: Allow multiple MDS hostnames in the mount command
5264 Details : Try to read the configuration from all specified MDS
5265 hostnames during a client mount in case the "primary"
5268 Severity : enhancement
5270 Description: Stop sending data to evicted clients as soon as possible.
5271 Details : Check if the client we are about to send or are sending data to
5272 was evicted already. (Check is done every second of waiting,
5273 for which l_wait_event interface was extended to allow checking
5274 of exit condition at specified intervals).
5277 Frequency : rare, normally only when NFS exporting is done from client
5279 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5281 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5284 Frequency : always, if srand() is called before liblustre initialization
5286 Description: Liblustre uses system PRNG disturbing its usage by user application
5287 Details : Introduce internal to lustre fast and high-quality PRNG for
5288 lustre usage and make liblustre and some other places in generic
5289 lustre code to use it.
5291 Severity : enhancement
5292 Bugzilla : 9477, 9557, 9870
5293 Description: Verify that the MDS configuration logs are updated when xml is
5294 Details : Check if the .xml configuration logs are newer than the config
5295 logs stored on the MDS and report an error if this is the case.
5296 Request --write-conf, or allow starting with --old_conf.
5298 Severity : enhancement
5300 Description: Handle symlinks in the path when checking if Lustre is mounted.
5301 Details : Resolve intermediate symlinks when checking if a client has
5302 mounted a filesystem to avoid duplicate client mounts.
5307 Description: lconf can hit an error exception but still return success.
5308 Details : The lconf command catches the Command error exception at the top
5309 level script context and will exit with the associated exit
5310 status, but doesn't ensure that this exit status is non-zero.
5315 Description: failure of ptlrpc thread startup can cause oops
5316 Details : Starting a ptlrpc service thread can fail if there are a large
5317 number of threads or the server memory is very fragmented.
5318 Handle this without oopsing.
5321 Frequency : always, only if liblustre and non-default acceptor port was used
5323 Description: liblustre cannot connect to servers with non-default acceptor port
5324 Details : tcpnal_set_default_params() was not called and was therefore
5325 ignoring the environment varaible TCPNAL_PORT, as well as other
5326 TCPNAL_ environment variables
5331 Description: two objects could be created on the same OST for a single file
5332 Details : If an OST is down, in some cases it was possible to create two
5333 objects on a single OST for a single file. No problems other
5334 than potential performance impact and spurious error messages.
5338 Bugzilla : 5681, 9562
5339 Description: Client may oops in ll_unhash_aliases
5340 Details : Client dcache may become inconsistent in race condition.
5341 In some cases "getcwd" can fail if the current directory is
5347 Description: Inode refcounting problems in NFS export code
5348 Details : link_raw functions used to call d_instantiate without obtaining
5349 extra inode reference first.
5353 Bugzilla : 9942, 9903
5354 Description: Referencing freed requests leading to crash, memleaks with NFS.
5355 Details : We used to require that call to ll_revalidate_it was always
5356 followed by ll_lookup_it. Also with revalidate_special() it is
5357 possible to call ll_revalidate_it() twice for the same dentry
5358 even if first occurence returned success. This fix changes semantic
5359 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5360 reference on a request referred from the intent.
5361 ll_intent_release() then releases such a request.
5364 Frequency : rare, normally benchmark loads only
5366 Description: unlinked inodes were kept in memory on the client
5367 Details : If a client is repeatedly creating and unlinking files it
5368 can accumulate a lot of stale inodes in the inode slab cache.
5369 If there is no other client load running this can cause the
5370 client node to run out of memory. Instead flush old inodes
5371 from client cache that have the same inode number as a new inode.
5374 Frequency : SLES9 2.6.5 kernel and long filenames only
5375 Bugzilla : 9969, 10379
5376 Description: utime reports stale NFS file handle
5377 Details : SLES9 uses out-of-dentry names in some cases, which confused
5378 the lustre dentry revalidation. Change it to always use the
5382 Frequency : rare, unless heavy write-truncate concurrency is continuous
5383 Bugzilla : 4180, 6984, 7171, 9963, 9331
5384 Description: OST becomes very slow and/or deadlocked during object unlink
5385 Details : filter_destroy() was holding onto the parent directory lock
5386 while truncating+unlinking objects. For very large objects this
5387 may block other threads for a long time and slow overall OST
5388 responsiveness. It may also be possible to get a lock ordering
5389 deadlock in this case, or run out of journal credits because of
5390 the combined truncate+unlink. Solution is to do object truncate
5391 first in one transaction without parent lock, and then do the
5392 final unlink in a new transaction with the parent lock. This
5393 reduces the lock hold time dramatically.
5396 Frequency : rare, 2.4 kernels only
5398 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5399 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5400 before unmounting their filesystem in order to avoid deadlock.
5401 The kernel_locked() function in 2.4 kernels only checks whether
5402 the kernel lock is held, not whether it is this process that is
5403 holding it as 2.6 kernels do.
5408 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5409 Details : The client ptlrpc code may be trying to reconnect to a down
5410 server before a previous connection attempt has timed out.
5411 Increase the reconnect interval to be longer than the connection
5412 timeout interval to avoid sending duplicate connections to
5416 Frequency : echo_client brw_test command
5418 Description: fix echo_client to work with OST preallocated code
5419 Details : OST preallocation code (5137) didn't take echo_client IO path
5420 into account: echo_client calls filter methods outside of any
5421 OST thread and, hence, there is no per-thread preallocated
5422 pages and buffers to use. Solution: hijack pga pages for IO. As
5423 a byproduct, this avoids unnecessary data copying.
5427 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5428 Description: Client can oops in mdc_commit_close() after open replay
5429 Details : It was possible for the MDS to return an open request with no
5430 transaction number in mds_finish_transno() if the client was
5431 evicted, but without actually returning an error. Clients
5432 would later try to replay that open and may trip an assertion
5433 Simplify the client close codepath, and always return an error
5434 from the MDS in case the open is not successful.
5437 Frequency : rare, 2.6 OSTs only
5439 Description: OST may deadlock under high load on fragmented files
5440 Details : If there was a heavy load and highly-fragmented OST filesystems
5441 it was possible to have all the OST threads deadlock waiting on
5442 allocation of biovecs, because the biovecs were not released
5443 until the entire RPC IO was completed. Instead, release biovecs
5444 as soon as they are complete to ensure forward IO progress.
5446 Severity : enhancement
5448 Description: Support for specifying external journal device at mount
5449 Details : If an OST or MDS device is formatted with an external journal
5450 device, this device major/minor is stored in the ext3 superblock
5451 and may not be valid for failover. Allow detecting and
5452 specifying the external journal at mount time.
5457 Description: Mounting an MDS with pending unlinked files may cause oops
5458 Details : target_finish_recovery() calls mds_postrecov() which returned
5459 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5460 considers this an error and immediately begins cleaning up the
5461 lov, just after starting the mds_lov process
5463 Severity : enhancement
5465 Description: Implement 'lfs df' to report actual free space on per-OST basis
5466 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5467 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5468 usage of objects; '-h' to report in human readable format.
5470 ------------------------------------------------------------------------------
5472 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5479 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5480 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5481 at startup during orphan unlinking. Instead, skip buddy update
5482 before setup, as it will be regenerated after recovery is complete.
5487 Description: If an OST is inactive, its locks might reference stale inodes.
5488 Details : lov_change_cbdata() must iterate over all namespaces, even if
5489 they are inactive to clear inode references from the lock.
5491 Severity : enhancement
5492 Frequency : occasional, if non-standard max_dirty_mb used
5494 Description: Client will block write RPCs if not enough grant
5495 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5496 then the client will block writes while waiting for another RPC
5497 to complete instead of consuming its dirty limit. With change
5498 we get improved performance when max_dirty_mb is small.
5500 Severity : enhancement
5501 Bugzilla : 3389, 6253
5502 Description: Add support for supplementary groups on the MDS.
5503 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5504 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5505 lookups for user supplementary groups into a cache.
5509 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5510 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5511 for liblustre O_EXCL is enforced by MDS.
5515 Description: readdir never returns NULL in liblustre.
5516 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5517 dentry in struct dirent.
5521 Frequency : liblustre only, depends on application IO pattern
5522 Description: liblustre clients evicted if not contacting servers
5523 Details : Don't put liblustre clients into the ping_evictor list, so
5524 they will not be evicted by the pinger ever.
5526 Severity : enhancement
5528 Description: Add ability to evict clients by NID from MDS.
5529 Details : By echoing "nid:$NID" string into
5530 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5531 $NID would be instantly evicted from this MDS and from all active
5532 OSTs connected to it.
5536 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5537 Details : lookup_it_finish() used to query file size from OSTs that was not
5542 Description: service threads change working directory to that of init
5543 Details : Starting lustre service threads may pin the working directory
5544 of the parent thread, making that filesystem busy. Threads
5545 now change to the working directory of init to avoid this.
5549 Frequency : during shutdown only
5550 Description: shutdown with a failed MDS or OST can cause unmount to hang
5551 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5556 Frequency : 2.6 only
5557 Description: chmod/chown may include an extra supplementary group
5558 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5559 supplementary group and if none is specified this is used.
5562 Bugzilla : 5479 (6816)
5564 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5565 Details : If lookup is in progress on a file that is unlinked we might try
5566 to revalidate the inode and fail in revalidate after lookup is
5567 complete and ll_file_open() enqueues the open again but
5568 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5571 Frequency : always, if lconf --abort_recovery used
5573 Description: lconf --abort_recovery fails with 'Operation not supported'
5574 Details : lconf was attempting to abort recovery on the MDT device and not
5577 ------------------------------------------------------------------------------
5579 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5584 Frequency : rare (only unsupported configurations with a node running as an
5586 Bugzilla : 6514, 5137
5587 Description: Mounting a Lustre file system on a node running as an OST could
5589 Details : OSTs now preallocates memory needed to write out data at
5590 startup, instead of when needed, to avoid having to
5591 allocate memory in possibly low memory situations.
5592 Specifically, if the file system is mounted on on OST,
5593 memory pressure could force it to try to write out data,
5594 which it needed to allocate memory to do. Due to the low
5595 memory, it would be unable to do so and the node would
5596 become unresponsive.
5598 Severity : enhancement
5600 Description: Addition of lconf --service command line option
5601 Details : lconf now accepts a '--service <arg>' option, which is
5602 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5604 Severity : enhancement
5606 Description: Failover mode is now the default for OSTs.
5607 Details : By default, OSTs will now run in failover mode. To return to
5608 the old behaviour, add '--failout' to the lmc line for OSTs.
5610 Severity : enhancement
5612 Description: Health checks are now provided for MDS and OSTs
5613 Details : Additional detailed health check information on MSD and OSTs
5614 is now provided through the procfs health_check value.
5617 Frequency : occasional, depends on IO load
5619 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5620 after numerous create/delete cycles
5621 Details : The ext3 inode allocation policy would not allocate new inodes
5622 very well on the OSTs because there are no new directories
5623 being created. Instead we look for groups with free space if
5624 the parent directories are nearly full.
5629 Description: Network or server problems during mount may cause partially
5630 mounted clients instead of returning an error.
5631 Details : The config llog parsing code may overwrite the error return
5632 code during mount error handling, returning success instead
5638 Description: MDS can fail to allocate large reply buffers
5639 Details : After long uptimes the MDS can fail to allocate large reply
5640 buffers (e.g. zconf client mount config records) due to memory
5641 fragmentation or consumption by the buffer cache. Preallocate
5642 some large reply buffers so that these replies can be sent even
5643 under memory pressure.
5647 Frequency : rare (liblustre)
5648 Description: fsx running with liblustre complained that using truncate() to
5649 extend the file doesn't work. This patch corrects that issue.
5650 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5651 ATTR_SIZE and lsm use in llu_setattr_raw.
5655 Frequency : rare, only 2.6 kernels
5656 Description: Unusual file access patterns on the MDS may result in inode
5657 data being lost in very rare circumstances.
5658 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5659 code in the 2.6 kernel caused the inode and/or EA data not to
5660 be read from disk, causing single-file corruption.
5664 Frequency : rare, only 2.6 filesystems using extents
5665 Description: Heavy concurrent write and delete load may cause data corruption.
5666 Details : It was possible under high-load situations to have an extent
5667 metadata block in the block device cache from a just-unlinked
5668 file overwrite a newly-allocated data block. We now unmap any
5669 metadata buffers that alias just-allocated data blocks.
5673 Frequency : filesystems with default stripe_count larger than 77
5674 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5675 Details : lconf specifies an inode size of 4096 bytes when the default
5676 stripe_count is larger than 77. This conflicts with the default
5677 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5678 this case to avoid pinning too much memory for large EAs.
5680 ------------------------------------------------------------------------------
5682 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5687 Frequency : rare (extremely heavy IO load with hundreds of clients)
5689 Description: Client is evicted, gets IO error writing to file
5690 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5691 caused clients to be evicted for AST timeouts. The fixes in
5692 bug 5192 mean we no longer need to have such short AST timeouts
5693 so ldlm_timeout has been increased.
5696 Frequency : occasional during --force or --failover shutdown under load
5697 Bugzilla : 5949, 4834
5698 Description: Server oops/LBUG if stopped with --force or --failover under load
5699 Details : a collection of import/export refcount and cleanup ordering
5700 issues fixed for safer force cleanup
5703 Frequency : only filesystems larger than 120 OSTs
5704 Bugzilla : 5990, 6223
5705 Description: lfs getstripe would oops on a very large filesystem
5706 Details : lov_getconfig used kfree on vmalloc'd memory
5709 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5710 Bugzilla : 6242, 6243
5711 Description: reading from files that had been truncated to a non-zero size
5712 but never opened returned no data
5713 Details : ll_file_read() reads zeros from no-object files to EOF
5718 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5719 Details : There's little harm in aborting MDS/OSS recovery and letting it
5720 try again, so I removed the LASSERT and return an error instead.
5722 Severity : enhancement
5724 Description: New debugging infrastructure for tracking down data corruption
5725 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5726 (b) cover more of the client-side code path, and (c) try to narrow
5727 down where problems occurred
5731 Bugzilla : 3819, 4364, 4397, 6313
5732 Description: Racing close and eviction MDS could cause assertion in mds_close
5733 Details : It was possible to get multiple mfd references during close and
5734 client eviction, leading to one thread referencing a freed mfd.
5736 Severity: : enhancement
5737 Bugzilla : 3262, 6359
5738 Description: Attempts to reconnect to servers are now more aggressive.
5739 Details : This builds on the enhanced upcall-less recovery that was added
5740 in 1.4.2. When trying to reconnect to servers, clients will
5741 now try each server in the failover group every 10 seconds. By
5742 default, clients would previously try one server every 25 seconds.
5747 Description: After recovery, certain operations trigger a failed
5748 assertion on a client.
5749 Details : Failing over an mds, using lconf -d --failover, while a
5750 client was doing a readdir() call would cause the client to
5751 LBUG after recovery completed and the readdir() was resent.
5753 Severity : enhancement
5755 Description: Default groups are now added by lconf
5756 Details : You can now run lconf --group <servicename> without having to
5757 manually add groups with lmc.
5760 Frequency : occasional
5762 Description: Nodes with an elan id of 0 trigger a failed assertion
5765 Frequency : always when accessing e.g. tty/console device nodes
5767 Description: tty and some other devices nodes cannot be used on lustre
5768 Details : file's private_data field is used by device data and lustre
5769 values in there got lost. New field was added to struct file to
5770 store fs-specific private data.
5773 Frequency : when exporting Lustre via NFS
5775 Description: NFSD failed occasionally when looking up a path component
5776 Details : NFSD is looking up ".." which was broken in ext3 directories
5777 that had grown large enough to become hashed.
5780 Frequency : Clusters with multiple interfaces not on the same subnet
5782 Description: Nodes will repeatedly try to reconnect to an interface which it
5783 cannot reach and report an error to the log.
5784 Details : Extra peer list entries will be created by lconf with some peers
5785 unreachable. lconf now validates the peer before adding it.
5788 Frequency : Only if a default stripe is set on the filesystem root.
5790 Description: Setting a default stripe on the filesystem root prevented the
5791 filesystem from being remounted.
5792 Details : The client was sending extra request flags in the root getattr
5793 request and did not allocate a reply buffer for the dir EA.
5796 Frequency : occasional, higher if lots of files are accessed by one client
5797 Bugzilla : 6159, 6097
5798 Description: Client trips assertion regarding lsm mismatch/magic
5799 Details : While revalidating inodes the VFS looks up inodes with ifind()
5800 and in rare cases can find an inode that is being freed.
5801 The ll_test_inode() code will free the lsm during ifind()
5802 when it finds an existing inode and then the VFS later attaches
5803 this free lsm to a new inode.
5807 Bugzilla : 6422, 7030
5808 Description: MDS deadlock between mkdir and client eviction
5809 Details : Creating a new file via mkdir or mknod (starting a transaction
5810 and getting the ns lock) can deadlock with client eviction
5811 (gets ns lock and trying to finish a synchronous transaction).
5814 Frequency : occasional
5815 Description: While starting a server, the fsfilt_ext3 module could not be
5817 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5818 kernels. Previously, lconf would still use the ext3 name
5819 when trying to load modules. Now, it will correctly use
5820 ext3 on 2.4 and ldiskfs on 2.6.
5822 Severity : enhancement
5823 Description: The default stripe count has been changed to 1
5824 Details : The interpretation of the default stripe count (0, to lfs
5825 or lmc) has been changed to mean striping across a single
5826 OST, rather than all available. For general usage we have
5827 found a stripe count of 1 or 2 works best.
5829 Severity : enhancement
5830 Description: Add support for compiling against Cray portals.
5831 Details : Conditional compiling for some areas that are different
5835 Frequency : occasional
5836 Bugzilla : 6409, 6834
5837 Description: Creating files with an explicit stripe count may lead to
5838 a failed assertion on the MDS
5839 Details : If some OSTs are full or unavailable, creating files may
5840 trigger a failed assertion on the MDS. Now, Lustre will
5841 try to use other servers or return an error to the
5845 Frequency : occasional
5847 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5848 caused lock timeout during readahead (since 1.4.2).
5849 Details : Processes doing readahead might match a lock that hasn't been
5850 granted yet if there are overlapping and conflicting lock
5851 requests. The readahead process waits on ungranted lock
5852 (original lock is CBPENDING), while OST waits for that process
5853 to cancel CBPENDING read lock and eventually evicts client.
5855 Severity : enhancement
5857 Description: Initial enabling of flock support for clients
5858 Details : Implements fcntl advisory locking and file status functions.
5859 This feature is provided as an optional mount flag (default
5860 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5861 locking are implemented yet, and those that are are not guaranteed
5862 to be completely correct in production environments.
5863 mount -t lustre -o [flock|noflock] ...
5866 Frequency : occasional
5868 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5869 assertion in the ext3 JBD (journaling) layer.
5870 Details : The b_committed_data struct is protected by the big kernel lock
5871 in 2.4 kernels, serializing journal_commit_transaction() and
5872 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5873 access to this struct. In 2.6 kernels there is finer grained
5874 locking to improve SMP performance of the JBD layer.
5878 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5883 Description: merge and cleanup kernel patches.
5886 -----------------------------------------------------------------------------
5888 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5890 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5891 versions of Lustre. Please update all servers and clients to
5892 version 1.4.2 or later at the same time. You must also run
5893 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5894 to update the configuration logs.
5896 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5897 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5898 - swab llog records of type '0' so we get proper header size/idx (5861)
5899 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5900 - fix rename of one directory over another leaking an inode (5953)
5901 - avoid SetPageDirty on 2.6 (5981)
5902 - don't re-add just-being-destroyed locks to the waiting list (5653)
5903 - when creating new directories, inherit the parent's custom
5904 striping settings if present parent (3048)
5905 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5906 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5907 - don't hold client locks on temporary worklist from l_lru (5666)
5908 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5909 - automatically evict dead clients (5921)
5910 - Update file size properly in create+truncate+fstat case (6196)
5911 - Do not unhash mountpoint dentries, do not allow removal of
5913 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5914 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5915 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5916 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5917 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5918 - Correctly handle reads of files with no objects (6243)
5919 - lctl recover will also mark a device active if deactivate used (5933)
5921 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5922 - allow --write-conf on an MDS with different nettype than client (5619)
5923 - don't write config llogs to MDS for mounts not from that MDS (5617)
5924 - lconf should create multiple TCP connections from a client (5201)
5925 - init scripts are now turned off by default; run chkconfig --on
5926 lustre and chkconfig --on lustrefs to use them
5927 - upcalls are no longer needed for clients to recover to failover
5929 - add --abort-recovery option to lconf to abort recovery on device
5931 - add support for an arbitrary number of OSTs (3026)
5932 - Quota support protocol changes.
5933 - forward compatibility changes to wire structs (6007)
5934 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5935 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5936 - improved "lustre status" script
5937 - initialize blocksize for non-regular files (6062)
5938 - added --disable-server and --disable-client configure options (5782)
5939 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5940 - Vanilla 2.4.29 support
5941 - increase maximum number of obd devices to 520 (6242)
5942 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5943 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5944 SLES 9 2.6.5 kernels
5945 - limit stripes per file to 160 (the maximum EA size) (6093)
5947 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5950 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5951 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5952 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5953 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5954 - small ext3 extents cleanups and fixes (5733)
5955 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5956 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5957 - avoid needless client->OST connect, fix handle mismatch (5317)
5958 - fix DLM error path that led to out-of-sync client, long delays (5779)
5959 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5960 - fix several locking issues related to i_size (5492,5624,5654,5672)
5961 - don't move pending lock onto export if it is already evicted (5683)
5962 - fix kernel oops when creating .foo in unlinked directory (5548)
5963 - fix deadlock in obdfilter statistics vs. object create (5811)
5964 - use time_{before,after} to avoid timer jiffies wrap (5882)
5965 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5966 - Do not leak request if server was not able to process it (5154)
5967 - If mds_open unable to find parent dir, make that negative lookup(5154)
5968 - don't create new directories with extent-mapping (5909, 5936)
5970 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5971 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5972 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5973 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5974 - support for RHEL4 kernel on i686 (5773)
5975 - provide error messages when incompatible logs are encountered (5898)
5977 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5978 * version 1.4.0.10 (1.4.1 release candidate 1)
5980 - don't keep a lock reference when lock is not granted (4238)
5981 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5982 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5983 - drop import inflight refcount on signal_completed_replay error (5255)
5984 - unlock page after async write error during send (3677)
5985 - handle missing objects in filter_preprw_read properly (5265)
5986 - no transno return for symlink open, don't save no-trasno open (3440)
5987 - don't try to complete elan receive that already failed (4012)
5988 - free RPC server reply state on error (5406)
5989 - clean up thread from ptlrpc_start_thread() on error (5160)
5990 - readahead could read extra page into cache that wasn't ejected (5388)
5991 - prevent races in class_attach/setup/cleanup/detach (5260)
5992 - don't dereference de->d_inode after l_dput of de (5458)
5993 - use "int" for stripe value returned from lock_to_stripe (5544)
5994 - mballoc allocation and error-checking fixes in 2.6 (5504)
5995 - block device patches to fix I/O request sizes in 2.6 (5482)
5996 - look up hostnames for IB nals (5602)
5997 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5998 - don't start multiple acceptors for the same port (5277)
5999 - fix incorrect LASSERT in mds_getattr_name (5635)
6000 - export a proc file for general "ping" checking (5628)
6001 - fix "lfs check" to not block when the MDS is down (5628)
6003 - service request history (4965)
6004 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6005 - create an "evict_client" /proc entry on OSTs, like the MDS has
6006 - fix mount usage message, return errors per mount(8) (5168)
6007 - change grep [] to grep "[]" in tests so they work in more UMLs
6008 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6009 - remove ancient LOV_MAGIC_V0 EA support (5047)
6010 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6011 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6012 - allow readahead allocations to fail when low on memory (5383)
6013 - mmap locking landed again, after considerable improvement (2828)
6014 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6016 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6019 - send OST transaction number in read/write reply to free req (4966)
6020 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6021 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6022 - the watchdog thread now runs as interruptible (5246)
6023 - flock/lockf fixes (but it's still disabled, pending 5135)
6024 - don't use EXT3 constants in llite code (5094)
6025 - memory shortage at startup could cause assertion (5176)
6027 - reorganization of lov code
6028 - single portals codebase
6030 - add extents/mballoc support (5025)
6031 - direct I/O reads in the obdfilter (4048)
6032 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6034 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6037 - send OST transaction number in read/write reply to free req (4966)
6038 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6039 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6040 - the watchdog thread now runs as interruptible (5246)
6041 - handle missing objects in filter_preprw_read properly (5265)
6042 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6043 - drop import inflight refcount on signal_completed_replay error (5255)
6044 - unlock page after async write error during send (3677)
6045 - return original error code on reconstructed replies (3761)
6046 - no transno return for symlink open, don't save no-trasno open (3440)
6048 - add pid to ldlm debugging output (4922)
6049 - bump the watchdog timeouts -- we can't handle 30sec yet
6050 - extra debugging for orphan dentry/inode bug (5259)
6052 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6055 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6056 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6057 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6058 - fix NULL dereference in /proc/sys/portals/routes (4827)
6059 - allow failed mdc_close() operations to be interrupted (4561)
6060 - stop precreate on OST before MDS would time out on it (4778)
6061 - don't send partial-page writes before EOF from client (4410)
6062 - discard client grant for sub-page writes on large-page clients (4520)
6063 - don't free dentries not owned by NFS code, check generation (4806)
6064 - fix lsm leak if mds_create_objects() fails (4801)
6065 - limit debug_daemon file size, always print CERROR messages (4789)
6066 - use transno after validating reply (3892)
6067 - process timed out requests if import state changes (3754)
6068 - update mtime on OST during writes, return in glimpse (4829)
6069 - add mkfsoptions to LDAP (4679)
6070 - use ->max_readahead method instead of zapping global ra (5039)
6071 - don't interrupt __l_wait_event() during strace
6073 - add software watchdogs to catch hung threads quickly (4941)
6074 - make lustrefs init script start after nfs is mounted
6075 - fix CWARN/ERROR duplication (4930)
6076 - return async write errors to application if possible (2248)
6077 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6078 - print NAL number in %x format (4645)
6079 - update barely-supported suse-2.4.21-171 series (4842)
6080 - support for sles 9 %post scripts
6081 - support for building 2.6 kernel-source packages
6082 - support for sles km_* packages
6084 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6087 - ignore -ENOENT errors in osc_destroy (3639)
6088 - notify osc create thread that OSC is being cleaned up (4600)
6089 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6090 - reconstruct ost_handle() like mds_handle() (4657)
6091 - create a new thread to do import eviction to avoid deadlock (3969)
6092 - let lconf resolve symlinked-to devices (4629)
6093 - don't unlink "objects" from directory with default EA (4554)
6094 - hold socknal file ref over connect in case target is down (4394)
6095 - allow more than 32000 subdirectories in a single directory (3244)
6096 - fix blocks count for O_DIRECT writes (3751)
6097 - OST returns ENOSPC from object create when no space left (4539)
6098 - don't send truncate RPC if file size isn't changing (4410)
6099 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6100 - bind to privileged port in socknal and tcpnal (3689)
6102 - rate limit CERROR/CWARN console message to avoid overload (4519)
6103 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6104 - basic mmap support (3918)
6105 - kernel patch series update from b1_4 (4711)
6107 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6110 - avoid crash during MDS cleanup with OST shut down (2775)
6111 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6112 - don't use bad inodes on the MDS (3744)
6113 - dynamic object preallocation to improve recovery speed (4236)
6114 - don't hold spinlock over lock dumping or change debug flags (4401)
6115 - don't zero obd_dev when it is force cleaned (3651)
6116 - print grants to console if they go negative (4431)
6117 - "lctl deactivate" will stop automatic recovery attempts (3406)
6118 - look for existing locks in ldlm_handle_enqueue() (3764)
6119 - don't resolve lock handle twice in recovery avoiding race (4401)
6120 - revalidate should check working dir is a directory (4134)
6122 - don't always mark "slow" obdfilter messages as errors (4418)
6124 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6127 - don't close LustreDB during write_conf until it is done (3860)
6128 - fix typo in lconf for_each_profile (3821)
6129 - allow dumping logs from multiple threads at one time (3820)
6130 - don't allow multiple threads in OSC recovery (3812)
6131 - fix debug_size parameters (3864)
6132 - fix mds_postrecov to initialize import for llog ctxt (3121)
6133 - replace config semaphore with spinlock (3306)
6134 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6135 - don't allow enqueue to complete on a destroyed export (3822)
6136 - down write_lock before checking llog header bitmap (3825)
6137 - recover from lock replay timeout (3764)
6138 - up llog sem before sending rpc (3652)
6139 - reduce ns lock hold times when setting kms (3267)
6140 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6141 - fix NULL deref and obd_dev leak on setup error (3312)
6142 - replace some LBUG about llog ops with error handling (3841)
6143 - don't match INVALID dentries from d_lookup and spin (3784)
6144 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6145 - fix invalid assertion in ptlrpc_set_wait (3880)
6147 - add libwrap support for the TCP acceptor (3996)
6148 - add /proc/sys/portals/routes for non-root route listing (3994)
6149 - allow setting MDS UUID in .xml (2580)
6150 - print the stack of a process that LBUGs (4228)
6152 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6155 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6156 - make sure to unset replay flag from failed open requests (3440)
6157 - if default stripe count is 0, use OST count for inode size (3636)
6158 - update parent mtime/ctime on client for create/unlink (2611)
6159 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6160 - free recovery state on server during a forced cleanup (3571)
6161 - unregister_reply for resent reqs (3063)
6162 - loop back devices mounting and status check on 2.6 (3563)
6163 - fix resource-creation race that can provoke i_size == 0 (3513)
6164 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6165 - more debugging for page-accounting assertion (3746)
6166 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6167 - avoid deadlock after precreation failure (3758)
6168 - fix race and lock order deadlock in orphan handling (3450, 3750)
6169 - add validity checks when grabbing inodes from l_ast_data (3599)
6171 - add /proc/.../recovery_status to obdfilter (3428)
6172 - lightweight CDEBUG infrastructure, debug daemon (3668)
6173 - change default OSC RPC parameters to be better on small clusters
6174 - turn off OST read cache for files smaller than 32MB
6175 - install man pages and include them in rpms (3100)
6176 - add new init script for (un)mounting lustre filesystems (2593)
6177 - run chkconfig in %post for init scripts (3701)
6178 - drop scimac NAL (unmaintained)
6180 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6183 - clean kiobufs before and after use (3485)
6184 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6185 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6186 - change init script to fail more clearly if not run as root (1528)
6187 - allow clients to reconnect during replay (1742)
6188 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6189 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6190 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6191 - fix cond_resched (3554)
6192 - only evict unfinished clients after recovery (3515)
6193 - allow bulk resend, prevent data loss (3570)
6194 - dynamic ptlrpc request buffer allocation (2102)
6195 - don't allow unlinking open directory if it isn't empty (2904)
6196 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6197 - remove extraneous obd dereference causing LASSERT failure (3334)
6198 - don't use get_cycles() when creating temp. files on the mds (3156)
6199 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6200 - handle EEXIST for set-stripe, set proper directory name (3336)
6202 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6203 - fix ksocknal_fmb_callback() error messages (2918)
6205 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6208 - don't copy lvb into (possibly NULL) reply on error (2983)
6209 - don't deref dentry after dput, don't free lvb on error (2922)
6210 - use the kms to determine writeback rpc length (2947)
6211 - increment oti_logcookies when osc is inactive (2948)
6212 - update client's i_blocks count via lvb messages (2543)
6213 - handle intent open/close of special files properly (1557)
6214 - mount MDS with errors=remount-ro, like obdfilter (2009)
6215 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6216 - don't use cancelling-locks' kms values (2947)
6217 - use highest lock extent for kms, not last one (2925)
6218 - don't dereference ERR_PTR() dentry in error handling path (3107)
6219 - fix thread race in portals_debug_dumplog() (3122)
6220 - create lprocfs device entries at setup instead of at attach (1519)
6221 - common AST error handler, don't evict client on completion race (3145)
6222 - zero nameidata in detach_mnt in 2.6 (3118)
6223 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6224 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6225 - initialize RPC timeout timer earlier for 2.6 (3219)
6226 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6227 - print nal/nid for unknown nid (3258)
6228 - additional checks for oscc recovery before doing precreate (3284)
6229 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6230 - don't crash in mdc_close for bad permissions on open (3285)
6231 - zero i_rdev for non-device files (3147)
6232 - clear page->private before handing to FS, better assertion (3119)
6233 - tune the read pipeline (3236)
6234 - fix incorrect decref of invalidated dentry (2350)
6235 - provide read-ahead stats and refine rpc in flight stats (3328)
6236 - don't hold journal transaction open across create RPC (3313)
6237 - update atime on MDS at close time (3265)
6238 - close LDAP connection when recovering to avoid server load (3315)
6239 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6240 - don't leak open file on MDS after open resend (3325)
6241 - serialize filter_precreate and filter_destroy_precreated (3329)
6242 - loop device shouldn't call sync_dev() for nul device (3092)
6243 - clear page cache after eviction (2766)
6244 - resynchronize MDS->OST in background (2824)
6245 - refuse to mount the same filesystem twice on same mountpoint (3394)
6246 - allow llmount to create routes for mounting behind routers (3320)
6247 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6248 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6249 - fix rare problem with rename on htree directories (3417)
6251 - allow default OST striping configuration per directory (1414)
6252 - fix compilation for qswnal for 2.6 kernels (3125)
6253 - increase maximum number of MDS request buffers for large systems
6254 - change liblustreapi to be useful for external progs like lfsck (3098)
6255 - increase local configuration timeout for slow disks (3353)
6256 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6258 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6261 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6262 - don't overwrite extent policy data in reply if lock was blocked (2901)
6263 - drop filter export grants atomically with removal from device (2663)
6264 - del obd_self_export from work_list in class_disconnect_exports (2908)
6265 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6266 - swab reply message in mdc_close, other PPC fixes (2464)
6267 - fix destroying of named logs (2325)
6268 - overwrite old logs when running lconf --write_conf (2264)
6269 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6270 - fix race in target_handle_connect (2898)
6271 - mds_reint_create() should take same inode create lock (2926)
6272 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6273 - don't close files for self_export to avoid uninitialized obd (2936)
6274 - allow MDS with the same name as client node (2939)
6275 - hold dentry reference for closed log files for unlink (2325)
6276 - reserve space for all logs during transactions (2059)
6277 - don't evict page beyond end of stripe extent (2925)
6278 - don't oops on a deleted current working directory (2399)
6279 - handle hard links to targets without a parent properly (2517)
6280 - don't dereference NULL lock when racing during eviction (2867)
6281 - don't grow lock extents when lots of conflicting locks (2919)
6283 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6286 - account for cache space usage on clients to avoid data loss (974)
6287 - lfsck support in lustre kernel code (2349)
6288 - reduce journal credits needed for BRW writes (2370)
6289 - orphan handling to avoid losing space on client/server crashes
6290 - ptlrpcd can be blocked, stopping ALL progress (2477)
6291 - use lock value blocks to assist in proper KMS, faster stat (1021)
6292 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6293 - recovery for initial connections (2355)
6294 - fixes for mds_cleanup_orphans (1934)
6295 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6296 - block all file creations until orphan recovery completes (1901)
6297 - client remove rq_connection from request struct (2423)
6298 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6299 - recovery timer race (2670)
6300 - mdc_close recovey bug (2532)
6301 - ptlrpc cleanup bug (2710)
6302 - mds timeout on local locks (2588)
6303 - namespace lock held during RPCs (2431)
6304 - handle interrupted sync write properly (2503)
6305 - don't try to handle a message that hasn't been replied to (2699)
6306 - client assert failure during cleanup after abort recovery (2701)
6307 - leak mdc device after failed mount (2712)
6308 - ptlrpc_check_set allows timedout requests to complete (2714)
6309 - wait for inflight reqs when ptlrpcd finishes (2710)
6310 - make sure unregistered services are removed from the srv_list
6311 - reset bulk XID's when resending them (caught by 1138 test)
6312 - unregister_bulk after timeout
6313 - fix lconf error (2694)
6314 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6315 - readahead locks pages, leaves pending causing memory pressure (2673)
6316 - increase OST request buffers to 4096 on large machines (2729)
6317 - fix up permission of existing directories in simple_mkdir (2661)
6318 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6319 - don't assign transno to errored transactions (2742)
6320 - don't delete objects on OST if given a bogus objid from MDS (2751)
6321 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6322 - if rq_no_resend, then timeout request after recovery (2432)
6323 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6324 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6325 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6326 - MDS recovery shouldn't delete objects that it has given out (2730)
6327 - if enqueue arrives after completion, don't clobber LVB (2819)
6328 - don't unlock pages twice when trigger_group_io returns error (2814)
6329 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6330 - don't write pages to disk if there was an error (1450)
6331 - don't ping imports that have recovery disabled (2676)
6332 - take buffered bytes into account when balancing socknal conn (2817)
6333 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6334 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6335 - remove little-endian swabbing of llog records (1987)
6336 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6337 - retry reposting request buffers if they fail (1191)
6338 - grow extent at grant time to avoid granting a revoked lock (2809)
6339 - lock revoke doesn't evict page if covered by a second lock (2765)
6340 - disable VM readahead to avoid reading outside lock extents (2805)
6342 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6343 - updated kernel patches for hp-2.4.20 kernel (2681)
6345 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6348 - fix truncated write corruption (2366)
6349 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6351 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6352 - ensure len(uuid) < 37 in lmc (1171)
6353 - fix ia64 OOPS in llog_test (2255)
6354 - zero end of page at obdfilter for partial page writes (2648)
6355 - don't leave stale dentries around after renames (bug 2428)
6356 - fix timeouts when evicting a client with a single lock held (2642)
6357 - set deadline for the initial HELLO message to drain (2634)
6358 - print out dotted-quad IP addresses in the socknal (2302)
6360 - additional debugging for MDS client eviction problem (2443)
6361 - fix mkfsoptions support for osts (2603, 2604)
6363 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6366 - add series for the vanilla 2.6.0 kernel
6367 - add series for the vanilla 2.4.24 kernel
6368 - add series for a cray x86/64 UL kernel drop
6369 - fix xattr patches for the vanilla 2.4.19 series
6371 - generate true UUIDs in lmc (1171)
6372 - have portals stack dumping break in UML (2466)
6373 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6374 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6375 - wake callback waiting threads on client eviction (2460)
6376 - Add --ptldebug and --subsystem to lmc (1719)
6377 - update assertion to allow safe interrupt allocation
6378 - set rq_no_resend for cancel requests (2432)
6379 - recalculate ptlrpcd timeout after resend (2494)
6380 - call vfs_rmdir when removing pending directories (2368)
6381 - fix renaming a file to itself (2429)
6382 - lmc creates a default one-stripe lov (2454)
6383 - expand procfs space to handle large clusters (2326)
6384 - increase UML stack to avoid overflow
6385 - update lconf's list of debug and subsystem masks
6386 - fix lfs find --obd (2510)
6387 - /proc tunable for disabling filter read caching (2591)
6388 - stop rpm packages from altering slapd.conf (2301)
6389 - disable nagle in the socknal under 0conf (2578)
6390 - choose mds inode size based on stripe count (2572)
6391 - fix kernel-source rpm problems (2516)
6393 - add --disable-doc to avoid pdf generation (2421)
6394 - update documentation, tests, type-os, comments
6395 - avoid format warnings on ia64
6396 - remove the TOE NAL
6397 - tiny code cleanups by removing unused fields
6399 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6402 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6403 - remove the most blatant lies from BUILDING file (2371)
6404 - change default debug level to reasonable production setting
6405 - reduce client side cache size to reduce cache flush time
6406 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6407 - make TCP zerocopy and pinger support enabled by default (2476)
6408 - sync writes completed after process exits caused crashes (2319)
6409 - maintain correct mount count on the MDS (2356)
6410 - backout 1557, because 2316 wasn't really fixed
6411 - better file I/O statistics gathering in /proc
6412 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6413 - another kernel patch to fix zero-copy TCP function export
6414 - don't take duplicate lock when processing re-sent getattr (2420)
6415 - lctl uses obd_self_export instead of creating new conn (2353)
6416 - MDS/OST recovery case which requires object creation asserted (2425)
6417 - move lfs from /usr/sbin to /usr/bin in packages
6418 - fix race between mds_client_add and mds_client_free (2417)
6419 - use kmalloc instead of slabs in portals (2430)
6420 - don't create duplicate records when a failover MDS is present (2442)
6421 - remove unnecessary mount age check (2332)
6422 - don't remove directory inodes from locks prematurely (2451)
6423 - don't break if MDS service name is the same as hostname (2103)
6424 - fix races in client write RPC generation when cache full (2482)
6426 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6429 - remove now-unused request->rq_obd (278)
6430 - if an allocation fails, print out how much memory we've used (1933)
6431 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6432 - add the "configurable stack size" patch to most series files (1256)
6433 - ability to write large log records, for 100+ OST configs (2306)
6434 - fix NULL deref when filter_prep fails (2314)
6435 - fix operator precedence error in filter_sync
6436 - dynamic allocation of socknal TX descriptors (2315)
6437 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6438 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6439 - ensure that CWARN messages in Portals always get to the syslog
6440 - __init/__exit are not for prototype decls (ldlm_init/exit)
6441 - x86-64 compile warning fixes
6442 - fix gateway LMC keyword conflict (2318)
6443 - fix MDS lock inversions in getattr/reint paths (1844)
6444 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6445 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6446 - DLM race condition prevented some lock evictions (2328)
6447 - ENOMEM detection and retry on socknal sends (2230)
6448 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6449 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6450 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6451 - better error messages when a client is rejected during recovery (1505)
6452 - avoid cancelling locks which were never granted, after failure (2330)
6453 - fix i_sem/journal inversion in mds_client_add (2333)
6454 - fix truncate/getattr lock cycle deadlock (2334)
6455 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6456 - fix two rare exit paths which could leak an l_lock() ref (2321)
6457 - fencepost error in MDS/OST orphan recovery (2226)
6458 - make log record alignment 8 bytes (1988)
6459 - lstripe now fails when requested offset > ost_count (2237)
6460 - ensure that all kernel series have a complete list.h (1607)
6461 - fix crashes in special-file operations (2316)
6462 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6464 - allow configurable automake binary, for testing new versions
6465 - small update to the lfs documentation
6467 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6469 * fix negative export reference count in fsfilt_sync (2312)
6471 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6472 * release candidate 0.9.1
6474 - orphans are moved into the PENDING directory for possible recovery
6475 - replayed opens now open by fid for orphan/rename safety (1042)
6476 - last close of an orphan inode generates a transno (683)
6477 - chdir() and mount() now pin the directory entry (1020)
6478 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6479 - discard very old requests without processing them (1502)
6480 - remove some common, well-understood CERRORs (1505)
6481 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6482 - clear "grant" flags in OST replies until OST grant code lands (1644)
6483 - fix read performance by not clobbering i_blksize on client (1598)
6484 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6485 - make LRU size a /proc tunable, clears locks when reduced (707)
6486 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6487 - server locks take references on exports now (1558)
6488 - build fixes for 2.4.20-rh trees (1663)
6489 - return an error from lov_create if all OSCs are inactive (1751)
6490 - fix import levels when a reconnect happens without a timeout (1597)
6491 - exit early from mds_open if we get a lookup error (1749)
6492 - partial page read at EOF wouldn't wait for disk before sending (1642)
6493 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6494 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6495 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6496 - write/truncate lock inversion (1639)
6497 - Don't auto-load obdclass, portals modules during cleanup (1495)
6498 - fix timestamps from jumping to "now" (1763)
6499 - extra journal assertions (1648)
6500 - add an extra multiunlink test (1771)
6501 - fix read_record/write_record API (1776)
6502 - fix leak of offset_extent, possible incorrect i_size later (1772)
6503 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6504 - Debugging for the kqswnal_get_idle_tx problems (1820)
6505 - Allow recovery to be attempted multiple times (1536)
6506 - Write out MDS last_rcvd file after it is first created (1600)
6507 - Fix tx_descriptor leak in failed transmit situations (1827)
6508 - ext3 journaling fixes for assertion failure after IO error (1871)
6509 - class_export_put() on freed export after completion AST error (1896)
6510 - Fix revalidate looping in VFS (1322)
6511 - Don't access a freed export during MDS_REINT timeout (1521)
6512 - Add open-unlink recovery support on the MDS (1673,1764)
6513 - Return an error if no MDS data was read from last_rcvd (1946)
6514 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6515 - Don't setup a disk device that doesn't match exported UUID (317)
6516 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6517 - avoid committing NULL handle in force close
6518 - local.sh is now a one-stripe LOV configuration
6519 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6520 - fix ext3 htree duplicate directory entry corruption (1516)
6521 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6522 - update to new LOV EA format (2097)
6523 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6525 2003-06-15 Phil Schwan <phil@clusterfs.com>
6528 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6529 - per-import/export recovery handling (958, 931, 959)
6530 - multiple last-rcvd slots, for serving multiple FSes (949)
6531 - connections are again shared between multiple imp/exports (963, 964)
6532 - "umount -f" would hang if any requests needed to be sent (393, 978)
6533 - avoid pinning large req buffer by copying for queued messages (989)
6534 - add "uuid" to "lctl device" command to help upcalls (991)
6535 - "open" RPCs with transnos would confuse recovery counters (1037)
6536 - do proper endian conversion of all wire messages (288, 340, 891)
6537 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6538 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6539 - fix ptlrpc_connection leak in target_handle_connect (1174)
6540 - fix import refcounting bug in OST and MDS cleanup (1134)
6541 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6542 - fix very unlikely obd_types race condition (501)
6543 - remove osc_open hack for echo_client (1187)
6544 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6545 - a failure in read_inode2 leads to deadlock (1139)
6546 - cancel ack-locks as soon as transaction is committed (1072)
6547 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6548 - make sure to commitrw after any preprw to avoid deadlock (1162)
6549 - failing to execute a file in a lustre FS would lock inode (1203)
6550 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6551 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6552 - obd_uuid2tgt would walk off the end of the list (1255)
6553 - on IA64 the osc would give portals incorrect bulk size (1258)
6554 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6555 - fix lock inversion caused by new llite matching code (1282)
6556 - limit the number of dirty pages on a client to 10MB (1286)
6557 - timed out locks were not being corrected cancelled (1289)
6558 - fix O_DIRECT above 4GB on IA-32 (1292)
6559 * major user-visible changes
6560 - fail out/fail over policy now controlled by the upcall (993)
6562 - add OBD_PING to check server availability and failure (954)
6563 - lustre messages are now sent in sending host order (288, 340, 891)
6564 - add eadatalen to MDS getattr reply (340)
6565 - OST read replies may contain second buffer, with per-page status (593)
6567 2003-03-11 Phil Schwan <phil@clusterfs.com>
6570 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6571 - failed open()s could cause deadlock; fixed (867, 869)
6572 - stop cancelling OST locks when files are closed (481)
6573 - overlapping XID spaces caused network corruption (851, 853)
6574 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6575 - setattr_raw added, to do single-RPC, server-side setattrs
6576 - lmc/lconf syntax change for OST UUIDs
6577 - fix crashy race condition between ptlrpc_free_req and osc_close
6578 - don't use request in mdc_enqueue if we hit a timeout (889)
6579 - don't set the inode i_size for regular files from the MDS (896)
6580 - handle out of order completion AST (842)
6581 - don't LBUG if a lock request times out after receiving AST (913)
6582 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6583 - if a bad lock AST arrives, send an error instead of dropping entirely
6584 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6585 - fix leak in bulk IO when only partially completed (899, 900, 926)
6586 - fix O_DIRECT for ia64 (55)
6587 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6588 - C-z after timeout could hang a process forever; fixed (977)
6590 - client-side I/O cache (678, 924, 929, 941, 970)
6592 - READPAGE and SETATTRs which don't take server-side locks get
6595 2003-02-11 Phil Schwan <phil@clusterfs.com>
6598 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6599 - Fix fsfilt_extN_readpage() to read a full page of directory
6600 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6601 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6602 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6603 - Most intent operations were removed, in favour of a new RPC mode
6604 that does a single RPC to the server and bypasses most of the VFS
6605 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6606 - Aggressively cancel local locks on DLM servers
6607 - mds_reint_unlink sends EA to the client if it's the last nlink.
6608 client uses that EA to unlink OST objects.
6609 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6610 - recursive symlinks were fixed (439)
6611 - fixed NULL deref in DEBUG_REQ
6612 - filter_update_lastobjid no longer calls sync, which annoyed extN
6613 - fixed multi-client small-writes to a single file problem (445)
6614 - fixed mtime updates during file writes (607)
6615 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6616 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6617 - fixed Linux OST size reporting problem (444, 656)
6618 - OST now updates object mtime with writes or setattr (607, 619)
6619 - client verifies file size before zeroing page past EOF (445)
6620 - OST now writes last allocated objid to disk with allocation (108)
6621 - LOV on echo now works (409)
6623 - mds_reint_unlink sends a new buffer, with the EA included. this
6624 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6625 set if a regular file was being unlinked, and it was the last link
6626 - use PtlGet from the target for bulk writes (315)
6627 - OST now updates object mtime with writes or setattr (607, 619)
6628 - LDLM now has a grant-time callback to revalidate locked items, if
6630 - Many MDS operations were reorganized to combat race conditions
6632 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6633 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6635 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6638 - Fully reactivate OST imports after reconnection (512, others)
6639 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6640 - More graceful error handling for truncating on dead OST (515)
6641 - Don't error out unless we're actually accessing dead stripes (474)
6642 - Fix garbage sizes when stripes are missing (410)
6643 - LRU counters were broken, causing constant lock purge (433, 432)
6644 - garbage on read from stripes with failed OSTs (441)
6645 - mark OSCs as active before reconnecting during recovery (438)
6646 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6647 - lfind did not preserve OST order in output (443)
6648 - symlinks cause hung clients, incorrect data (439)
6649 - stop dereferencing request after dropping refcount (457)
6650 - don't LASSERT(spin_is_locked) on non-SMP (455)
6651 - fixes for many rename() bugs
6652 - fstat didn't correctly synchronize attributes (399)
6653 - server must handle lock cancellation during blocking AST prep (487)
6654 - bulk descriptors were free()d too soon (511)
6655 - fix paths in lconf, which would load incorrect modules (451, 507)
6656 - fix confusing lconf 'host not found' error message (386)
6657 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6658 - fix race condition in mdc_blocking_ast() for inode access (526)
6659 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6660 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6661 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6662 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6663 - fix lstripe in several ways (kernel side) (527)
6664 - fix request leak in ldlm_cli_enqueue (262)
6665 - incorrect OSC was marked inactive after OST failure
6666 - call mds_fs_cleanup before unmounting filesystem (524)
6667 - fix races between taking ns_lock and ldlm_lock_change_resource
6668 - fix races updating LOV export open file list
6669 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6670 - fix recovery NULL deref in ldlm_cli_cancel_unused
6671 - fix some DLM races by using new hash table for lock handles (419)
6672 - permit the client to specify desired inodes, at replay
6673 - duplicate requests when we queue them for replay reintegration
6674 - fix last_rcvd offset calculation
6675 - sync after each recovered transaction, so we always make progress
6676 - never, not always, ERESTART requests without transnos
6677 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6679 - skip replay if the MDS says that the client is already connected
6680 - don't check for a recovery-enabled export to match lctl's UUID
6681 - don't INC_USE_COUNT for phantom exports
6682 - don't crash when cleaning up phantom exports (567)
6683 - don't double-finish or set replay data for errored mdc_open requests
6684 - abort requests when they time out, so we don't get old replies
6685 - send/receive replies for AST messages again
6686 - if the client says that it doesn't have the lock, cancel it on the
6688 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6689 mark it as destroyed, it will all work out when decref is called
6690 - fix module use counts (22, 581)
6692 - ASTs now expect a reply (server cancels lock on error reply)
6694 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6697 - fix many simultaneous client startup (392)
6698 - fix dentry->d_it clobbering
6699 - credentials weren't being shipped for readdir/getattr operations
6700 - remove invalid assertions triggered during some concurrent MD
6702 - proper Lustre versions added (336, 389)
6703 - fix memory leak for create error case (398)
6704 - fix LOV locking bug that would get cli/srv out of sync
6705 - fix echo client over LOV (409)
6706 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6707 - fix double-O_EXCL intent crash (424)
6708 - avoid sending multiple lock CANCELs (352)
6710 - MDS can do multi-client recovery (modulo bugs in new code)
6712 - many updates, edits, cleanups
6714 2002-11-18 Phil Schwan <phil@clusterfs.com>
6717 - fix null d_it dereference (346)
6718 - fix full OST/dbench hang (333)
6719 - fix permission problem with file removal (286)
6720 - fix removal of OSCs from LOV when they fail
6721 - fix NULL deref during bulk timeout (214)
6722 - fix problems related to multiple filesystems on one MDS (241)
6723 - fixed serious subtle metadata locking bugs
6724 - free locks on clients when inodes are removed due to memory
6726 - fix inode pointer in lock data (285)
6727 - partial support for multiple MDS on a single host (241)
6728 - data locks weren't cancelled at clear_inode time (290, 311)
6729 - intent locks could lead to unbounded lock growth (205)
6730 - added a maximum lock count, an LRU list, and a flusher
6731 - fix multiple rename (365)
6732 - properly abstracted the echo client
6733 - OSC locked 1 byte too many; fixed
6734 - rewrote brw callback code:
6735 - fixed recovery bugs related to LOVs (306)
6736 - fixed too-many-pages-in-one-write crash (191)
6737 - fixed (again) crash in sync_io_timeout (214)
6738 - probably fixed callback-related race (385)
6740 - Add capability to MDS protocol
6741 - LDLM cancellations and callbacks on different portals
6743 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6746 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6747 - defer open object destruction to close time (601981, 138)
6748 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6749 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6750 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6751 - don't call obd_setattr in ll_file_release for destroyed objects
6752 * protocol change to lustre_msg: move |version| and add |flags|
6753 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6754 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6755 * added replay of create, unlink, link and rename operations during
6756 MDS failover; recovery should be much more robust now
6757 * remove failed OSCs from LOVs (only lov_create uses this so far)
6758 * the lustre-HOWTO was brought (more) up to date (582544)
6760 2002-10-23 Phil Schwan <phil@clusterfs.com>
6763 - in-use dentries weren't being reused properly (617851)
6764 - prevent multiple LDLM setup (599178)
6765 - fix LOV size calculations for truncate (617853)
6766 - fix client handling of MDS intent errors (POSIX)
6767 - fix permission bug in lovstripe.c test (624321)
6768 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6769 - truncate past end of file could corrupt data
6770 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6771 - a race in recovery could return ETIMEDOUT to apps (623947)
6772 - building outside the source directory was fixed
6773 * the lustre-HOWTO was brought (more) up to date (582544)
6774 * major progress was made on recovery functionality
6776 2002-10-10 Phil Schwan <phil@clusterfs.com>
6779 - recovery deadlock fix
6780 - rm -rf causes LBUG fix (617817)
6781 - file open by multiple tasks fix (618962)
6782 - directory permissions bugs (602707 and 620007)
6783 - journal_stop fixed with locking (611313)
6784 - O_APPEND failures resolved (618273, perhaps 614459)
6785 - lconf PATH fix (619770)
6786 - IA64 build fix (621450)
6787 - RPC buffer sizes scale with amount of memory
6789 2002-10-01 Phil Schwan <phil@clusterfs.com>
6792 - locks would be cancelled without throwing away data pages,
6793 resulting in inconsistent data (605627)
6794 - inode attributes were not always being refreshed (605627, 612449)
6795 - lconf now continues to cleanup after lctl reports an error
6796 - MDS now enforces user permissions (602707)
6797 - lprocfs cleanup fixed, but not yet enabled (614157)
6798 - fixed infinite server hang, should a client not respond to an AST
6799 - avoid going into recovery if user calls readlink() with a buffer
6800 that's too small (613941)
6801 - AST RPCs no longer require replies (614867) -- this may be changed
6802 - don't crash server if client sends an IOV that's too big (611336)
6803 - fixed lock conversion deadlock (611892)
6804 - fixed the following of symlinks (614622)
6805 * recovery: the server can remove locks from a client that dies, other
6806 clients can make progress
6807 * more extN patch fixes
6808 * compile-time configurable ptlrpc buffer allocations
6810 - collaborative read cache document
6811 - Lustre Lite Performance CDR document-in-progress
6813 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6816 - fix typo in patch-2.4.18
6818 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6821 - clear ptlrpc request each time in handle_incoming_request()
6822 - unlink of files now destroys the object on the OST
6824 2002-09-19 Peter Braam <braam@clusterfs.com>
6826 * add hard link support
6827 * change obdfile creation method
6828 * kernel patch changed
6830 2002-09-19 Peter Braam <braam@clusterfs.com>
6833 - stack overflow bug in extN fixed
6835 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6837 * documentation updates
6838 - add man pages for config tools
6839 - update tests/README to describe testing with new config tools
6840 - finish metadata API descriptions
6841 * bug fixes and cleanups
6842 - statfs workaround for 16TB limit
6843 - LOV stripe allocation improved, can stripe on subset of OSTs
6844 - LOV file size/IO offset was wrong for files > 4GB in size
6845 - object EA data was being dropped, caused files to be unreadable
6846 - memory overflow with non-LOV OST caused memory corruption
6847 - fixed regression tests to work with new config tools, obdfilter
6848 - fixed bug when directory size became larger than 1 block
6849 - fixed bug (for single client case) when PWD was deleted
6850 - invalidate local directory pages when doing intent-based ops
6851 - avoid LDLM oops when lock callback contained bad data
6853 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6855 * documentation updates
6856 * bug fixes and cleanups
6857 - configuration tools
6860 - 64-bit compile warnings
6861 - 64-bit internal statfs data
6863 * test_brw on persistent OST devices
6865 * lprocfs (disabled)
6867 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6869 * documentation updates
6870 * bug fixes and cleanups
6871 * configuration tools
6873 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6876 * many small fixes to 0_5_4
6877 * io/network handling
6878 * thinkos in MDS operations
6880 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6883 * crucial basic fixes to 0.5.3
6884 * IOR, Iozone work over Elan
6887 2002-08-07 Phil Schwan <phil@clusterfs.com>
6888 * version 0_5_3, our first alpha
6889 * we use the new Portals iovs
6890 * documentation updates
6891 * bug fixes and cleanups
6892 * small changes in the DLM wire protocol
6894 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6895 * version 0_5_1 with some initial stability,
6896 * locking on MD and file I/O.
6897 * documentation updates
6898 * several bug fixes since 0.5.0
6899 * small changes in wire protocol
6901 2002-07-18 Phil Schwan <phil@clusterfs.com>
6903 * delivered as Lustre Light Alpha
6904 * fixed a crash after handling invalid MDS requests
6905 * fixed directory pages for architectures with non-4k pages sizes
6907 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6908 * release version v0_4_4
6909 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6911 * Fixes a number of interruption problems with OST operations.
6912 * Update documentation for portals header changes
6913 * Move all wire protocol structs/defines to lustre_idl.h
6914 * Fixes symlink length bug.
6915 * Add tcpdump to repository.
6917 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6918 * release version v0_4_3
6919 * Fixes statfs for inodes on extN.
6920 * Fixes bug in runtests which would delete /etc/hosts.
6921 * Use 64-bit object IDs wherever possible (not into VFS though)
6922 Remove ost_get_info, which is unused by lustre, and out of date.
6924 2002-07-03 Peter Braam <braam@clusterfs.com>
6925 * release version v0_4_2 Fixes a lookup error (type not passed)
6926 * move forward to head of Portals
6927 * move forward to latest Lustre kernel
6929 2002-06-25 Peter Braam <braam@clusterfs.com>
6930 * release version v0_4_1. Hopefully stable on single node use.