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 3.10.0-1160.2.1.el7 (RHEL7.9)
23 4.18.0-80.11.2.el8 (RHEL8.0)
24 4.4.180-94.97 (SLES12 SP3)
25 4.12.14-95.54 (SLES12 SP4)
26 4.12.14-122.20 (SLES12 SP5)
27 4.15.0-45 (Ubuntu 18.04)
28 * Client known to build on unpatched kernels (others may also work):
29 2.6.32-696.18.7.el6 (RHEL6.9)
30 2.6.32-754.6.3.el6 (RHEL6.10)
31 3.10.0-862.14.4.el7 (RHEL7.5)
32 4.14.0-49.13.1.el7a (RHEL7.5)
33 4.14.0-115.2.2.el7a (RHEL7.6)
34 4.18.0-147.3.1.el8 (RHEL8.1)
35 4.18.0-193.6.3.el8 (RHEL8.2)
36 4.4.120-92.70 (SLES12 SP2)
37 4.4.0-142 (Ubuntu 16.04)
38 vanilla linux 4.15.0 (ZFS only)
39 * Recommended e2fsprogs version: 1.45.6.wc1 or newer
40 * Recommended ZFS / SPL version: 0.7.13
41 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
42 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
43 stack overflow. For more information, please refer to bugzilla 17630.
44 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
45 the re-exporting nodes to be mounted with "32bitapi" mount option
46 * Note for Ubuntu 14.04.5:
47 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
48 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
52 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
53 for currently supported client and server kernel versions.
54 * Server primary kernels built and tested during release cycle:
55 3.10.0-957.27.2.el7 (RHEL7.6)
56 3.10.0-1062.18.1.el7 (RHEL7.7)
57 3.10.0-1127.8.2.el7 (RHEL7.8)
58 * Server known to build on the following kernels (others may also work):
59 2.6.32-696.18.7.el6 (RHEL6.9)
60 3.10.0-862.14.4.el7 (RHEL7.5)
61 4.4.120-92.70 (SLES12 SP2)
62 4.4.180-94.97 (SLES12 SP3)
63 4.4.0-142 (Ubuntu 16.04)
64 4.15.0-45 (Ubuntu 18.04)
65 vanilla linux 4.6.7 (ZFS only)
66 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
67 * Client primary unpatched kernels built and tested during release cycle:
68 3.10.0-957.27.2.el7 (RHEL7.6)
69 3.10.0-1062.18.1.el7 (RHEL7.7)
70 3.10.0-1127.8.2.el7 (RHEL7.8)
71 4.18.0-80.11.2.el8 (RHEL8.0)
72 4.4.180-94.97 (SLES12 SP3)
73 4.12.14-95.45 (SLES12 SP4)
74 4.12.14-122.20 (SLES12 SP5)
75 4.15.0-45 (Ubuntu 18.04)
76 * Client known to build on unpatched kernels (others may also work):
77 2.6.32-696.18.7.el6 (RHEL6.9)
78 2.6.32-754.6.3.el6 (RHEL6.10)
79 3.10.0-862.14.4.el7 (RHEL7.5)
80 4.14.0-49.13.1.el7a (RHEL7.5)
81 4.14.0-115.2.2.el7a (RHEL7.6)
82 4.18.0-147.8.1.el8 (RHEL8.1)
83 4.18.0-193.1.2.el8 (RHEL8.2)
84 4.4.120-92.70 (SLES12 SP2)
85 4.4.0-142 (Ubuntu 16.04)
86 vanilla linux 4.15.0 (ZFS only)
87 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
88 * Recommended ZFS / SPL version: 0.7.13
89 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
90 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
91 stack overflow. For more information, please refer to bugzilla 17630.
92 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
93 the re-exporting nodes to be mounted with "32bitapi" mount option
94 * Note for Ubuntu 14.04.5:
95 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
96 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
100 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
101 for currently supported client and server kernel versions.
102 * Server primary kernels built and tested during release cycle:
103 3.10.0-957.27.2.el7 (RHEL7.6)
104 3.10.0-1062.9.1.el7 (RHEL7.7)
105 * Server known to build on the following kernels (others may also work):
106 2.6.32-696.18.7.el6 (RHEL6.9)
107 3.10.0-862.14.4.el7 (RHEL7.5)
108 4.4.120-92.70 (SLES12 SP2)
109 4.4.180-94.97 (SLES12 SP3)
110 4.4.0-142 (Ubuntu 16.04)
111 4.15.0-45 (Ubuntu 18.04)
112 vanilla linux 4.6.7 (ZFS only)
113 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
114 * Client primary unpatched kernels built and tested during release cycle:
115 3.10.0-957.27.2.el7 (RHEL7.6)
116 3.10.0-1062.9.1.el7 (RHEL7.7)
117 4.18.0-80.11.2.el8 (RHEL8.0)
118 4.4.180-94.97 (SLES12 SP3)
119 4.12.14-95.29 (SLES12 SP4)
120 4.15.0-45 (Ubuntu 18.04)
121 * Client known to build on unpatched kernels (others may also work):
122 2.6.32-696.18.7.el6 (RHEL6.9)
123 2.6.32-754.6.3.el6 (RHEL6.10)
124 3.10.0-862.14.4.el7 (RHEL7.5)
125 4.14.0-49.13.1.el7a (RHEL7.5)
126 4.14.0-115.2.2.el7a (RHEL7.6)
127 4.18.0-147.3.1.el8 (RHEL8.1)
128 4.4.120-92.70 (SLES12 SP2)
129 4.4.0-142 (Ubuntu 16.04)
130 vanilla linux 4.15.0 (ZFS only)
131 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
132 * Recommended ZFS / SPL version: 0.7.13
133 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
134 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
135 stack overflow. For more information, please refer to bugzilla 17630.
136 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
137 the re-exporting nodes to be mounted with "32bitapi" mount option
138 * Note for Ubuntu 14.04.5:
139 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
140 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
144 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
145 for currently supported client and server kernel versions.
146 * Server primary kernels built and tested during release cycle:
147 3.10.0-957.27.2.el7 (RHEL7.6)
148 3.10.0-1062.1.1.el7 (RHEL7.7)
149 * Server known to build on the following kernels (others may also work):
150 2.6.32-696.18.7.el6 (RHEL6.9)
151 3.10.0-862.14.4.el7 (RHEL7.5)
152 4.4.120-92.70 (SLES12 SP2)
153 4.4.180-94.97 (SLES12 SP3)
154 4.4.0-142 (Ubuntu 16.04)
155 4.15.0-45 (Ubuntu 18.04)
156 vanilla linux 4.6.7 (ZFS only)
157 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
158 * Client primary unpatched kernels built and tested during release cycle:
159 3.10.0-957.27.2.el7 (RHEL7.6)
160 3.10.0-1062.1.1.el7 (RHEL7.7)
161 4.18.0-80.7.1.el8 (RHEL8.0)
162 4.4.180-94.97 (SLES12 SP3)
163 4.12.14-95.29 (SLES12 SP4)
164 4.15.0-45 (Ubuntu 18.04)
165 * Client known to build on unpatched kernels (others may also work):
166 2.6.32-696.18.7.el6 (RHEL6.9)
167 2.6.32-754.6.3.el6 (RHEL6.10)
168 3.10.0-862.14.4.el7 (RHEL7.5)
169 4.14.0-49.13.1.el7a (RHEL7.5)
170 4.14.0-115.2.2.el7a (RHEL7.6)
171 4.4.120-92.70 (SLES12 SP2)
172 4.4.0-142 (Ubuntu 16.04)
173 vanilla linux 4.15.0 (ZFS only)
174 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
175 * Recommended ZFS / SPL version: 0.7.13
176 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
177 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
178 stack overflow. For more information, please refer to bugzilla 17630.
179 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
180 the re-exporting nodes to be mounted with "32bitapi" mount option
181 * Note for Ubuntu 14.04.5:
182 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
183 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
187 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
188 for currently supported client and server kernel versions.
189 * Server known to build on the following kernels (others may also work):
190 2.6.32-696.18.7.el6 (RHEL6.9)
191 3.10.0-862.14.4.el7 (RHEL7.5)
192 3.10.0-957.10.1.el7 (RHEL7.6)
193 4.4.120-92.70 (SLES12 SP2)
194 4.4.162-94.69 (SLES12 SP3)
195 4.4.0-142 (Ubuntu 16.04)
196 4.15.0-45 (Ubuntu 18.04)
197 vanilla linux 4.6.7 (ZFS only)
198 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
199 * Client known to build on unpatched kernels (others may also work):
200 2.6.32-696.18.7.el6 (RHEL6.9)
201 2.6.32-754.6.3.el6 (RHEL6.10)
202 3.10.0-862.14.4.el7 (RHEL7.5)
203 3.10.0-957.10.1.el7 (RHEL7.6)
204 4.14.0-49.13.1.el7a (RHEL7.5)
205 4.14.0-115.2.2.el7a (RHEL7.6)
206 4.4.120-92.70 (SLES12 SP2)
207 4.4.162-94.69 (SLES12 SP3)
208 4.12.14-95.13 (SLES12 SP4)
209 4.4.0-142 (Ubuntu 16.04)
210 4.15.0-45 (Ubuntu 18.04)
211 vanilla linux 4.15.0 (ZFS only)
212 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
213 * Recommended ZFS / SPL version: 0.7.13
214 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
215 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
216 stack overflow. For more information, please refer to bugzilla 17630.
217 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
218 the re-exporting nodes to be mounted with "32bitapi" mount option
219 * Note for Ubuntu 14.04.5:
220 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
221 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
226 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
227 for currently supported client and server kernel versions.
228 * Server known to build on the following kernels (others may also work):
229 2.6.32-696.18.7.el6 (RHEL6.9)
230 3.10.0-862.14.4.el7 (RHEL7.5)
231 3.10.0-957.10.1.el7 (RHEL7.6)
232 4.4.120-92.70 (SLES12 SP2)
233 4.4.162-94.69 (SLES12 SP3)
234 4.4.0-142 (Ubuntu 16.04)
235 4.15.0-45 (Ubuntu 18.04)
236 vanilla linux 4.6.7 (ZFS only)
237 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
238 * Client known to build on unpatched kernels (others may also work):
239 2.6.32-696.18.7.el6 (RHEL6.9)
240 2.6.32-754.6.3.el6 (RHEL6.10)
241 3.10.0-862.14.4.el7 (RHEL7.5)
242 3.10.0-957.10.1.el7 (RHEL7.6)
243 4.14.0-49.13.1.el7a (RHEL7.5)
244 4.14.0-115.2.2.el7a (RHEL7.6)
245 4.4.120-92.70 (SLES12 SP2)
246 4.4.162-94.69 (SLES12 SP3)
247 4.12.14-95.13 (SLES12 SP4)
248 4.4.0-142 (Ubuntu 16.04)
249 4.15.0-45 (Ubuntu 18.04)
250 vanilla linux 4.15.0 (ZFS only)
251 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
252 * Recommended ZFS / SPL version: 0.7.13
253 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
254 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
255 stack overflow. For more information, please refer to bugzilla 17630.
256 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
257 the re-exporting nodes to be mounted with "32bitapi" mount option
258 * Note for Ubuntu 14.04.5:
259 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
260 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
264 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
265 for currently supported client and server kernel versions.
266 * Server known to build on the following kernels (others may also work):
267 2.6.32-696.18.7.el6 (RHEL6.9)
268 3.10.0-862.14.4.el7 (RHEL7.5)
269 3.10.0-957.el7 (RHEL7.6)
270 4.4.120-92.70 (SLES12 SP2)
271 4.4.162-94.69 (SLES12 SP3)
272 4.4.0-131 (Ubuntu 16.04)
273 4.15.0-32 (Ubuntu 18.04)
274 vanilla linux 4.6.7 (ZFS only)
275 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
276 * Client known to build on unpatched kernels (others may also work):
277 2.6.32-696.18.7.el6 (RHEL6.9)
278 2.6.32-754.6.3.el6 (RHEL6.10)
279 3.10.0-862.14.4.el7 (RHEL7.5)
280 3.10.0-957.el7 (RHEL7.6)
281 4.14.0-49.13.1.el7a (RHEL7.5)
282 4.14.0-115.2.2.el7a (RHEL7.6)
283 4.4.120-92.70 (SLES12 SP2)
284 4.4.162-94.69 (SLES12 SP3)
285 4.4.0-131 (Ubuntu 16.04)
286 4.15.0-32 (Ubuntu 18.04)
287 vanilla linux 4.15.0 (ZFS only)
288 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
289 * Recommended ZFS / SPL version: 0.7.9
290 * Tested with ZFS / SPL version: 0.7.9
291 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
292 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
293 stack overflow. For more information, please refer to bugzilla 17630.
294 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
295 the re-exporting nodes to be mounted with "32bitapi" mount option
296 * Note for Ubuntu 14.04.5:
297 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
298 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
300 --------------------------------------------------------------------------------
303 2018-04-03 Intel Corporation
305 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
306 for currently supported client and server kernel versions.
307 * Server known to build on patched kernels:
308 2.6.32-431.29.2.el6 (RHEL6.5)
309 2.6.32-504.30.3.el6 (RHEL6.6)
310 2.6.32-573.26.1.el6 (RHEL6.7)
311 2.6.32-642.15.1.el6 (RHEL6.8)
312 2.6.32-696.18.7.el6 (RHEL6.9)
313 3.10.0-693.21.1.el7 (RHEL7.4)
314 3.0.101-0.47.71 (SLES11 SP3)
315 3.0.101-107 (SLES11 SP4)
316 3.12.74-60.64.40 (SLES12 SP1)
317 4.4.114-92.64 (SLES12 SP2)
318 4.4.114-94.11 (SLES12 SP3)
319 3.13.0-101 (Ubuntu 14.04, ZFS only)
320 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
321 4.4.0-116 (Ubuntu 16.04)
322 vanilla linux 4.6.7 (ZFS only)
323 * Client known to build on unpatched kernels:
324 2.6.32-431.29.2.el6 (RHEL6.5)
325 2.6.32-504.30.3.el6 (RHEL6.6)
326 2.6.32-573.26.1.el6 (RHEL6.7)
327 2.6.32-642.15.1.el6 (RHEL6.8)
328 2.6.32-696.18.7.el6 (RHEL6.9)
329 3.10.0-693.21.1.el7 (RHEL7.4)
330 3.0.101-0.47.71 (SLES11 SP3)
331 3.0.101-107 (SLES11 SP4)
332 3.12.74-60.64.40 (SLES12 SP1)
333 4.4.114-92.64 (SLES12 SP2)
334 4.4.114-94.11 (SLES12 SP3)
335 3.13.0-101 (Ubuntu 14.04)
336 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
337 4.4.0-116 (Ubuntu 16.04)
339 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
340 * Recommended ZFS / SPL version: 0.7.6
341 * Tested with ZFS / SPL version: 0.7.6
342 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
343 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
344 stack overflow. For more information, please refer to bugzilla 17630.
345 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
346 the re-exporting nodes to be mounted with "32bitapi" mount option
347 * Note for Ubuntu 14.04.5:
348 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
349 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
351 --------------------------------------------------------------------------------
353 2017-07-05 Intel Corporation
355 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
356 for currently supported client and server kernel versions.
357 * Server known to build on patched kernels:
358 2.6.32-431.29.2.el6 (RHEL6.5)
359 2.6.32-504.30.3.el6 (RHEL6.6)
360 2.6.32-573.26.1.el6 (RHEL6.7)
361 2.6.32-642.15.1.el6 (RHEL6.8)
362 2.6.32-696.3.1.el6 (RHEL6.9)
363 3.10.0-514.21.1.el7 (RHEL7.3)
364 3.0.101-0.47.71 (SLES11 SP3)
365 3.0.101-100 (SLES11 SP4)
366 3.12.74-60.64.40 (SLES12 SP1)
367 4.4.59-92.17 (SLES12 SP2)
368 vanilla linux 4.6.7 (ZFS only)
369 * Client known to build on unpatched kernels:
370 2.6.32-431.29.2.el6 (RHEL6.5)
371 2.6.32-504.30.3.el6 (RHEL6.6)
372 2.6.32-573.26.1.el6 (RHEL6.7)
373 2.6.32-642.15.1.el6 (RHEL6.8)
374 2.6.32-696.3.1.el6 (RHEL6.9)
375 3.10.0-514.21.1.el7 (RHEL7.3)
376 3.0.101-0.47.71 (SLES11 SP3)
377 3.0.101-100 (SLES11 SP4)
378 3.12.74-60.64.40 (SLES12 SP1)
379 4.4.59-92.17 (SLES12 SP2)
381 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
382 * Recommended ZFS / SPL version: 0.6.5.9
383 * Tested with ZFS / SPL version: 0.6.5.9
384 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
385 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
386 stack overflow. For more information, please refer to bugzilla 17630.
387 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
388 the re-exporting nodes to be mounted with "32bitapi" mount option
390 --------------------------------------------------------------------------------
392 12-07-2016 Intel Corporation
394 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
395 for currently supported client and server kernel versions.
396 * Server known to build on patched kernels:
397 2.6.32-431.29.2.el6 (RHEL6.5)
398 2.6.32-504.30.3.el6 (RHEL6.6)
399 2.6.32-573.26.1.el6 (RHEL6.7)
400 2.6.32-642.11.1.el6 (RHEL6.8)
401 3.10.0-514.el7 (RHEL7.3)
402 3.0.101-0.47.71 (SLES11 SP3)
403 3.0.101-80 (SLES11 SP4)
404 3.12.62-60.64.8 (SLES12 SP1)
405 vanilla linux 4.5.7 (ZFS only)
406 * Client known to build on unpatched kernels:
407 2.6.32-431.29.2.el6 (RHEL6.5)
408 2.6.32-504.30.3.el6 (RHEL6.6)
409 2.6.32-573.26.1.el6 (RHEL6.7)
410 2.6.32-642.11.1.el6 (RHEL6.8)
411 3.10.0-514.el7 (RHEL7.3)
412 3.0.101-0.47.71 (SLES11 SP3)
413 3.0.101-80 (SLES11 SP4)
414 3.12.62-60.64.8 (SLES12 SP1)
416 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
417 * Recommended ZFS / SPL version: 0.6.5.8
418 * Tested with ZFS / SPL version: 0.6.5.8
419 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
420 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
421 stack overflow. For more information, please refer to bugzilla 17630.
422 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
423 the re-exporting nodes to be mounted with "32bitapi" mount option
425 --------------------------------------------------------------------------------
427 02-29-2016 Intel Corporation
429 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
430 for currently supported client and server kernel versions.
431 * Server known to build on patched kernels:
432 2.6.32-431.29.2.el6 (RHEL6.5)
433 2.6.32-504.30.3.el6 (RHEL6.6)
434 2.6.32-573.12.1.el6 (RHEL6.7)
435 3.10.0-327.3.1.el7 (RHEL7.2)
436 3.0.101-0.47.71 (SLES11 SP3)
437 3.0.101-68 (SLES11 SP4)
438 vanilla linux 4.2.1 (ZFS only)
439 * Client known to build on unpatched kernels:
440 2.6.32-431.29.2.el6 (RHEL6.5)
441 2.6.32-504.30.3.el6 (RHEL6.6)
442 2.6.32-573.12.1.el6 (RHEL6.7)
443 3.10.0-327.3.1.el7 (RHEL7.2)
444 3.0.101-0.47.71 (SLES11 SP3)
445 3.0.101-68 (SLES11 SP4)
448 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
449 * Recommended ZFS / SPL version: 0.6.4.2
450 * Tested with ZFS / SPL version: 0.6.4.2
451 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
452 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
453 stack overflow. For more information, please refer to bugzilla 17630.
454 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
455 the re-exporting nodes to be mounted with "32bitapi" mount option
457 --------------------------------------------------------------------------------
459 03-10-2015 Intel Corporation
461 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
462 for currently supported client and server kernel versions.
463 * Server known to build on patched kernels:
464 2.6.32-431.29.2.el6 (RHEL6.5)
465 2.6.32-504.8.1.el6 (RHEL6.6)
466 3.0.101-0.46 (SLES11 SP3)
467 * Client known to build on unpatched kernels:
468 2.6.32-431.29.2.el6 (RHEL6.5)
469 2.6.32-504.8.1.el6 (RHEL6.6)
470 3.10.0-123.20.1.el7 (RHEL7)
471 3.0.101-0.46 (SLES11 SP3)
472 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
473 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
474 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
475 stack overflow. For more information, please refer to bugzilla 17630.
476 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
477 the re-exporting nodes to be mounted with "32bitapi" mount option
479 Severity : enhancement
481 Description: control OST-index in IDIF via ROCOMPAT flag.
482 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
483 last_rcvd file. For new formatted OST device, it will be auto set;
484 for the case of upgrading from old OST device, you can enable it
485 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
486 enabled, for new created OST-object, its IDIF-in-LMA will contain
487 the OST-index; for the existing OST-object, the OSD will convert
488 old format IDIF as new format IDIF with OST-index stored in the
489 LMA EA when accessing such OST-object or via OI scrub. Once such
490 flag is enabled, it cannot be reverted back, so the system cannot
491 be downgraded to the orignal incompatible version.
493 --------------------------------------------------------------------------------
495 07-30-2014 Intel Corporation
497 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
498 for currently supported client and server kernel versions.
499 * Server known to build on patched kernels:
500 2.6.32-431.20.3.el6 (RHEL6.5)
501 3.0.101-0.31 (SLES11 SP3)
502 * Client known to build on unpatched kernels:
503 2.6.32-431.20.3.el6 (RHEL6.5)
504 3.0.101-0.31 (SLES11 SP3)
507 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
508 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
509 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
510 stack overflow. For more information, please refer to bugzilla 17630.
511 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
512 the re-exporting nodes to be mounted with "32bitapi" mount option
514 --------------------------------------------------------------------------------
516 10-31-2013 Intel Corporation
518 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
519 * Server builds for kernels:
520 2.6.32-358.18.1.el6 (RHEL6.4)
521 * Client builds for unpatched kernels:
522 2.6.32-358.18.1.el6 (RHEL6.4)
523 3.0.92-0.8 (SLES11 SP3)
525 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
526 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
527 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
528 stack overflow. For more information, please refer to bugzilla 17630.
529 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
530 the re-exporting nodes to be mounted with "32bitapi" mount option
532 --------------------------------------------------------------------------------
534 05-30-2013 Intel Corporation
536 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
537 * Server builds for kernels:
538 2.6.32-358.6.2.el6 (RHEL6.4)
539 * Client builds for unpatched kernels:
540 2.6.32-358.6.2.el6 (RHEL6.4)
541 3.0.74-0.6.6 (SLES11 SP2)
543 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
544 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
545 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
546 stack overflow. For more information, please refer to bugzilla 17630.
547 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
548 the re-exporting nodes to be mounted with "32bitapi" mount option
550 --------------------------------------------------------------------------------
552 9-30-2012 Whamcloud, Inc.
554 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
555 * Server support for kernels:
556 2.6.32-279.5.1.el6 (RHEL6)
557 * Client support for unpatched kernels:
558 2.6.32-279.5.1.el6 (RHEL6)
560 2.6.18-238.19.1.el5 (RHEL5)
561 2.6.16.60-0.69.1 (SLES10)
562 * Recommended e2fsprogs version: 1.41.90.wc4
563 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
564 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
565 stack overflow. For more information, please refer to bugzilla 17630.
566 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
567 the re-exporting nodes to be mounted with "32bitapi" mount option
569 Severity : enhancement
571 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
573 Severity : enhancement
575 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
577 Severity : enhancement
579 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
581 Severity : enhancement
583 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
585 Severity : enhancement
587 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
589 --------------------------------------------------------------------------------
591 04-30-2012 Whamcloud, Inc.
593 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
594 * Server support for kernels:
595 2.6.32-220.4.2.el6 (RHEL6)
596 * Client support for unpatched kernels:
597 2.6.32-220.4.2.el6 (RHEL6)
599 2.6.18-238.19.1.el5 (RHEL5)
600 2.6.16.60-0.69.1 (SLES10)
601 * Recommended e2fsprogs version: 1.41.90.wc4
602 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
603 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
604 stack overflow. For more information, please refer to bugzilla 17630.
605 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
606 the re-exporting nodes to be mounted with "32bitapi" mount option
608 Severity : enhancement
610 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
612 Severity : enhancement
614 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
616 Severity : enhancement
618 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
620 Severity : enhancement
622 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
624 --------------------------------------------------------------------------------
626 09-30-2011 Whamcloud, Inc.
628 * Server support for kernels:
629 2.6.18-238.19.1.el5 (RHEL5)
630 2.6.32-131.6.1.el6 (RHEL6)
631 * Client support for unpatched kernels:
632 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
633 2.6.18-238.19.1.el5 (RHEL5)
634 2.6.32-131.6.1.el6 (RHEL6)
635 2.6.29.4-167.fc11 (FC11)
636 2.6.16.60-0.69.1 (SLES10)
638 * Recommended e2fsprogs version: 1.41.90.wc3
639 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
640 Lustre filesystem with 4K stack may cause a stack overflow. For more
641 information, please refer to bugzilla 17630.
642 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
643 mounted with "32bitapi" mount option
645 Severity : enhancement
646 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
648 Severity : enhancement
649 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
651 Severity : enhancement
652 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
655 Frequency : only with SLES10
657 Description: Use OFED "KMP" provided by Novell
658 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
659 Lustre is now built against this package. That means you need to
660 install the ofed-kmp package from Novell for the patchless client.
661 Note that the ofed-kmp that Novell ships may not exactly match the
662 kernel version but should still be compatible.
664 Severity : enhancement
666 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
667 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
669 --------------------------------------------------------------------------------
671 2010-07-15 Oracle, Inc.
673 * Server support for kernels:
674 2.6.18-164.11.1.el5 (RHEL 5)
675 2.6.18-164.11.1.0.1.el5 (OEL 5)
676 * Client support for unpatched kernels:
677 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
678 2.6.18-164.11.1.el5 (RHEL 5),
679 2.6.18-164.11.1.0.1.el5 (OEL 5)
680 2.6.16.60-0.42.8 (SLES 10),
682 2.6.29.4-167.fc11 (FC11)
683 * Recommended e2fsprogs version: 1.41.10-sun2
685 Severity : enhancement
687 Description: Distribute lfs_migrate script with lustre
688 Details : The lfs_migrate mentioned in the manual is now included in the
689 lustre tree in lustre/scripts and will distributed in lustre RPMs.
691 Severity : enhancement
693 Description: Failover nid list added to import proc file.
695 Severity : enhancement
697 Description: "som_preview" mount option required to use SOM
698 Details : The clients will need to use som_preview mount option
702 Description: Lustre does not do 1MB IOs to HW RAID
703 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
705 Severity : enhancement
707 Description: lfs quota output is not very convenient for awk/sed-parsing
708 Details : Some positions in lfs quota output table could be empty or
709 non-empty which made it hard to parse it with scripts, now
710 a dash is put instead of space where there is not supposed
715 Description: Update to OFED 1.5.1
716 Details : OFED 1.5.1 will be used in environments that don't use
717 vendor supplied OFED.
721 Description: Remove lnet. prefix from lctl params display
722 Details : Also renamed lnet.memused to lnet_memused to
727 Description: Use conf_param -d to remove permanant settings.
731 Description: Add list_param -R
732 Details : lctl list_param now supports recursive mode.
735 Bugzilla : 21829, 22312
736 Description: Fix llobdstat.
737 Details : 1. Now updating the write_bytes counter
738 2. Port script fixes from 1.8.
742 Description: decrease the usage of memory on clients.
743 Details : 1. On clients, recycle dentries and inodes unused.
744 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
749 Description: llobdstat fix and enhancement
750 Details : add a counter to set a limit to how many samples will be returned
751 fix a wildcard in the path to limit to obdfilter stats only
755 Description: lfs check only allowed for root
756 Details : Code cleanups and added a regresion test of non-root use of
761 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
762 Details : Fix hr->hr_index overload.
766 Description: Fix FIEMAP functionality.
770 Description: FC11 patchless client support added
774 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
775 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
779 Description: New acc-sm script and various test-framework enhancments.
783 Description: Previously landed patch was reverted.
787 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
791 Description: Add ldiskfs maxdirsize mount option.
795 Description: Fix mkfs.lustre support for 16TB LUNs.
799 Description: Print mmp_check_interval
800 Details : print mmp_check_interval and make it possible to abort
801 mount operation in case it takes too long.
805 Description: Add cascading_rw to lustre mpi tests.
809 Description: Store SOM data in EA.
810 Details : Also several SOM related cleanups.
814 Description: Add refcount to md_open_data
818 Description: Use refcount on lov_request_set to prevent memory corruption.
822 Description: Statahead: use parent dir's i_mutex for synchronization.
826 Description: loadgen improvements.
830 Description: Enable peer health detection in ksocklnd and ko2iblnd.
834 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
838 Description: Fix race on obd_devcie:md_stats
842 Description: Fix SLES 11 client builds.
843 Details : Resolved uninitalized pointer gcc warnings.
847 Description: Allow non-root access for "lfs check".
848 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
852 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
854 Severity : enhancement
856 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
857 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
861 Description: Sanity test_27p fails.
862 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
863 Sanity tests 27 were cleaned up and improved.
867 Description: Unify req format on client/servers
868 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
869 buffers and swabbers approach..
874 Description: drop join file stuff.
876 Severity : enhancement
878 Description: Add 2.6.27 and SLES11 (patchless client) support
882 Description: Fix in ptlrpc_expire_one_request() to print the signed time
885 Severity : enhancement
887 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
890 Severity : enhancement
892 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
894 Severity : enhancement
896 Description: correctly handle big reply message.
897 Details : send LNet event if reply is bigger than buffer and adjust this buffer
902 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
904 Severity : enhancement
906 Description: Add support for OEL5.
910 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
912 Severity : enhancement
914 Description: "weak-modules" support
915 Details : Implement "weak-modules" support which enables kernel modules
916 to be used with any kernel that implements the same kABI. In
917 order to achieve this modules are now installed in
918 /lib/modules/$(uname -r)/updates/kernel on all distributions.
922 Description: client could not reconnect to OST because of an active request.
923 Details : The client now retries to reconnect to the same server, if a
924 connect request failed with EBUSY or -EAGAIN. Test case 35b
925 conf-sanity.sh has been created.
929 Description: Don't allow make backward step on assignin osc next id.
930 Details : race between allocation next id and ll_sync thread can be cause
931 of set wrong osc next id and can be kill valid ost objects.
935 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
939 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
943 Description: Error handling in osc_statfs_interpret() has been improved.
944 Details : Check in osc_statfs_interpret() for EBADR.
948 Description: Do not send statfs() requests to OSTs disabled by administrator.
949 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
951 Severity : enhancement
953 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
955 Severity : enhancement
957 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
959 Severity : enhancement
961 Description: Enhancement for lfs(1) command to use numeric uid/gid.
965 Description: Improved cleanup procedure for conf-sanity test_46a.
969 Description: Do not start bulk transfer, if client tried to reconnect.
970 Details : Do not start bulk transfer, if client tried to reconnect,
971 but failed due to active rpcs. Check for exp_abort_active_req
972 has been implemented.
976 Description: Prevent inconsistences between linux and lustre mount structures.
977 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
978 Make the sleep interruptible.
980 Severity : enhancement
982 Description: provide server to client comms path
983 Details : server to client communications path via new LDLM_SET_INFO rpc
987 Description: truncate starts GFP_FS allocation under transaction causing deadlock
988 Details : ldiskfs_truncate calls grab_cache_page which may start page
989 allocation under an open transaction. This may lead to
990 calling prune_icache with consequent lustre reentrance.
994 Description: send correctly lsm on open replay
995 Details : MDS is trust to LSM size on replay open, but client can set wrong size
1000 Description: enable client interface failover
1001 Details : When a child reconnects from another NID, properly update export
1002 nid hash position and ldlm reverse import.
1004 Severity : enhancement
1006 Description: Communicate OST degraded/readonly state via statfs to MDS
1007 Details : Flags in the statfs returned from OSTs indicate whether the
1008 OST is in a degraded RAID state, or if the filesystem has
1009 turned read-only after a filesystem error is detected.
1014 Description: don't panic if EPROTO was hit when reading symlink
1015 Details : correctly handling request reference in error cases.
1017 Severity : enhancement
1019 Description: Add LustreNetLink, a kernel-userspace communcation path.
1021 Severity : enhancement
1023 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1028 Description: access to llog context before init.
1029 Details : move handling CATALOGS file at osc layer and forbid access to llog
1030 context before init.
1033 Frequency : rare, only when client is writing to OST on local node
1035 Description: Avoid deadlock for local client writes
1036 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1037 memory freeing context. This allows OST threads to set the
1038 PF_MEMALLOC flag on task structures in order to allocate memory
1039 from reserved pools and complete IO.
1040 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1041 so that the OST threads generate memory pressure and allow
1042 inactive pages to be reclaimed.
1044 Severity : enhancement
1046 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1051 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1052 Details : move ldlm namespace creation in setup phase to avoid grab
1053 _lprocfs_lock with cli_sem held.
1057 Description: Temporarily disable grant shrink.
1058 Details : Disable the feature for debugging.
1061 Frequency : only when formatting test filesystems
1063 Description: Unable to run several mkfs.lustre on loop devices at the same time
1064 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1065 formatting. The solution is to proper handle the error.
1067 Severity : enhancement
1069 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1071 Severity : enhancement
1073 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1078 Description: resolve race between obd_disconnect and class_disconnect_exports
1079 Details : if obd_disconnect will be called to already disconnected export he
1080 forget release one reference and osc module can't unloaded.
1082 Severity : enhancement
1084 Description: Allow tuning service thread via /proc
1085 Details : For each service a new
1086 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1087 created that can be used to set min/max thread counts, and get the
1088 current number of running threads.
1093 Descriptoin: don't return error if have partially created objects for file.
1094 Details : lov_update_create_set uses set->set_success as index for created
1095 objects, so if some requests will be failed, they will have hole
1096 at end of array and qos_shrink_lsm for allocate correct lsm.
1098 Severity : enhancement
1100 Description: Update OFED support to 1.4.1
1105 Description: Reduce small size read RPC
1106 Details : Set read-ahead limit for every file and only do read-ahead when
1107 available read-ahead pages are bigger than 1M to avoid small size
1110 Severity : enhancement
1112 Description: move AT tunable parameters for more consistent usage
1113 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1115 Severity : enhancement
1117 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1118 Details : allow skip disconnected ost for send statfs request and hide error
1124 Description: fix racy locking of mballoc block bitmaps causing BUG
1125 Details : The locking of the mballoc buddy bitmap and the in-memory
1126 block bitmap was using two different spin locks in some
1127 cases. This made it possible to incorrectly access the
1128 mballoc bitmap while another process was modifying it,
1129 causing a sanity assertion to fail. While no on-disk corruption
1130 was reported, there was some risk of this happening.
1133 Frequency : rare, on llog test 6
1135 Descriptoin: don't allow connect to already connected import
1136 Details : allowing connect to already connected import is hide connecting problem.
1141 Descriptoin: allow kill process which wait statahead result
1142 Details : for some reasons 'ls' can stick in waiting result from statahead,
1143 in this case need way for kill this process.
1145 Severity : enhancement
1147 Description: Add state history info file, enhance import info file
1148 Details : Track import connection state changes in a new osc/mdc proc file;
1149 add overview-type data to the osc/mdc import proc file.
1151 Severity : enhancement
1153 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1155 Severity : enhancement
1157 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1162 Descriptoin: OST grant too much space to client even there are not enough space.
1163 Details : Client will shrink its grant cache to OST if there are no write
1164 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1165 this grant cache if there are already not enough avaible space
1166 (left_space < total_clients * 32M).
1169 Frequency : start MDS on uncleanly shutdowned MDS device
1171 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1172 Details : stay in waiting mds<>ost recovery finished produce random bugs
1173 due race between two ll_sync thread for one lov target. send
1174 ACTIVATE event only if connect realy finished and import have
1178 Frequency : rare, connect and disconnect target at same time
1180 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1181 Details : don't call obd_disconnect under lov_lock. this long time
1182 operation and can block ptlrpcd which answer to connect request.
1187 Descriptoin: don't lose wakeup for imp_recovery_waitq
1188 Details : recover_import_no_retry or invalidate_import and import_close can
1189 both sleep on imp_recovery_waitq, but we was send only one wakeup
1193 Frequency : always with long access acl
1195 Descriptoin: mds can't pack reply with long acl.
1196 Details : mds don't control size of acl but they limited by reint/getattr
1199 Severity : enhancement
1201 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1203 Severity : enhancement
1205 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1208 Frequency : start MDS on uncleanly shutdowned MDS device
1210 Descriptoin: aborting recovery hang on MDS
1211 Details : don't throttle destroy RPCs for the MDT.
1214 Frequency : on remount
1216 Description: external journal device not working after the remount
1217 Details : clear dev_rdonly flag for external journal devices in
1223 Description: shutdown vs evict race
1224 Details : client_disconnect_export vs connect request race.
1225 if client will evicted at this time - we start invalidate
1226 thread without referece to import and import can be freed
1230 Frequency : rare, need acl's on inode.
1232 Description: client can't handle ost additional correctly
1233 Details : if ost was added after client connected to mds client can have
1234 hit lnet_try_match_md ... to big messages to wide striped files.
1235 in this case need teach client to handle config events about add
1236 lov target and update client max ea size at that event.
1238 Severity : enhancement
1240 Description: Changelogs
1241 Details : Changelogs are a lightweight mechanism to track filesystem
1242 metadata and namespace changes. The changelog is recorded
1243 permanently on the MDTs, and is periodically "consumed" / purged
1244 when records are no longer needed.
1246 Severity : enhancement
1248 Description: compact fld format with extents
1249 Details : Store range of seq rather than every seq in FLD. Seq
1250 controller update FLD rather than clients. In Case of CMD, mdt0
1251 has FLD, all other metadata server act as non persistent proxy
1252 for FLD queries and cache fld entries in fld cache.
1257 Description: don't skip ost target if they assigned to file
1258 Details : Drop slow OSCs if we can, but not for requested start idx.
1259 This means "if OSC is slow and it is not the requested
1260 start OST, then it can be skipped, otherwise skip it only
1261 if it is inactive/recovering/out-of-space.
1265 Description: more cleanup in mds_lov
1266 Details : not send LOV EA under replay, we can't know about they size at this
1267 time. Don't allow client connect to mds before any ost connected,
1268 for avoid problems with LOV EA size and returning EIO to client.
1270 Severity : enhancement
1272 Description: Interoperability at server side (Disk interoperability)
1274 Severity : enhancement
1276 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1278 Severity : enhancement
1280 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1282 Severity : enhancement
1284 Description: New client IO stack (CLIO).
1286 Severity : enhancement
1288 Description: Commit on sharing. Eliminate inter-client dependencies between
1289 uncommitted transactions by doing transaction commits.
1290 Thereby clients may recovery independently.
1293 Frequency : Create a symlink file with a very long name
1295 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1296 Details : If there is no extra space in the request for early cancels,
1297 ldlm_req_handles_avail() returns 0 instead of a negative value.
1299 Severity : enhancement
1301 Description: Add /proc entry for import status
1302 Details : The mdc, osc, and mgc import directories now have
1303 an import directory that contains useful import data for debugging
1304 connection problems.
1306 Severity : enhancement
1308 Description: Re-disable certain /proc logging
1309 Details : Enable and disable client's offset_stats, extents_stats and
1310 extents_stats_per_process stats logging on the fly.
1313 Frequency : Only on FC kernels 2.6.22+
1315 Description: oops in statahead
1316 Details : Do not drop reference count for the dentry from VFS when lookup,
1317 VFS will do that by itself.
1319 Severity : enhancement
1321 Description: Generic /proc file permissions
1322 Details : Set /Proc file permissions in a more generic way to enable non-
1323 root users operate on some /proc files.
1327 Description: Hitting mdc_commit_close() ASSERTION
1328 Details : Properly handle request reference release in
1329 ll_release_openhandle().
1333 Description: quota recovery deadlock during mds failover
1334 Details : This patch includes att18982, att18236, att18237 in bz14840.
1336 1. fix osts hang when mds does failover with quotaon
1337 2. prevent watchdog storm when osts threads wait for the
1342 Frequency : only patchless client
1343 Description: add workaround for race between add/remove dentry from hash
1345 Severity : enhancement
1347 Description: Allow OST glimpses to return PW locks
1351 Description: LBUG when llog conf file is full
1352 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1356 Description: Prevent import from entering FULL state when server in recovery
1360 Description: service mount cannot take device name with ":"
1361 Details : Only when device name contains ":/" will mount treat it as
1367 Description: replace ptlrpcd with the statahead thread to interpret the async
1368 statahead RPC callback
1372 Frequency : on recovery
1373 Description: I/O failures after umount during fail back
1374 Details : if client reconnected to restarted server we need join to recovery
1375 instead of find server handler is changed and process self
1376 eviction with cancel all locks.
1378 Severity : enhancement
1380 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1382 Severity : enhancement
1384 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1386 Severity : enhancement
1388 Description: Upcall on Lustre log has been dumped
1389 Details : Allow for a user mode script to be called once a Lustre log has
1390 been dumped. It passes the filename of the dumped log to the
1391 script, the location of the script can be specified via
1392 /proc/sys/lnet/debug_log_upcall.
1397 Description: avoid idr_remove called for id which is not allocated.
1398 Details : Move assigment s_dev for clustered nfs to end of initialization,
1399 to avoid problem with error handling.
1404 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1405 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1406 into CONN_USED_HASH and this prodice warning when put connection
1407 again in unused hash.
1410 Severity : enhancement
1412 Description: File striping can now be set to use an arbitrary pool of OSTs.
1414 Severity : enhancement
1416 Description: Export bytes_read/bytes_write count on OSC/OST.
1420 Description: Early reply size mismatch, MGC loses connection
1421 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1422 the connect flags are properly negotiated.
1427 Description: quotas are not honored with O_DIRECT
1428 Details : all writes with the flag O_DIRECT will use grants which leads to
1429 this problem. Now using OBD_BRW_SYNC to guard this.
1433 Description: add quota statistics
1434 Details : 1. sort out quota proc entries and proc code.
1435 2. add quota statistics
1437 Severity : enhancement
1439 Description: enable quota support for HEAD.
1443 Description: Properly propagate oinfo flags from lov to osc for statfs
1444 Details : restore missing copy oi_flags to lov requests.
1446 Severity : enhancement
1448 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1449 : l_getgroups(8), lst(8), routerstat(8)
1450 Details : included man pages for llobdstat(8), llstat(8),
1451 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1453 Severity : enhancement
1455 Description: configure's --enable-quota should check the
1456 : kernel .config for CONFIG_QUOTA
1457 Details : configure is terminated if --enable-quota is passed but
1458 : no quota support is in kernel
1462 Description: Remove portals compatibility
1463 Details : Remove portals compatibility, not interoperable with releases
1468 Description: Resolve device initialization race
1469 Details : Prevent proc handler from accessing devices added to the
1470 obd_devs array but yet be intialized.
1472 Severity : enhancement
1474 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1476 Severity : enhancement
1478 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1483 Description: Using wrong pointer in osc_brw_prep_request
1484 Details : Access to array[-1] can produce panic if kernel compiled with
1485 CONFIG_PAGE_ALLOC enabled
1489 Description: Client runs out of low memory
1490 Details : Consider only lowmem when counting initial number of llap pages
1494 Description: *optional* service tags registration
1495 Details : if the "service tags" package is installed on a Lustre node
1496 When the filesystem is mounted, a local-node service tag will
1497 be created. See http://inventory.sun.com/ for more information
1498 about the Service Tags asset management system.
1502 Description: Kernel BUG tries to release flock
1503 Details : Lustre does not destroy flock lock before last reference goes
1504 away. So always drop flock locks when client is evicted and
1505 perform unlock regardless of successfulness of speaking to MDS.
1509 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1513 Description: sanity test 65a fails if stripecount of -1 is set
1514 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1519 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1520 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1523 Severity : enhancement
1525 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1529 Description: do not process already freed flock
1530 Details : flock can possibly be freed by another thread before it reaches
1531 to ldlm_flock_completion_ast.
1535 Description: LBUG during stress test
1536 Details : Need properly lock accesses the flock deadlock detection list.
1540 Description: oops in page fault handler
1541 Details : kernel page fault handler can return two special 'pages' in error
1542 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1546 Description: timeout with invalidate import.
1547 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1548 should be handled by ptlrpcd. This produce long age waiting and
1549 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1551 Severity : enhancement
1553 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1557 Description: enable MGS and MDT services start separately
1558 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1559 a co-located MGS without starting the MGS, which is a complement
1560 to 'nosvc' mount option.
1564 Frequency : after recovery
1565 Description: precreate to many object's after del orphan.
1566 Details : del orphan st in oscc last_id == next_id and this triger growing
1567 count of precreated objects. Set flag LOW to skip increase count
1568 of precreated objects.
1572 Frequency : rare, on clear nid stats
1573 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1574 Details : when clean nid stats sometimes try destroy live entry,
1575 and this produce panic in free.
1579 Description: Stack overflow during MDS log replay
1580 ease stack pressure by using a thread dealing llog_process.
1584 Description: wait until IO finished before start new when do lock cancel.
1585 Details : VM protocol want old IO finished before start new, in this case
1586 need wait until PG_writeback is cleared until check dirty flag and
1587 call writepages in lock cancel callback.
1589 Severity : enhancement
1591 Description: using special macro for print time and cleanup in includes.
1595 Description: mds_mfd_close() ASSERTION(rc == 0)
1596 Details : In mds_mfd_close(), we need protect inode's writecount change
1597 within its orphan write semaphore to prevent possible races.
1601 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1605 Frequency : rare, on shutdown ost
1606 Description: don't hit live lock with umount ost.
1607 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1608 use shrink_dcache_sb instead.
1612 Description: don't panic with use echo client
1613 Details : echo client pass NULL as client nid pointer and this produce null
1614 pointer dereference.
1618 Description: fix build on ppc32
1619 Details : compile code with -m64 flag produce wrong object file for ppc32.
1623 Description: add message levels for liblustreapi
1627 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1628 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1629 will result in an -EAGAIN and in the exit call path we call
1630 alloc_rr() which will with increasing aggressiveness attempt to
1631 aquire precreated objects on the minimum number of required OSCs.
1635 Description: Use old size assignment to avoid deadlock
1636 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1637 the scheduling while holding a spinlock. We do not need locking
1638 for size in ll_update_inode() because size is only updated from
1639 the MDS for directories or files without objects, so there is no
1640 other place to do the update, and concurrent access to such inodes
1641 are protected by the inode lock.
1645 Description: resolve "_IOWR redefined" build error on SLES10
1649 Description: dump the memory debugging after all modules are unloaded to
1650 suppress false negative in conf_sanity test 39
1652 Severity : enhancement
1654 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1659 Description: fix SLES kernel versioning
1660 Details : the kernel version for our SLES 10 kernel did not include a "-"
1661 before the "smp" at the end. while this was not a problem in
1662 general, it did mean that software trying to use the kernel
1663 version to try to detect a vendor specific kernel would fail.
1664 this was most evident by the OFED build scripts.
1668 Description: Don't update lov_desc members until making sure they are valid
1669 Details : When updating lov_desc members via proc fs, need fix their
1670 validities before doing the real update.
1674 Description: don't put request into delay list while invalidate in flight.
1675 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1676 import in flight. this produce timeout for invalidate and sometimes
1677 can cause stale data.
1681 Frequency : on ppc only
1682 Description: not convert ost objects for directory because it's not exist.
1683 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1687 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1689 Severity : enhancement
1691 Description: Update to support 2.6.22.14 vanilla kernel.
1695 Frequency : rare, on recovery
1696 Description: read procfs can produce deadlock in some situation
1697 Details : Holding lprocfs lock which send rpc can produce block for destroy
1698 obd objects and this also block reconnect with -EALREADY. This
1699 isn't fix all lprocfs bugs - but make it rare.
1701 Severity : enhancement
1703 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1706 Frequency : frequent on X2 node
1708 Description: mdc_set_open_replay_data LBUG
1709 Details : Set replay data for requests that are eligible for replay.
1713 Description: lustre_mgs: operation 101 on unconnected MGS
1714 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1715 MGC, and late on MGC cannot successfully connect to MGS and a lot
1716 of the error messages complaining that MGS is not connected.
1720 Frequency : on network error
1721 Description: panic with double free request if network error
1722 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1723 but it's true only for synchronus enqueue, for async enqueue
1724 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1727 Severity : enhancement
1729 Description: client-side metadata stat-ahead during readdir(directory readahead)
1730 Details : perform client-side metadata stat-ahead when the client detects
1731 readdir and sequential stat of dir entries therein
1734 Frequency : on start mds
1736 Description: Implement get_info(last_id) in obdfilter.
1739 Frequency : occasional
1741 Description: Correctly check stale fid, not start epoch if ost not support SOM
1742 Details : open with flag O_CREATE need set old fid in op_fid3 because
1743 op_fid2 was overwritten with new generated fid, but mds can answer
1744 with one of these two fids and both is not stale. Setattr
1745 incorrectly started an epoch and assume will be called
1746 done_writing, but without SOM done_writing ever being called.
1749 Frequency : rare, depends on device drivers and load
1751 Description: MDS or OSS nodes crash due to stack overflow
1752 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1753 In some cases, in conjunction with device drivers that use a lot
1754 of stack the MDS (or possibly OSS) service threads could overflow
1755 the stack. One change which was identified to consume additional
1756 stack has been reworked to avoid the extra stack usage.
1759 Frequency : occasional
1761 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1762 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1763 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1764 oscc_has_objects and see OSC has no objests and call
1765 oscc_internal_create to resend create request.
1767 Severity : enhancement
1769 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1771 Severity : enhancement
1773 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1776 Frequency : very rare
1778 Description: Fix replay if there is an un-replied request and open
1779 Details : In some cases, older replay request will revert the
1780 mcd->mcd_last_xid on MDS which is used to record the client's
1781 latest sent request.
1783 Severity : enhancement
1785 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1787 Severity : enhancement
1789 Description: Add rhel5 support to HEAD.
1791 Serverity : enhancement
1793 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1798 Description: Don't allow skipping OSTs if index has been specified.
1799 Details : Don't allow skipping OSTs if index has been specified, make
1800 locking in internal create lots better.
1804 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1805 Details : access to bitfield in structure is always rounded to long
1806 and this produce problem with not atomic change any bit.
1810 Description: Lustre make rpms failed.
1811 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1812 builds Lustre rpms from tarball.
1815 Frequency : rare on shutdown ost
1817 Description: If llog cancel was not send before clean_exports phase, this can
1818 produce deadlock in llog code.
1819 Details : If llog thread has last reference to obd and call class_import_put
1820 this produce deadlock because llog_cleanup_commit_master wait when
1821 last llog_commit_thread exited, but this never success because was
1822 called from llog_commit_thread.
1826 Description: allow userland application know is lost one of stripes.
1827 Details : fill lvb_blocks with error code on ost and return it to
1828 application if error flag found.
1832 Description: NULL lov_tgts causing MDS oops
1833 Details : more safe checks for NULL lov_tgts for avoid oops.
1835 Severity : enhancement
1837 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1841 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1842 Details : If an OST is down the MDS will hang indefinitely in
1843 obd_statfs_rqset() waiting for the statfs data. While for
1844 MDS QOS usage of statfs, it should not stuck in waiting.
1846 Severity : enhancement
1848 Description: remote_acl support
1849 Details : Support ACL-based permission check for remote user.
1850 Support setfacl/getfacl for remote user with the utils
1851 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1852 the system "{s,g}etfacl" utils.
1854 Severity : enhancement
1856 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1858 Severity : enhancement
1860 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1864 Description: make mgs_setparam() handle fsname containing dash
1865 Details : fsname containing a dash does not work with lctl conf_param
1867 Severity : enhancement
1869 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1871 Severity : enhancement
1873 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1877 Frequency : rare, at shutdown
1878 Description: access already free / zero obd_namespace.
1879 Details : if client_disconnect_export was called without force flag set,
1880 and exist connect request in flight, this can produce access to
1881 NULL pointer (or already free pointer) when connect_interpret
1882 store ocd flags in obd_namespace.
1886 Frequency : only at startup
1887 Description: not alloc memory with spinlock held.
1888 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1889 if any spinlock held.
1891 Severity : enhancement
1893 Description: make lustre randomly fail allocating memory
1894 Details : Make lustre randomly failed allocating memory for testing purpose.
1896 Severity : enhancement
1898 Description: lost problems with lov objid file
1899 Details : Fixes some scability and access to not inited memory problems
1900 in work with lov objdid file.
1905 Description: lfs find does not continue on file error
1906 Details : Continue other files processing when a file/dir is absent.
1910 Description: Inconsistent usage of lustre_pack_reply()
1911 Details : Standardize the usage of lustre_pack_reply() such that it
1912 always generate a CERROR on failure.
1915 Frequency : occasional
1917 Description: MDS hang or stay in waiting lock
1918 Details : If client receive lock with CBPENDING flag ldlm need send lock
1919 cancel as separate rpc, to avoid situation when cancel request
1920 can't processed due all i/o threads stay in wait lock.
1924 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1928 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1932 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1936 Description: Update to RHEL4 latest kernel.
1938 Severity : enhancement
1940 Description: Build SLES10 patchless client fails
1941 Details : The configure was broken by run ./configure with
1942 --with-linux-obj=.... argument for patchless client. When the
1943 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1944 can't search header adequately. Use absolute path such as
1945 -I($LINUX)/include instead.
1949 Description: interrupt oig_wait produce painc on resend.
1950 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1951 private set, and this produce situation when rq_ptlrpcd_data not
1952 copyed to new allocated request and triggered LBUG on assert
1953 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1954 ptlrpcd set if request is changed and can be safety changed to use
1959 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1960 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1961 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1962 these macros and replace them with their difinitions instead.
1966 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1967 Details : Update which_patch & target file for SLES10 latest kernel.
1970 Description: Remove client patches from SLES 10 kernel.
1971 Details : This causes SLES 10 clients to behave as patchless clients
1972 even on a Lustre-patched (server) kernel.
1974 Severity : enhancement
1976 Description: self-adjustable client's lru lists
1977 Details : use adaptive algorithm for managing client cached locks lru
1978 lists according to current server load, other client's work
1979 pattern, memory activities, etc. Both, server and client
1980 side namespaces provide number of proc tunables for controlling
1983 Severity : enhancement
1985 Description: light-weight GSS support
1986 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1987 and reduce performance overhead.
1989 Severity : enhancement
1991 Description: Linux keyring support
1992 Details : Support using service of Linux keyring for Lustre GSS internal
1993 context refresh/cache mechanism.
1997 Description: Fix errors in lfs documentation
1998 Details : Fixes man pages
2002 Description: don't use GFP_* in generic Lustre code.
2003 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2007 Description: obdclass is limited by single OBD_ALLOC(idarray)
2008 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2012 Description: warnings with build patchless client with vanila 2.6.19 and up
2013 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2014 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2018 Frequency : Only for SLES
2019 Description: sanity_quota fail test_1
2020 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2021 file, which makes TSTID[2] inunique.
2025 Frequency : Only for SLES9
2026 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2027 Details : Update target/ChangeLog/which_patch .
2031 Frequency : Only for test_10 in sanity-quota.sh
2032 Description: a bug in quota test code
2033 Details : checking a test flag in a wrong place
2037 Frequency : Only for test_13 in sanity-quota.sh
2038 Description: a bug in quota test script
2039 Details : shouldn't check overfull allocated quota
2043 Description: jbd statistics
2044 Details : Port older jbd statistics patch for sles10
2046 Severity : enhancement
2048 Description: Add printing inode info into message about error in writepage.
2053 Description: reply_lock_interpret crash due to race with it and lock cancel.
2054 Details : Do not replay locks that are being cancelled. Do not reference
2055 locks by their address during replay, just by their handle.
2059 Frequency : When flocks are used.
2060 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2061 Details : Flock locks might destroy just granted lock if it could be merged
2062 with another existing flock, this is done in completion handler,
2063 so teach ldlm_cli_enquque_fini that this is a valid case for
2069 Description: Oops in read and write path when failing to allocate lock.
2070 Details : Check if lock allocation failed and return error back.
2074 Description: lstripe command fails for valid OST index
2075 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2076 instead of lov->desc.ld_active_tgt_count.
2080 Description: sanity.sh failed test 103
2081 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2086 Description: df doesn't work properly if diskfs blocksize != 4K
2087 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2091 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2092 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2094 Severity : enhancement
2096 Description: Adaptive timeouts
2097 Details : RPC timeouts adapt to changing server load and network
2098 conditions to reduce resend attempts and improve recovery time.
2102 Description: llapi_file_create() does not allow some changes
2103 Details : add llapi_file_open() that allows specifying the mode and
2104 open flags, and also returns an open file handle.
2108 Description: merge and cleanup kernel patches.
2109 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2113 Description: Add journal checksum support.(Kernel part)
2114 Details : The journal checksum feature adds two new flags i.e
2115 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2116 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2117 indicates that the commit block contains the checksum for
2118 the blocks described by the descriptor blocks. Now commit
2119 record can be sent to disk without waiting for descriptor
2120 blocks to be written to disk. This behavior is controlled
2121 using JBD2_FEATURE_ASYNC_COMMIT flag.
2125 Description: OSS needs mutliple precreate threads
2126 Details : Add ability to start more than one create thread per OSS.
2130 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2131 Details : execute lfs setstripe on client
2135 Description: mds_obd_create error creating tmp object
2136 Details : When the user sets quota on root, llog will be affected and can't
2137 create files and write files.
2140 Frequency : Always on ia64 patchless client, and possibly others.
2142 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2143 Details : This allows the patchless client to be loaded on architectures
2144 without this export.
2148 Description: RedHat Update kernel for RHEL5
2149 Details : Add the kernel config file for RHEL5.
2153 Description: RedHat Update kernel for RHEL5
2154 Details : Modify the kernel config file more closer RHEL5.
2158 Description: Build failure against Centos5 (RHEL5)
2159 Details : Define PAGE_SIZE when it isn't present.
2163 Description: test_7 on sanity-quota.sh can't be used on separate machines
2164 Details : add facet to handle it
2168 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2169 Details : with NFS, the anon dentry's parent was set to itself in
2170 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2171 corresponding dentry other than use rec->ur_name.
2173 Severity : enhancement
2175 Description: lfs setstripe enhancement
2176 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2180 Description: enable data checksumming by default
2181 Details : enable checksum by default, allow --disable-checksum
2182 configure option and "-o nochecksum" mount option. Checksums
2183 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2187 Description: lustre support for RHEL5
2188 Details : Add support for RHEL5.
2192 Description: Client eviction due to failover config
2193 Details : after a connection loss, the lustre client should attempt to
2194 reconnect to the last active server first before trying the
2195 other potential connections.
2199 Description: when mds and osts use different quota unit(32bit and 64bit),
2200 quota will be released repeatly.
2201 Details : void sending multiple quota reqs to mds, which will keep the
2202 status between the reqs.
2206 Description: osts not allocated evenly to files
2207 Details : change the condition to increase offset_idx
2210 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2212 Description: Data corruption for OSTs that are formatted larger than 2TB
2214 Details : When generating the bio request for lustre file writes the
2215 sector number would overflow a temporary variable before being
2216 used for the IO. The data reads correctly from Lustre (which
2217 will overflow in a similar manner) but other file data or
2218 filesystem metadata may be corrupted in some cases.
2222 Description: Tune the kernel for good SCSI performance.
2223 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2224 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2229 Description: rewrite ext2-derived code in obdclass/uuid.c
2230 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2231 from scratch preserving functionality.
2235 Description: "lfs find -obd UUID" prints directories
2236 Details : "lfs find -obd UUID" will return all directory names instead
2237 of just file names. It is incorrect because the directories
2238 do not reside on the OSTs.
2242 Description: use i_size_read and i_size_write in 2.6 port
2243 Details : replace inode->i_size access with i_size_read/write()
2246 Frequency : when removing large files
2248 Description: scheduling issue during removal of large Lustre files
2249 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2250 It causes scheduling issues when removing large files (17TB in the
2254 Frequency : only with liblustre clients on XT3
2256 Description: evictions taking too long
2257 Details : allow llrd to evict clients directly on OSTs
2260 Frequency : only on ppc
2262 Description: /proc/fs/lustre/devices broken on ppc
2263 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2264 We should make sure the type of 'index' is loff_t and then cast
2265 explicitly as needed below. Do not assign an explicitly cast
2269 Frequency : only for rhel5
2271 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2272 Details : Modify the target file & which_kernel.
2274 Severity : enhancement
2276 Description: omit set fsid for export NFS
2277 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2278 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2282 Frequency : Always, for kernels after 2.6.16
2283 Description: Fix warning idr_remove called for id=.. which is not allocated.
2284 Details : Last kernels save old s_dev before kill super and not allow
2285 to restore from callback - restore it before call kill_anon_super.
2289 Description: buffer overruns could theoretically occur
2290 Details : llapi_semantic_traverse() modifies the "path" argument by
2291 appending values to the end of the origin string, and a
2292 overrun may occur. Adding buffer overrun check in liblustreapi.
2296 Description: Fix error on 'ls .' at the top of the Lustre mount.
2297 Details : Don't revalidate dentry if it is a root dentry.
2301 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2302 Details : Modify vm-tunables-rhel4.patch.
2306 Description: Kernel config for 2.6.18-vanilla.
2307 Details : Modify targets/2.6-vanilla.target.in.
2308 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2309 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2310 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2311 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2315 Description: improve handling recoverable errors
2316 Details : if request processig with error which can be recoverable on server
2317 request should be resend, otherwise page released from cache and
2322 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2323 Details : Modify target file & which_patch.
2324 A flaw was found in the IA32 system call emulation provided
2325 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2326 value could be stored in the %RAX register, which could trigger an
2327 out-of-bounds system call table access. An untrusted local user
2328 could exploit this flaw to run code in the kernel
2329 (ie a root privilege escalation). (CVE-2007-4573).
2333 Description: O_DIRECT bypasses client statistics.
2334 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2335 read_ahead_stats were updated. copied the stats section from
2336 osc_send_oap_rpc() into async_internal().
2340 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2344 Description: change order of libsysio includes
2345 Details : '#include sysio.h' should always come before '#include xtio.h'
2349 Description: handle "serious error: objid * already exists" more gracefully
2350 Details : If LAST_ID value on disk is smaller than the objects existing in
2351 the O/0/d* directories, it indicates disk corruption and causes an
2352 LBUG(). If the object is 0-length, then we should use the existing
2353 object. This will help to avoid a full fsck in most cases.
2355 Severity : enhancement
2357 Description: adapt the lustre_config script to support the upgrade case
2358 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2359 server targets to 1.6 in parallel.
2363 Description: To avoid grant space > avaible space when the disk is almost
2364 full. Without this patch you might see the error "grant XXXX >
2365 available" or some LBUG about grant, when the disk is almost
2367 Details : In filter_check_grant, for non_grant cache write, we should
2368 check the left space by if (*left > ungranted + bytes), instead
2369 of (*left > ungranted), because only we are sure the left space
2370 is enough for another "bytes", then the ungrant space should be
2371 increase. In client, we should update cl_avail_grant only there
2372 is OBD_MD_FLGRANT in the reply.
2376 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2377 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2378 described in CVE-2007-4573. This problem could allow untrusted
2379 local users to gain root access.
2382 Frequency : when using O_DIRECT and quotas
2384 Description: Incorrect file ownership on O_DIRECT output files
2385 Details : block usage reported by 'lfs quota' does not take into account
2386 files that have been written with O_DIRECT.
2391 Description: touch file failed when fs is not full
2392 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2393 otherwise we can get ENOSP while fs is not full.
2397 Description: parallel lock callbacks
2398 Details : Instead of sending blocking and completion callbacks as separated
2399 requests, adding them to a set and sending in parallel.
2402 Frequency : only for Cray XT3
2403 Bugzilla : 12829/13455
2404 Description: Changing primary group doesn't change the group lustre assigns to
2406 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2407 overridden with the primary group provided by the group upcall,
2408 whereas the supplied fsgid can be trusted if it is in the list of
2409 supplementary groups returned by the group upcall.
2411 Severity : enhancement
2413 Description: Allow masking D_WARNING, D_ERROR messages from console
2414 Details : Console messages can now be disabled via lnet.printk.
2418 Description: User code with malformed file open parameter crashes client node
2419 Details : Before packing join_file req, all the related reference should be
2420 checked carefully in case some malformed flags cause fake
2421 join_file req on client.
2425 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2426 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2427 mode is equal to the granted mode. However, it should also check
2428 that LDLM_FL_CP_REQD is not set.
2432 Description: Heavy nfs access might result in deadlocks
2433 Details : After ELC code landed, it is now improper to enqueue any mds
2434 locks under och_sem, because enqueue might want to decide to
2435 cancel open locks for same inode we are holding och_sem for.
2439 Description: Client eviction while running blogbench
2440 Details : A lot of unlink operations with concurrent I/O can lead to a
2441 deadlock causing evictions. To address the problem, the number of
2442 oustanding OST_DESTROY requests is now throttled to
2443 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2444 callbacks are processed in priority.
2448 Description: enable ACLs on MDS by default
2449 Details : ACLs must be enabled on MDS by default.
2452 Frequency : PPC/PPC64 only
2454 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2455 Details : fix duplicated definitions between asm-ppc64/types.h and
2456 lustre_types.h on PPC.
2459 Frequency : PPC/PPC64 only
2461 Description: asm-ppc/segment.h does not exist
2462 Details : fix compile issue on PPC.
2466 Description: better handle error messages in extents code
2469 Frequency : RHEL4 only
2471 Description: mkfs is very slow on IA64/RHEL4
2472 Details : A performance regression has been discovered in the MPT Fusion
2473 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2474 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2475 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2477 Severity : enhancement
2479 Description: SNMP support enhancement
2480 Details : Adding total number of sampled request for an MDS node in snmp
2483 Severity : enhancement
2485 Description: Optimize ldlm waiting list processing for PR extent locks
2486 Details : When processing waiting list for read extent lock and meeting read
2487 lock that is same or wider to it that is not contended, skip
2488 processing rest of the list and immediatelly return current
2489 status of conflictness, since we are guaranteed there are no
2490 conflicting locks in the rest of the list.
2494 Description: Time out and refuse to reconnect
2495 Details : When the failover node is the primary node, it is possible
2496 to have two identical connections in imp_conn_list. We must
2497 compare not conn's pointers but NIDs, otherwise we can defeat
2498 connection throttling.
2502 Description: port llog fixes from b1_6 into HEAD
2503 Details : Port llog reference couting and some llog cleanups from b1_6
2504 (bug 10800) into HEAD, for protect from panic and access to already
2505 free llog structures.
2509 Description: Detect stride IO mode in read-ahead
2510 Details : When a client does stride read, read-ahead should detect that and
2511 read-ahead pages according to the detected stride pattern.
2515 Description: data checksumming impacts single node performance
2516 Details : add support for several checksum algorithm. Currently, only CRC32
2517 and Adler-32 are supported. The checksum type can be changed on
2518 the fly via /proc/fs/lustre/osc/*/checksum_type.
2522 Description: use adler32 for page checksums
2523 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2528 Description: build for x2 fails
2529 Details : fix compile issue on Cray systems.
2533 Description: Properly match for duplicate locks
2534 Details : Due to different lock order from skiplists code, we need to
2535 traverse entire list for now
2538 Frequency : only on PPC/SLES10
2540 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2541 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2542 be defined. Add a hack in mkfs_lustre.c to work around this compile
2547 Description: LASSERT on MDS when client holding flock lock dies
2548 Details : ldlm pool logic depends on number of granted locks equal to
2549 number of released locks which is not true for flock locks, so
2550 just exclude such locks from consideration.
2554 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2555 Details : Use fsfilt_sync() for both the whole filesystem sync and
2556 individual file sync to eliminate dangerous inode locking
2557 with I_LOCK that can lead to a deadlock.
2561 Description: performance in 1.6.3
2562 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2563 sure that 1MB requests are not fragmented by the block layer.
2565 Severity : enhancement
2567 Description: organize the server-side client stats on per-nid basis
2568 Details : Change the structure of stats under obdfilter and mds to
2576 The "uuid"s file would list the uuids of _active_ exports.
2577 And the clear entry is to clear all stats and stale nids.
2579 Severity : enhancement
2581 Description: eliminate client locks in face of contention
2582 Details : file contention detection and lockless i/o implementation
2583 for contended files.
2587 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2590 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2592 Description: MDS may lose file striping (and hence file data) in some cases
2593 Details : If there are additional extended attributes stored on the MDS,
2594 in particular ACLs, SELinux, or user attributes (if user_xattr
2595 is specified for the client mount options) then there is a risk
2596 of attribute loss. Additionally, the Lustre file striping
2597 needs to be larger than default (e.g. striped over all OSTs),
2598 and an additional attribute must be stored initially in the
2599 inode and then increase in size enough to be moved to the
2600 external attribute block (e.g. ACL growing in size) for file
2605 Description: skiplist implementation simplification
2606 Details : skiplists are used to group compatible locks on granted list
2607 that was implemented as tracking first and last lock of each
2608 lock group the patch changes that to using doubly linked lists
2612 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2613 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2614 because LASSERT on both the data supplied by a client, and the
2615 data on disk is dangerous and incorrect.
2617 Severity : enhancement
2619 Description: Slow truncate/writes to huge files at high offsets.
2620 Details : Directly associate cached pages to lock that protect those pages,
2621 this allows us to quickly find what pages to write and remove
2622 once lock callback is received.
2626 Description: more ldlm soft lockups
2627 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2628 starve other threads from the resource lock for a long time in
2629 case of long waiting queue, so change the debug level from
2630 D_OTHER to the less frequently used D_INFO.
2632 Severity : enhancement
2634 Description: add -gid, -group, -uid, -user options to lfs find
2638 Description: Hung threads in invalidate_inode_pages2_range
2639 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2640 once one is completed. As a result, some RPCs are stuck in the
2641 queue and are never sent.
2645 Description: filter threads hungs on waiting journal commit
2646 Details : Cleanup filter group llog code, then only filter group llog will
2647 be only created in the MDS/OST syncing process.
2651 Description: Procfs and llog threads access destoryed import sometimes.
2652 Details : Sync the import destoryed process with procfs and llog threads by
2653 the import refcount and semaphore.
2655 Severity : enhancement
2657 Description: openlock cache of b1_6 port to HEAD
2662 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2663 Details : If insertion of an extent fails, then discard the inode
2664 preallocation and free data blocks else it can lead to duplicate
2669 Description: don't always update ctime in ext3_xattr_set_handle()
2670 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2671 In some cases the ctime should not be updated, for example for
2672 2.0->1.8 compatibility it is necessary to delete an xattr and it
2673 should not update the ctime.
2677 Bugzilla : 15713/16362
2678 Description: Assertion in iopen_connect_dentry in 1.6.3
2679 Details : looking up an inode via iopen with the wrong generation number can
2680 populate the dcache with a disconneced dentry while the inode
2681 number is in the process of being reallocated. This causes an
2682 assertion failure in iopen since the inode's dentry list contains
2683 both a connected and disconnected dentry.
2687 Description: assertion failure in ldlm_handle2lock()
2688 Details : fix a race between class_handle_unhash() and class_handle2object()
2689 introduced in lustre 1.6.5 by bug 13622.
2694 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2695 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2700 Description: oops in sd_iostats_seq_show()
2701 Details : unloading/reloading the scsi low level driver triggers a kernel
2702 bug when trying to access the sd iostat file.
2707 Description: Kernel panics during QLogic driver reload
2708 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2709 patch, causing memory corruption.
2714 Description: journal_dev option does not work in b1_6
2715 Details : pass mount option during pre-mount.
2717 Severity : enhancement
2719 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2720 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2721 extent information of a file. It can be used to map logical blocks
2722 in a file to physical blocks in the block device.
2726 Description: LDLM soft lockups - improvement
2727 Details : It is be possible to send the lock handle along with each read
2728 or write request because the client is already doing a lock match
2729 itself so there isn't any reason the OST should have to re-do that
2733 Frequency : only on Cray X2
2735 Description: X2 build failures
2736 Details : fix build failures on Cray X2.
2740 Description: xid & resent requests
2741 Details : Initialize RPC XID from clock at startup (randomly if clock is
2744 Severity : enhancement
2746 Description: Add lustre_start utility to start or stop multiple Lustre servers
2751 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2752 Details : In case of memory pressure, list_del() can be called twice on
2753 req->rq_history_list, causing a kernel oops.
2757 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2758 Details : fix stack overflow in the distributed lock manager by defering
2759 export eviction after a failed AST to the elt thread instead of
2760 handling it in the dlm interpret routine.
2764 Description: Convert some comments to new format.
2765 Details : Update documenting comments to match doxygen conventions.
2769 Description: Grammar fixes.
2770 Details : A couple of trivial spelling fixes.
2774 Description: OSD_COUNTERS-mandatory
2775 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2776 to simplify the code.
2780 Description: simplify cmm_device freeing logic.
2781 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2785 Description: Add lockdep support to dt_object_operations locking interface.
2786 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2787 parameter indicating lock ordering. Update mdd code to use new
2792 Description: Introduce failloc constants for lockless IO tests.
2793 Details : Add two new failloc constants to test lockless IO. Only one of
2794 them in implemented---another is checked in yet to be landed
2799 Description: Add lockdep support for inode mutex.
2800 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2801 in the situations where Lustre has to lock more than one inode
2806 Description: Add optional invariants checking support.
2807 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2808 configure switch. This macro is to be used for consistency and
2809 sanity checks that are too expensive to be left in `production'
2814 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2815 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2820 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2821 Details : Use spin_lock_nested() in (the only) situation where more than
2822 one ldlm_lock is locked simultaneously. Also, fix possible
2823 dead-lock in ldlm_lock_change_resource() by enforcing particular
2828 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2829 Details : Instead of specifying each ldlm_lock call-back through separate
2830 parameter, wrap them into struct ldlm_callback_suite.
2834 Description: Kill join_lru obd method and its callers.
2835 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2836 in memory---a requirement implemented through ->o_join_lru() obd
2837 method in HEAD. Remove this method and its users.
2841 Description: Add asynchronous ldlm ENQUEUE completion handler.
2842 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2843 case that stock ldlm_completion_ast() cannot handle as it waits
2844 until lock is granted. Introduce new ldlm_completion_ast_async()
2845 for this. Also comment ldlm_completion_ast().
2849 Description: ldlm_error <-> errno conversion.
2850 Details : Add functions to map (rather arbitrary) between LDLM error codes
2851 and standard errno values. CLIO needs this to prevent LDLM specific
2852 constants from escaping ldlm and osc.
2856 Description: Kill unused ldlm_handle2lock_ns() function.
2857 Details : Kill unused ldlm_handle2lock_ns() function.
2861 Description: Add lu_ref support to ldlm_lock
2862 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2863 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2864 and ldlm_resource. LDLM interface has to be changed, because code that
2865 releases a reference on a lock, has to "know" what reference this is.
2866 In the most frequent case
2868 lock = ldlm_handle2lock(handle);
2870 LDLM_LOCK_PUT(lock);
2872 no changes are required. When any other reference (received _not_ from
2873 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2874 instead of LDLM_LOCK_PUT().
2876 Arguably, changes are pervasive, and interface requires some discipline
2877 for proper use. On the other hand, it was very instrumental in finding
2878 a few leaked lock references.
2882 Description: Add ldlm_lock_addref_try().
2883 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2884 attempts to addref a lock that might be being canceled
2889 Description: Add ldlm_weigh_callback().
2890 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2891 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2892 hard-coded `number of pages' logic.
2896 Description: Add lockdep annotations to llog code.
2897 Details : Use appropriately tagged _nested() locking calls in the places
2898 where llog takes more than one ->lgh_lock lock.
2902 Description: Add loi_kms_set().
2903 Details : Wrap kms updates into a helper function.
2907 Description: Constify instances of struct lsm_operations.
2908 Details : Constify instances of struct lsm_operations.
2912 Description: lu_conf support.
2913 Details : On a server, a file system object is uniquely identified
2914 by a fid, which is sufficient to locate and load all object
2915 state (inode). On a client, on the other hand, more data are
2916 necessary instantiate an object. Change lu_object_find() and
2917 friends to take additional `lu_conf' argument describing object.
2918 Typically this includes layout information.
2922 Description: lu_context fixes.
2923 Details : Introduce new lu_context functions that are needed on the client
2924 side, where some system threads (ptlrpcd) are shared by multiple
2925 modules, and so cannot be stopped during module shutdown.
2929 Description: Add start and stop methods to lu_device_type_operations.
2930 Details : Introduce two new methods in lu_device_type_operations, that are
2931 invoked when first instance of a given type is created and last one
2932 is destroyed respectively. This is need by CLIO.
2936 Description: Add lu_ref support to struct lu_device.
2937 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2938 track leaked references.
2942 Description: Introduce lu_kmem_descr.
2943 Details : lu_kmem_descr and its companion interface allow to create
2944 and destroy a number of kmem caches at once.
2948 Description: Fix lu_object finalization race.
2949 Details : Fix a race between lu_object_find() finding an object and its
2950 concurrent finalization. This race is (most likely) not possible
2951 on the server, but might happen on the client.
2955 Description: Introduce lu_ref interface.
2956 Details : lu_ref is a debugging module allowing to track references to
2957 a given object. It is quite cpu expensive, and has to be
2958 explicitly enabled with --enable-lu_ref. See usage description
2963 Description: Factor lu_site procfs stats into a separate function.
2964 Details : Separate lu_site stats printing code into a separate function
2965 that can be reused on a client.
2969 Description: Constify instances of struct {lu,dt,md}_device_operations.
2970 Details : Constify instances of struct {lu,dt,md}_device_operations.
2974 Description: Introduce struct md_site and move meta-data specific parts of
2975 struct lu_site here.
2976 Details : Move md-specific fields out of struct lu_site into special struct
2977 md_site, so that lu_site can be used on a client.
2981 Description: Kill mdd_lov_destroy().
2982 Details : Remove unused mdd code.
2986 Description: Add st_block checking to multistat.c.
2987 Details : Add st_block checking to multistat.c.
2991 Description: Add lu_ref support to struct obd_device.
2992 Details : Add lu_ref tracking to obd_device.
2996 Description: Kill obd_set_fail_loc().
2997 Details : Remove unused code.
3001 Description: Add special type for ptlrpc_request interpret functions.
3002 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
3003 there. Actual usage will be in CLIO.
3007 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
3008 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
3009 doesn't work with lockdep.
3013 Description: Add rwv.c test program.
3014 Details : New testing program exercising readv(2) and writev(2) (Qian).
3018 Description: Add sendfile.c test program.
3019 Details : New testing program exercising sendfile(2) (Jay).
3023 Description: Ratelimit a message that can be very frequent.
3024 Details : Ratelimit a memory allocation failure message that can
3029 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3030 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3034 Description: Liblustre build fixes.
3035 Details : Liblustre build fixes.
3039 Description: libcfs: add cfs_{need,cond}_resched() interface.
3040 Details : libcfs: add cfs_{need,cond}_resched() definition and
3041 implementations for Linux, NT, and liblustre.
3043 Severity : enhancement
3045 Description: More exported tunables for mballoc
3046 Details : Add support for tunable preallocation window and new tunables for
3047 large/small requests
3051 Description: Detect corruption of block bitmap and checking for preallocations
3052 Details : Checks validity of on-disk block bitmap. Also it does better
3053 checking of number of applied preallocations. When corruption is
3054 found, it turns filesystem readonly to prevent further corruptions.
3058 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3059 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3060 window to 0. If it does hit cache miss, and read-pattern is still
3061 stride-io mode, does not reset the stride window, but also does not
3062 increase the stride window length in this case.
3066 Frequency : only for big-endian servers
3067 Description: Check if system is big-endian while mounting fs with extents feature
3068 Details : Mounting a filesystem with extents feature will fail on big-endian
3069 systems since ext3-based ldiskfs is not supported on big-endian
3070 systems. This can be over-riden with "bigendian_extents" mount option.
3072 Severity : enhancement
3074 Description: The root squash functionality
3075 Details : A security feature, which is to prevent users from being able
3076 to mount lustre on their desktop, run as root, and delete
3077 all of the files in the filesystem. The goal is accomplished by
3078 remapping user id (UID) and group id (GID) of the root user to
3079 a UID and GID specified by the system administartor via Lustre
3080 configuration management server (MGS). The functionality also
3081 allows to specify sets of clients for which the remapping does
3086 Description: Excessive recovery window
3087 Details : With AT enabled, the recovery window can be excessively long (6000+
3088 seconds). To address this problem, we no longer use
3089 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3090 timeout no longer depends on the service time, it is set to
3091 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3092 time via pb_service_time.
3096 Description: Watchdog triggered on MDS failover
3097 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3098 the OSTs know that the MDS "UUID" can be reused for the same export
3099 from a different NID, so we do not need to wait for the export to be
3103 Frequency : rare, only if using MMP with Linux RAID
3105 Description: MMP doesn't work with Linux RAID
3106 Details : While using HA for Lustre servers with Linux RAID, it is possible
3107 that MMP will not detect multiple mounts. To make this work we
3108 need to unplug the device queue in RAID when the MMP block is being
3109 written. Also while reading the MMP block, we should read it from
3110 disk and not the cached one.
3112 Severity : enhancement
3114 Description: open file using fid
3115 Details : A file can be opened using just its fid, like
3116 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3119 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3122 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3123 Details : add proper locking for creation of jbd_4k slab cache
3127 Description: MMP check in ext3_remount() fails without displaying any error
3128 Details : When multiple mount protection fails during remount, proper error
3131 Severity : enhancement
3133 Description: Allow stripe size to be up to 4G-64k
3134 Details : Fix math logic to allow large stripe sizes.
3138 Description: add check for >8TB ldiskfs filesystems
3139 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3140 Don't allow >8TB ldiskfs filesystems to be mounted without
3141 force_over_8tb mount option
3146 Descriptoin: OSS complains "too many blocks" and crashes
3147 Details : In filter_commitrw_write():
3148 1. lquota_chkquota();
3149 2. obdo_from_inode();
3150 3. lquota_pending_commit();
3151 "2." has changed uid and gid in this case, then caused
3152 lquota_chkquota() and lquota_pending_commit() to use
3153 different uid and gid. That is the root of the bug.
3157 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3158 Details : Detect and print out the 'stripe count = -1' case properly.
3162 Description: Out or order replies might be lost on replay
3163 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3164 request with tid smaller than one currently being added, add it
3165 to the start, not end of the list.
3167 --------------------------------------------------------------------------------
3169 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3171 * Support for kernels:
3172 2.6.5-7.283 (SLES 9),
3173 2.6.9-55.EL (RHEL 4),
3174 2.6.16.46-0.14 (SLES 10),
3175 2.6.18.8 vanilla (kernel.org)
3176 * Client support for unpatched kernels:
3177 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3178 2.6.16 - 2.6.22 vanilla (kernel.org)
3179 * Due to recently discovered recovery problems, we do not recommend
3180 using patchless RHEL 4 clients with this or any earlier release.
3181 * Recommended e2fsprogs version: 1.39.cfs8
3182 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3183 * Starting with this release, the ldiskfs backing filesystem required
3184 by Lustre is now in its own package, lustre-ldiskfs. This package
3185 should be installed. It is versioned separately from Lustre and
3186 may be released separately in future.
3190 Description: block reactivating mgc import until all deactivates complete
3191 Details : Fix race when failing back MDT/MGS to itself (testing)
3194 Frequency : at statup only
3196 Description: mds_lov_synchronize race leads to various problems
3197 Details : simultaneous MDT->OST connections at startup can cause the
3198 sync to abort, leaving the OSC in a bad state.
3200 Severity : enhancement
3202 Description: add optional extra BUILD_VERSION info
3203 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3204 to override the lustre version.
3207 Frequency : 2.6.18 servers only
3209 Description: ll_kern_mount() doesn't release the module reference
3210 Details : The ldiskfs module reference count never drops down to 0
3211 because ll_kern_mount() doesn't release the module reference.
3216 Description: server LBUG when using old ost_num_threads parameter
3217 Details : Accept the old ost_num_threads parameter but warn that it
3218 is deprecated, and fix an off-by-one error that caused an LBUG.
3223 Description: Transient SCSI error results in persistent IO issue
3224 Details : iobuf->dr_error is not reinitialized to 0 between two
3228 Frequency : sometimes when underlying device returns I/O errors
3230 Description: OSTs not going read-only during write failures
3231 Details : OSTs are not remounted read-only when the journal commit threads
3232 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3233 instead of the ext3 wrappers.
3237 Description: poor connect scaling with increasing client count
3238 Details : Don't run filter_grant_sanity_check for more than 100 exports
3239 to improve scaling for large numbers of clients.
3242 Frequency : SLES10 only
3244 Description: sanity-quota.sh quotacheck failed: rc = -22
3245 Details : Quotas cannot be enabled on SLES10.
3248 Frequency : liblustre clients only
3250 Description: getdirentries does not give error when run on compute nodes
3251 Details : getdirentries does not fail when the size specified as an argument
3252 is too small to contain at least one entry
3254 Severity : enhancement
3256 Description: Add LNET router traceability for debug purposes
3257 Details : If a checksum failure occurs with a router as part of the
3258 IO path, the NID of the last router that forwarded the bulk data
3259 is printed so it can be identified.
3264 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3265 Details : Due to a race condition, liblustre clients were occasionally
3266 evicted incorrectly.
3268 Severity : enhancement
3270 Description: lfs setstripe use optional parameters instead of postional
3273 Severity : enhancement
3275 Description: Nanosecond timestamp support for ldiskfs
3276 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3277 resolution timestamps. There is not yet support for this at
3278 the Lustre filesystem level.
3281 Frequency : during server recovery
3283 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3284 Details : request with rq_no_resend flag not awake l_wait_event if they get
3288 Frequency : nfs export on patchless client
3290 Description: connectathon hang when test nfs export over patchless client
3291 Details : Disconnected dentry cannot be found with lookup, so we do not need
3292 to unhash it or make it invalid
3295 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3296 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3297 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3302 Description: open req refcounting wrong on reconnect
3303 Details : If reconnect happened between getting open reply from server and
3304 call to mdc_set_replay_data in ll_file_open, we will schedule
3305 replay for unreferenced request that we are about to free.
3306 Subsequent close will crash in variety of ways.
3307 Check that request is still eligible for replay in
3308 mdc_set_replay_data().
3313 Description: disable writes to filesystem when reading health_check file
3314 Details : the default for reading the health_check proc file has changed
3315 to NOT do a journal transaction and write to disk, because this
3316 can cause reads of the /proc file to hang and block HA state
3317 checking on a healthy but otherwise heavily loaded system. It
3318 is possible to return to the previous behaviour during configure
3319 with --enable-health-write.
3321 Severity : enhancement
3323 Description: 64-bit inode version
3324 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3325 to the inode. This will be required for version-based recovery.
3330 Description: MDS fails to start if a duplicate client export is detected
3331 Details : in some rare cases it was possible for a client to connect to
3332 an MDS multiple times. Upon recovery the MDS would detect this
3333 and fail during startup. Handle this more gracefully.
3335 Severity : enhancement
3337 Description: Add -o localflock option to simulate old noflock
3339 Details : This will achieve local-only flock/fcntl locks
3345 Description: log_commit_thread vs filter_destroy race leads to crash
3346 Details : Take import reference before releasing llog record semaphore
3351 Description: Wrong request locking in request set processing
3352 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3353 imp_delayed_list, in this place should be used imp_lock.
3356 Frequency : when reconnection
3358 Description: Grant Leak when osc reconnect to OST
3359 Details : When osc reconnect ost, OST(filter) should check whether it
3360 should grant more space to client by comparing fed_grant and
3361 cl_avail_grant, and return the granted space to client instead
3362 of "new granted" space, because client will call osc_init_grant
3363 to update the client grant space info.
3366 Frequency : when client reconnect to OST
3368 Description: Grant Leak when osc do resend and replay bulk write
3369 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3370 bulk write request, because the grant info while be sync between
3371 OSC and OST when reconnect, and we should ignore the grant info
3372 these of resend/replay write req.
3377 Description: Grant space more than available left space sometimes.
3378 Details : When then OST is about to be full, if two bulk writing from
3379 different clients came to OST. Accord the avaliable space of the
3380 OST, the first req should be permitted, and the second one
3381 should be denied by ENOSPC. But if the seconde arrived before
3382 the first one is committed. The OST might wrongly permit second
3383 writing, which will cause grant space > available space.
3386 Frequency : when client is evicted
3388 Description: Grant might be wrongly erased when osc is evicted by OST
3389 Details : when the import is evicted by server, it will fork another
3390 thread ptlrpc_invalidate_import_thread to invalidate the
3391 import, where the grant will be set to 0. While the original
3392 thread will update the grant it got when connecting. So if
3393 the former happened latter, the grant will be wrongly errased
3394 because of this race.
3399 Description: Checking Stale with correct fid
3400 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3401 to check whether it is stale, because sometimes, we want the
3402 enqueue happened anyway, and op_data.fid2 will not be initialized.
3404 Severity : enhancement
3406 Description: update patchless client
3407 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3410 Frequency : only with 2.4 kernel
3412 Description: random memory corruption
3413 Details : size of struct ll_inode_info is to big for union inode.u and this
3414 can be cause of random memory corruption.
3419 Description: Memory leak in recovery
3420 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3421 It should also check obd_fail before fsfilt_start, otherwise if
3422 fsfilt_start return -EROFS,(failover mds during mds recovery).
3423 then the req will return with repmsg->transno = 0 and rc = EROFS.
3424 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3425 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3426 be freed no matter whether fsfilt_commit success or not.
3429 Frequency : only with huge count clients
3431 Description: Prevents from taking the superblock lock in llap_from_page for
3433 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3434 for llap_from_page call in ll_removepage prevents from taking the
3435 superblock lock for a soon died page.
3440 Description: Not check open intent error before release open handle
3441 Details : in some rare cases, the open intent error is not checked before
3442 release open handle, which may cause
3443 ASSERTION(open_req->rq_transno != 0), because it tries to release
3444 the failed open handle.
3449 Description: Set cat log bitmap only after create log success.
3450 Details : in some rare cases, the cat log bitmap is set too early. and it
3451 should be set only after create log success.
3455 Description: Accessing a block bevice can re-enable I/O when Lustre is
3456 tearing down a device.
3457 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3462 Description: service threads may hog cpus when there are a lot of requests
3463 Details : Insert cond_resched to give other threads a chance to use some CPU
3468 Description: the cat log was not initialized in recovery
3469 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3470 unlink log on mds will not be initialized until mds post
3471 recovery. And also in mds post recovery, the unlink log will
3472 initialization will be done asynchronausly, so there will be race
3473 between add unlink log and unlink log initialization.
3477 Description: brw_stats were being printed incorrectly
3478 Details : brw_stats were being printed as log2 but all of them were not
3479 recorded as log2. Also remove some code duplication arising from
3480 filter_tally_{read,write}.
3484 Frequency : rare, only in recovery.
3485 Description: ASSERTION(req->rq_type != LI_POISON) failed
3486 Details : imp_lock should be held while iterating over imp_sending_list for
3487 prevent destroy request after get timeout in ptlrpc_queue_wait.
3491 Description: replay-single.sh test 52 fails
3492 Details : A lock's skiplist need to be cleanup when it being unlinked
3493 from its resource list.
3497 Description: Short directio read returns full requested size rather than
3499 Details : Direct I/O operations should return actual amount of bytes
3500 transferred rather than requested size.
3502 Severity : enhancement
3504 Description: metadata RPC reduction (e.g. for rm performance)
3505 Details : decrease the amount of synchronous RPC between clients and servers
3506 by canceling conflicing lock before the operation on the client
3507 and packing thier handles into the main operation RPC to server.
3509 Severity : enhancement
3511 Description: Async OSC create to avoid the blocking unnecessarily.
3512 Details : If an OST has no remaining object, system will block on the
3513 creation when it needs to create a new object on this OST. Now,
3514 ways use pre-created objects when available, instead of blocking on
3515 an empty osc while others are not empty. If we must block, we
3516 block for the shortest possible period of time.
3520 Description: improve handling recoverable errors
3521 Details : if request processig with error which can be recoverable on server
3522 request should be resend, otherwise page released from cache and
3525 Severity : enhancement
3527 Description: refine locking for avoid write wrong info into lov_objid file
3528 Details : fix possible races with add new target and write/update data in
3531 --------------------------------------------------------------------------------
3533 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3538 Frequency : on some architectures
3540 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3541 Details : Uninitialized flags sometimes cause configuration commands to
3545 Frequency : patchless clients only
3547 Description: missing __iget() symbol export
3548 Details : The __iget() symbol export is missing. To avoid the need for
3549 this on patchless clients the deathrow inode reaper is turned
3550 off, and we depend on the VM to clean up old inodes. This
3551 dependency was during via the fix for bug 12181.
3553 --------------------------------------------------------------------------------
3555 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3557 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3558 INTEROPERATE with older versions automatically. In many cases a
3559 special upgrade step is needed. Please read the
3560 user documentation before upgrading any part of a live system.
3561 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3562 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3563 * WARNING: Lustre configuration and startup changes are required with
3564 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3568 Severity : enhancement
3570 Description: MountConf
3571 Details : Lustre configuration is now managed via mkfs and mount
3572 commands instead of lmc and lconf. New obd types (MGS, MGC)
3573 are added for dynamic configuration management. See
3574 https://mail.clusterfs.com/wikis/lustre/MountConf for
3577 Severity : enhancement
3579 Description: dynamic OST addition
3580 Details : OSTs can now be added to a live filesystem
3582 Severity : enhancement
3584 Description: startup order invariance
3585 Details : MDTs and OSTs can be started in any order. Clients only
3586 require the MDT to complete startup.
3588 Severity : enhancement
3590 Description: parallel, asynchronous orphan cleanup
3591 Details : orphan cleanup is now performed in separate threads for each
3592 OST, allowing parallel non-blocking operation.
3594 Severity : enhancement
3596 Description: optimized stripe assignment
3597 Details : stripe assignments are now made based on ost space available,
3598 ost previous usage, and OSS previous usage, in order to try
3599 to optimize storage space and networking resources.
3601 Severity : enhancement
3603 Description: Permanently set tunables
3604 Details : All writable /proc/fs/lustre tunables can now be permanently
3605 set on a per-server basis, at mkfs time or on a live system.
3607 Severity : enhancement
3609 Description: Lustre message v2
3610 Details : Add lustre message format v2.
3612 Severity : enhancement
3614 Description: client OST exclusion list
3615 Details : Clients can be started with a list of OSTs that should be
3616 declared "inactive" for known non-responsive OSTs.
3620 Description: SPEC SFS validation failure on NFS v2 over lustre.
3621 Details : Changes the blocksize for regular files to be 2x RPC size,
3622 and not depend on stripe size.
3624 Severity : enhancement
3626 Description: Multiple MD RPCs in flight.
3627 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3628 To avoid overly-overloading MDS, introduce a limit on number of
3629 MDS RPCs in flight for a single client and add /proc controls
3630 to adjust this limit.
3632 Severity : enhancement
3634 Description: client read/write statistics
3635 Details : Add client read/write call usage stats for performance
3636 analysis of user processes.
3637 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3638 file access. extents_stats shows chunk size distribution.
3639 extents_stats_per_process show chunk size distribution per
3642 Severity : enhancement
3644 Description: mds statistics
3645 Details : Add detailed mds operations statistics in
3646 /proc/fs/lustre/mds/*/stats.
3650 Description: Failure of copying files with lustre special EAs.
3651 Details : Client side always return success for setxattr call for lustre
3652 special xattr (currently only "trusted.lov").
3657 Description: Refcount LNET uuids
3658 Details : The global LNET uuid list grew linearly with every startup;
3659 refcount repeated list entries instead of always adding to
3662 Severity : enhancement
3664 Description: Dynamic service threads
3665 Details : Within a small range, start extra service threads
3666 automatically when the request queue builds up.
3669 Frequency : mixed-endian client/server environments
3671 Description: mixed-endian crashes
3672 Details : The new msg_v2 system had some failures in mixed-endian
3675 Severity : enhancement
3677 Description: Easy OST removal
3678 Details : OSTs can be permanently deactivated with e.g. 'lctl
3679 conf_param lustre-OST0001.osc.active=0'
3681 Severity : enhancement
3683 Description: MGS proc entries
3684 Details : Added basic proc entries for the MGS showing what filesystems
3687 Severity : enhancement
3689 Description: provide MGS failover
3690 Details : Added config lock reacquisition after MGS server failover.
3692 Severity : enhancement
3694 Description: add Linux 2.4 support
3695 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3699 Description: a large application tries to do I/O to the same resource and dies
3700 in the middle of it.
3701 Details : Check the req->rq_arrival time after the call to
3702 ost_brw_lock_get(), but before we do anything about
3703 processing it & sending the BULK transfer request. This
3704 should help move old stale pending locks off the queue as
3705 quickly as obd_timeout.
3708 Frequency : when an incorrect nid is specified during startup
3710 Description: ptlrpc connect to non-existant node causes kernel crash
3711 Details : LNET can't be re-entered from an event callback, which
3712 happened when we expire a message after the export has been
3713 cleaned up. Instead, hand the zombie cleanup off to another
3716 Severity : enhancement
3718 Description: plain/inodebits lock performance improvement
3719 Details : Grouping plain/inodebits in granted list by their request modes
3720 and bits policy, thus improving the performance of search through
3724 Frequency : only if OST filesystem is corrupted
3726 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3727 Details : for a short time RPCs with bulk IO are in the replay list,
3728 but replay of bulk IOs is unimplemented. If the OST filesystem
3729 is corrupted due to disk cache incoherency and then replay is
3730 started it is possible to trip an assertion. Avoid putting
3731 committed RPCs into the replay list at all to avoid this issue.
3734 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3736 Description: System hang on startup
3737 Details : This bug allowed the liblustre (e.g. catamount) client to
3738 return to the app before handling all startup RPCs. This
3739 could leave the node unresponsive to lustre network traffic
3740 and manifested as a server ptllnd timeout.
3742 Severity : enhancement
3744 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3745 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3746 boolean to control whether to print peer debug info when a
3747 client's RPC times out.
3750 Frequency : only for kernels with patches from Lustre below 1.4.3
3752 Description: Remove old rdonly API
3753 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3756 Frequency : only for devices with external journals
3758 Description: Set external device read-only also
3759 Details : During a commanded failover stop, we set the disk device
3760 read-only while the server shuts down. We now also set any
3761 external journal device read-only at the same time.
3764 Frequency : when upgrading from 1.4 while trying to change parameters
3766 Description: The wrong (new) MDC name was used when setting parameters for
3767 upgraded MDT's. Also allows changing of OSC (and MDC)
3768 parameters if --writeconf is specified at tunefs upgrade time.
3771 Frequency : when setting specific ost indicies
3773 Description: QOS code breaks on skipped indicies
3774 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3775 created with --index need not be sequential.
3777 Severity : enhancement
3779 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3780 Details : The uninit_groups feature works in conjunction with the kernel
3781 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3782 up the pass1 processing of e2fsck. This is a read-only feature
3783 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3784 mount filesystems that have had this feature enabled.
3786 Severity : enhancement
3788 Description: Improve multi-block allocation algorithm to avoid fragmentation
3789 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3790 allocation locality and avoid filesystem fragmentation.
3792 ------------------------------------------------------------------------------
3794 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3796 * Support for kernels:
3797 2.4.21-47.0.1.EL (RHEL 3)
3798 2.6.5-7.283 (SLES 9)
3799 2.6.9-42.0.10.EL (RHEL 4)
3800 2.6.12.6 vanilla (kernel.org)
3801 2.6.16.27-0.9 (SLES 10)
3802 * Recommended e2fsprogs version: 1.39.cfs5
3804 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3808 Frequency : occasional, depends on client load and configuration
3809 Bugzilla : 12181, 12203
3810 Description: data loss for recently-modified files
3812 Details : In some cases it is possible that recently written or created
3813 files may not be written to disk in a timely manner (this should
3814 normally be within 30s unless client IO load is very high).
3815 The problem appears as zero-length files or files that are a
3816 multiple of 1MB in size after a client crash or client eviction
3817 that are missing data at the end of the file.
3819 This problem is more likely to be hit on clients where files are
3820 repeatedly created and unlinked in the same directory, clients
3821 have a large amount of RAM, have many CPUs, the filesystem has
3822 many OSTs, the clients are rebooted frequently, and/or the files
3823 are not accessed by other nodes after being written.
3825 The presence of the problem can be detected by looking at
3826 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3827 smaller than the second (nr_unused) then dirty files will not
3828 be flushed automatically to disk. "sync; sleep 10" should be
3829 run several times on the node before unmounting it to update
3830 Lustre (this is also safe to run on nodes without this problem).
3832 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3833 that can cause this same problem, so customers using that kernel
3834 also need to update the kernel in addition to Lustre. In order
3835 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3837 It is normal that files written just before a client crash (less
3838 than 30s) may not yet have been flushed to disk, even for local
3842 Frequency : frequent on thin XT3 nodes
3844 Description: UUID collision on thin XT3 Linux nodes
3845 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3846 randomly, since we relied on an insufficiently-seeded PRNG.
3851 Description: OSS hangs after "All ost request buffers busy"
3852 Details : A deadlock between quota and journal operations caused OSS
3853 hangs after printing "All ost request buffers busy."
3856 Frequency : always on liblustre builds
3858 Description: Cleanup compiler warnings on liblustre
3861 Frequency : always on liblustre builds on XT3
3863 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3867 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3868 > 32000 subdirectories
3869 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3870 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3871 subdirectory count crosses 32000. This will aid e2fsck to
3872 correctly handle more than 32000 subdirectories.
3875 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3877 Description: System hang on startup
3878 Details : This bug allowed the liblustre (e.g. catamount) client to
3879 return to the app before handling all startup RPCs. This
3880 could leave the node unresponsive to lustre network traffic
3881 and manifested as a server ptllnd timeout.
3883 Severity : enhancement
3885 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3886 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3887 boolean to control whether to print peer debug info when a
3888 client's RPC times out.
3893 Description: make O_SYNC working on 2.6 kernels
3894 Details : 2.6 kernels use different method for mark pages for write,
3895 so need add a code to lustre for O_SYNC work.
3900 Description: Failure to close file and release space on NFS
3901 Details : Put inode details into lock acquired in ll_intent_file_open.
3902 Use mdc_intent_lock in ll_intent_open to properly
3903 detect all kind of errors unhandled by mdc_enqueue.
3908 Description: proc file read during shutdown sometimes raced obd removal,
3910 Details : Add lock to prevent obd access after proc file removal.
3913 Frequency : Only for files larger than 4GB on 32-bit clients.
3915 Description: improperly doing page alignment of locks
3916 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3917 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3922 Description: RPCs being resent when they shouldn't be
3923 Details : Some RPCs that should not be resent are being resent. This
3924 can cause inconsistencies in the RPC state machine. Do not
3925 resend such requests.
3928 Frequency : rare, only with NFS export
3930 Description: Crash on NFS re-export node
3931 Details : under very unusual load conditions an assertion is hit in
3932 ll_intent_file_open()
3935 Frequency : only if OST filesystem is corrupted
3937 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3938 Details : for a short time RPCs with bulk IO are in the replay list,
3939 but replay of bulk IOs is unimplemented. If the OST filesystem
3940 is corrupted due to disk cache incoherency and then replay is
3941 started it is possible to trip an assertion. Avoid putting
3942 committed RPCs into the replay list at all to avoid this issue.
3947 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3948 Details : Large single O_DIRECT read and write calls can fail to allocate
3949 a sufficiently large buffer to process the request. In case of
3950 allocation failure the allocation is retried with a smaller
3951 buffer and broken into smaller requests.
3953 Severity : enhancement
3955 Description: Add -o localflock option to simulate old noflock behaviour.
3956 Details : This will achieve local-only flock/fcntl locks coherentness.
3961 Description: versioning check is incomplete
3962 Details : Checking the version difference of client vs. server, report
3963 error if the gap is too big.
3968 Description: add support PG_writeback bit
3969 Details : add support for PG_writeback bit for Lustre, for more carefull
3970 work with page cache in 2.6 kernel. This also fix some deadlocks
3971 and remove hack for work O_SYNC with 2.6 kernel.
3973 Severity : enhancement
3975 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3976 Details : The uninit_groups feature works in conjunction with the kernel
3977 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3978 up the pass1 processing of e2fsck. This is a read-only feature
3979 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3980 mount filesystems that have had this feature enabled.
3982 Severity : enhancement
3984 Description: Improve multi-block allocation algorithm to avoid fragmentation
3985 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3986 allocation locality and avoid filesystem fragmentation.
3988 ------------------------------------------------------------------------------
3990 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3992 * Support for kernels:
3993 2.6.9-42.0.3.EL (RHEL 4)
3994 2.6.5-7.276 (SLES 9)
3995 2.4.21-47.0.1.EL (RHEL 3)
3996 2.6.12.6 vanilla (kernel.org)
3997 2.6.16.21-0.8 (SLES10)
3998 * Recommended e2fsprogs version: 1.39.cfs2-0
4000 * The backwards-compatible /proc/sys/portals symlink has been removed
4001 in this release. Before upgrading, please ensure that you change
4002 any configuration scripts or /etc/sysctl.conf files that access
4003 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4004 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4005 in advance of the upgrade on any system running Lustre 1.4.6 or
4006 newer, since /proc/sys/lnet was added in that version.
4007 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4011 Frequency : only when quota is used
4013 Description: avoid scanning export list for quota master
4014 Details : Change the algorithms to avoid scanning export list in order
4015 to improve the efficiency.
4018 Frequency : MDS failover only, very rarely
4020 Description: "went back in time" messages on mds failover
4021 Details : The greatest transno may be lost when the current operation
4022 finishes with an error (transno==0) and the client's last_rcvd
4023 record is over-written. Save the greatest transno in the
4024 mds_last_transno for this case.
4027 Frequency : always for specific kernels and striping counts
4029 Description: client may get "Matching packet too big" without ACL support
4030 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4031 when trying to access files in certain configurations. The
4032 clients should in fact be denied when mounting because they do
4033 not understand ACLs.
4036 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4038 Description: many clients connecting with IO in progress causes connect timeouts
4039 Details : Avoid synchronous journal commits to avoid delays caused by many
4040 clients connecting/disconnecting when bulk IO is in progress.
4041 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4042 OST_IO_PORTAL to avoid delays behind potentially many pending
4046 Frequency : occasionally with multiple writers to a single file
4048 Description: shared writes to file may result in wrong size reported by stat()
4049 Details : Allow growing of kms when extent lock is cancelled
4052 Frequency : always with random mmap IO to multi-striped file
4054 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4055 Details : Only if the hole is at the end of OST object so that kms is too
4056 small. Fix is to increase kms accordingly in ll_nopage.
4059 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4061 Description: writes to a missing object would leak memory on the OST
4062 Details : If there is an inconsistency between the MDS and OST filesystems,
4063 such that the MDS references an object that doesn't exist, writes
4064 to that object will leak memory due to incorrect cleanup in the
4065 error handling path, eventually running out of memory on the OST.
4070 Description: Creating too long symlink causes lustre errors
4071 Details : Check symlink and name lengths before sending requests to MDS.
4074 Frequency : only if flock is enabled (not on by default)
4076 Description: posix locks not released on fd closure on 2.6.9+
4077 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4078 kernels, this caused such locks not to be released on fd close and
4079 then assertions on fs unmount about still used locks.
4082 Frequency : MDS failover only, very rarely
4084 Description: clients may get ASSERTION(granted_lock != NULL)
4085 Details : When request was taking a long time, and a client was resending
4086 a getattr by name lock request. The were multiple lock requests
4087 with the same client lock handle and
4088 mds_getattr_name->fixup_handle_for_resent_request found one of the
4089 lock handles but later failed with ASSERTION(granted_lock != NULL).
4094 Description: handle->h_buffer_credits > 0, assertion failure
4095 Details : h_buffer_credits is zero after truncate, causing assertion
4096 failure. This patch extends the transaction or creates a new
4100 Frequency : NFS re-export or patchless client
4101 Bugzilla : 11179, 10796
4102 Description: Crash on NFS re-export node (__d_move)
4103 Details : We do not want to hash the dentry if we don't have a lock.
4104 But if this dentry is later used in d_move, we'd hit uninitialised
4105 list head d_hash, so we just do this to init d_hash field but
4106 leave dentry unhashed.
4109 Frequency : NFS re-export or patchless client
4111 Description: NFS exports has problem with symbolic link
4112 Details : lustre client didn't properly install dentry when re-exported
4113 to NFS or running patchless client.
4116 Frequency : NFS re-export or patchless client
4118 Description: Various nfs/patchless fixes.
4119 Details : fixes reuse disconected alias for lookup process - this fixes
4120 warning "find_exported_dentry: npd != pd",
4121 fix permission error with open files at nfs.
4122 fix apply umask when do revalidate.
4125 Frequency : occasional
4127 Description: Crash on NFS re-export node
4128 Details : calling clear_page() on the wrong pointer triggered oops in
4129 generic_mapping_read().
4132 Frequency : rarely, using O_DIRECT IO
4134 Description: unaligned directio crashes client with LASSERT
4135 Details : check for unaligned buffers before trying any requests.
4138 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4140 Description: stale data returned from RAID cache
4141 Details : If only a small amount of IO is done to the RAID device before
4142 reading it again it is possible to get stale data from the RAID
4143 cache instead of reading it from disk.
4146 Frequency : always for sles10 kernel
4148 Description: sles10 support
4149 Details : ll_follow_link: compile fixes and using of nd_set_link
4150 under newer kernels.
4153 Frequency : depends on arch, kernel and compiler version, always on sles10
4156 Description: recursive or deep enough symlinks cause stack overflow
4157 Details : getting rid of large stack-allocated variable in
4161 Frequency : depends on hardware
4163 Description: lustre write performance loss in the SLES10 kernel
4164 Details : the performance loss is caused by using of write barriers in the
4165 ext3 code. The SLES10 kernel turns barrier support on by
4166 default. The fix is to undo that change for ldiskfs.
4168 ------------------------------------------------------------------------------
4170 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4172 * Support for kernels:
4173 2.6.9-42.0.3EL (RHEL 4)
4174 2.6.5-7.276 (SLES 9)
4175 2.4.21-47.0.1.EL (RHEL 3)
4176 2.6.12.6 vanilla (kernel.org)
4180 Frequency : quota enabled and large files being deleted
4182 Description: releasing more than 4GB of quota at once hangs OST
4183 Details : If a user deletes more than 4GB of files on a single OST it
4184 will cause the OST to spin in an infinite loop. Release
4185 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4190 Description: statfs data retrieved from /proc may be stale or zero
4191 Details : When reading per-device statfs data from /proc, in the
4192 {kbytes,files}_{total,free,avail} files, it may appear
4193 as zero or be out of date.
4196 Frequency : systems with MD RAID1 external journal devices
4198 Description: lconf's call to blkid is confused by RAID1 journal devices
4199 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4200 of returning all block devices that match the journal UUID.
4203 Frequency : always, for aggregate stripe size over 4GB
4205 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4206 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4207 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4210 Frequency : always if "lfs find" used on a local file/directory
4212 Description: "lfs find" segfaults if used on a local file/directory
4213 Details : The case where a directory component was not specified wasn't
4214 handled correctly. Handle this properly.
4217 Frequency : always on ppc64
4219 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4220 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4221 perform endianness conversion.
4224 Frequency : rarely (truncate to non-zero file size after write under load)
4225 Bugzilla : 10730, 10687
4226 Description: Files padded with zeros to next 4K multiple
4227 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4228 it is possible that files that are truncated to a non-zero size
4229 immediately after being written are filled with zero bytes beyond
4230 the truncated size. No file data is lost.
4232 Severity : enhancement
4234 Description: Allow recovery/failover for liblustre clients.
4235 Details : liblustre clients were unaware of failover configurations until
4238 Severity : enhancement
4240 Description: user file locks should fail when not mounting with flock option
4241 Details : Set up an error-returning stub in ll_file_operations.lock field
4242 to prevent incorrect behaviour when client is mounted without
4243 flock option. Also, set up properly f_op->flock field for
4247 Frequency : always on ia64
4249 Description: "lfs df" loops on printing out MDS statfs information
4250 Details : The obd_ioctl_data was not initialized and in some systems
4251 this caused a failure during the ioctl that did not return
4252 an error. Initialize the struct and return an error on failure.
4255 Frequency : SLES 9 only
4257 Description: Error of copying files with lustre special EAs as root
4258 Details : Client side always return success for setxattr call for lustre
4259 special xattr (currently only "trusted.lov").
4262 Frequency : rarely on clusters with both ia64+i386 clients
4264 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4265 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4266 clients are doing concurrent writes to the same file it is
4267 possible that the ia64 clients may LASSERT because the OST
4268 extent locks are not PAGE_SIZE aligned. Ensure that grown
4269 locks are always aligned on the request boundary.
4272 Frequency : specific use, occasional
4274 Description: Overwriting in use executable truncates on-disk binary image
4275 Details : If one node attempts to overwrite an executable in use by
4276 another node, we now correctly return ETXTBSY instead of
4277 truncating the file.
4279 Severity : enhancement
4281 Description: Async OSC create to avoid the blocking unnecessarily.
4282 Details : If an OST has no remaining object, system will block on the creating
4283 when need to create a new object on this OST. Now, ways use
4284 pre-created objects when available, instead of blocking on an
4285 empty osc while others are not empty. If we must block, we block
4286 for the shortest possible period of time.
4291 Description: chmod on Lustre root is propagated to other clients
4292 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4293 invalid by the follow_mount time.
4296 Frequency : liblustre clients only
4298 Description: Race in 'instant cancel' lock handling could lead to such locks
4299 never to be granted in case of SMP MDS
4300 Details : Do not destroy not yet granted but cbpending locks in
4304 Frequency : replay/resend of open
4306 Description: non null lock assetion failure in mds_intent_policy
4307 Details : Trying to replay/resend lockless open requests resulted in
4308 mds_open() returning 0 with no lock. Now it sets a flag if
4309 a lock is going to be returned.
4311 Severity : enhancement
4313 Description: Checksum enhancements
4314 Details : New checksum enhancements allow for resending RPCs that failed
4317 Severity : enhancement
4319 Description: Tunables on number of dirty pages in cacche
4320 Details : Allow to set limit on number of dirty pages cached.
4325 Description: client crash on unmount - lock still has references
4326 Details : In some error handling cases it was possible to leak a lock
4327 reference on a client while accessing a file. This was not
4328 harmful to the client during operation, but would cause the
4329 client to crash when the filesystem is unmounted.
4332 Frequency : specific case, rare
4334 Description: ETXTBSY on mds though file not in use by client
4335 Details : ETXTBSY is no longer incorrectly returned when attempting to
4336 chmod or chown a directory that the user previously tried to
4337 execute or a currently-executing binary.
4340 Frequency : extremely rare except on liblustre-based clients
4342 Description: Lustre space not freed when files are deleted
4343 Details : Clean up open-unlinked files after client eviction. Previously
4344 the unlink was skipped and the files remained as orphans.
4349 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4350 Details : In some cases it was possible to send a blocking callback to
4351 a client doing a glimpse, even though that client didn't get
4352 a lock granted. When the glimpse lock is cancelled on the OST
4353 the freed lock is left on the waiting list and corrupted the list.
4356 Frequency : all core dumps
4358 Description: Broke core dumps to lustre
4359 Details : Negative dentry may be unhashed if parent does not have UPDATE
4360 lock, but some callers, e.g. do_coredump, expect dentry to be
4361 hashed after successful create, hash it in ll_create_it.
4363 ------------------------------------------------------------------------------
4365 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4367 * Support for kernels:
4368 2.6.9-42.0.2.EL (RHEL 4)
4369 2.6.5-7.276 (SLES 9)
4370 2.4.21-40.EL (RHEL 3)
4371 2.6.12.6 vanilla (kernel.org)
4375 Frequency : always on RHEL 3
4377 Description: Number of open files grows over time
4378 Details : The number of open files grows over time, whether or not
4379 Lustre is started. This was due to a filp leak introduced
4380 by one of our kernel patches.
4382 ------------------------------------------------------------------------------
4384 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4386 * Support for kernels:
4387 2.6.9-42.EL (RHEL 4)
4388 2.6.5-7.276 (SLES 9)
4389 2.4.21-40.EL (RHEL 3)
4390 2.6.12.6 vanilla (kernel.org)
4395 Bugzilla : 5719, 9635, 9792, 9684
4396 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4397 Details : If a server is under heavy load and cannot reply to new
4398 connection requests before the client resends the (re)connect,
4399 the connection handling code can behave badly if two service
4400 threads are concurrently handing separate (re)connections from
4401 the same client. Add better locking to the connection handling
4402 code, and ensure that only a single connection will be processed
4403 for a given client UUID, even if the lock is dropped.
4405 Severity : enhancement
4407 Description: add TCP zero-copy support to kernel
4408 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4409 sends if the hardware supports scatter-gather and checksumming.
4410 This allows socklnd to do client-write and server-read more
4411 efficiently and reduce CPU utilization from skbuf copying.
4414 Frequency : only if NFS exporting from client
4416 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4417 Details : If setting attributes on a file created under NFS that had
4418 never been opened it would be possible to oops the client
4419 if the file had no objects.
4422 Frequency : always for liblustre
4424 Description: liblustre client does MDS+OSTs setattr RPC for each write
4425 Details : When doing a write from a liblustre client, the client
4426 incorrectly issued an RPC to the MDS and each OST the file was
4427 striped over in order to update the timestamps. When writing
4428 with small chunks and many clients this could overwhelm the MDS
4429 with RPCs. In all cases it would slow down the write because
4430 these RPCs are unnecessary.
4432 Severity : enhancement
4434 Description: allow number of MDS service threads to be changed at module load
4435 Details : It is now possible to change the number of MDS service threads
4436 running. Adding "options mds mds_num_threads={N}" to the MDS's
4437 /etc/modprobe.conf will set the number of threads for the next
4438 time Lustre is restarted (assuming the "mds" module is also
4439 reloaded at that time). The default number of threads will
4440 stay the same, 32 for most systems.
4445 Description: OST crash if filesystem is unformatted or corrupt
4446 Details : If an OST is started on a device that has never been formatted
4447 or if the filesystem is corrupt and cannot even mount then the
4448 error handling cleanup routines would dereference a NULL pointer.
4453 Description: NULL pointer deref in llap_from_page.
4454 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4455 incorrect) mapping if the page was truncated/reclaimed while it was
4456 searched for. Check for this condition and skip such pages when
4457 doing readahead. Introduce extra check to llap_from_page() to
4458 verify page->mapping->host is non-NULL (so page is not anonymous).
4461 Frequency : Sometimes when using sys_sendfile
4463 Description: "page not covered by a lock" warnings from ll_readpage
4464 Details : sendfile called ll_readpage without right page locks present.
4465 Now we introduced ll_file_sendfile that does necessary locking
4466 around call to generic_file_sendfile() much like we do in
4470 Frequency : with certain MDS communication failures at client mount time
4472 Description: NULL pointer deref after failed client mount
4473 Details : a client connection request may delayed by the network layer
4474 and not be sent until after the PTLRPC layer has timed out the
4475 request. If the client fails the mount immediately it will try
4476 to clean up before the network times out the request. Add a
4477 reference from the request import to the obd device and delay
4478 the cleanup until the network drops the request.
4481 Frequency : occasionally during client (re)connect
4483 Description: assertion failure during client (re)connect
4484 Details : processing a client connection request may be delayed by the
4485 client or server longer than the client connect timeout. This
4486 causes the client to resend the connection request. If the
4487 original connection request is replied in this interval, the
4488 client may trip an assertion failure in ptlrpc_connect_interpret()
4489 which thought it would be the only running connect process.
4492 Frequency : only with obd_echo servers and clients that are rebooted
4494 Description: kernel BUG accessing uninitialized data structure
4495 Details : When running an obd_echo server it did not start the ping_evictor
4496 thread, and when a client was evicted an uninitialized data
4497 structure was accessed. Start the ping_evictor in the RPC
4498 service startup instead of the OBD startup.
4500 Severity : enhancement
4501 Bugzilla : 10193 (patchless)
4502 Description: Remove dependency on various unexported kernel interfaces.
4503 Details : No longer need reparent_to_init, exit_mm, exit_files,
4504 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4507 Frequency : rare (only users of deprecated and unsupported LDAP config)
4509 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4510 Details : LDAP apparently contains 'lustreName' attributes instead of
4511 'name'. A simple remapping of the name is sufficient.
4514 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4516 Description: waiting_locks_callback trips kernel BUG if client is evicted
4517 Details : Running with the dump_on_timeout debug flag turned on makes
4518 it possible that the waiting_locks_callback() can try to dump
4519 the Lustre kernel debug logs from an interrupt handler. Defer
4520 this log dumping to the expired_lock_main() thread.
4522 Severity : enhancement
4524 Description: Support NFS exporting on 2.6 kernels.
4525 Details : Implement non-rawops metadata methods for NFS server to use without
4526 changing NFS server code.
4529 Frequency : very rare (synthetic metadata workload only)
4531 Description: two racing renames might cause an MDS thread to deadlock
4532 Details : Running the "racer" program may cause one MDS thread to rename
4533 a file from being the source of a rename to being the target of
4534 a rename at exactly the same time that another thread is doing
4535 so, and the second thread has already enqueued these locks after
4536 doing a lookup of the target and is trying to relock them in
4537 order. Ensure that we don't try to re-lock the same resource.
4540 Frequency : only very large systems with liblustre clients
4542 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4543 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4544 all OSTs in parallel. This allows the eviction of stale liblustre
4545 clients to proceed much faster than if they were done in series,
4546 and also offers similar improvements for other set_info RPCs.
4551 Description: excessive CPU usage during initial read phase on client
4552 Details : During the initial read phase on a client, it would agressively
4553 retry readahead on the file, consuming too much CPU and impacting
4554 performance (since 1.4.5.8). Improve the readahead algorithm
4555 to avoid this, and also improve some other common cases (read
4556 of small files in particular, where "small" is files smaller than
4557 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4562 Description: MDS crash when receiving packet with unknown intent.
4563 Details : Do not LBUG in unknown intent case, just return -EFAULT
4565 Severity : enhancement
4566 Bugzilla : 9293, 9385
4567 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4568 Details : Do not serialize getattr (non-intent version) and statfs.
4571 Frequency : occasional, when OST network is overloaded/intermittent
4573 Description: client evicted by OST after bulk IO timeout
4574 Details : If a client sends a bulk IO request (read or write) the OST
4575 may evict the client if it is unresposive to its data GET/PUT
4576 request. This is incorrect if the network is overloaded (takes
4577 too long to transfer the RPC data) or dropped the OST GET/PUT
4578 request. There is no need to evict the client at all, since
4579 the pinger and/or lock callbacks will handle this, and the
4580 client can restart the bulk request.
4583 Frequency : Always when mmapping file with no objects
4585 Description: client crashes when mmapping file with no objects
4586 Details : Check that we actually have objects in a file before doing any
4587 operations on objects in ll_vm_open, ll_vm_close and
4593 Description: Request leak when working with deleted CWD
4594 Details : Introduce advanced request refcount tracking for requests
4595 referenced from lustre intent.
4597 Severity : Enhancement
4599 Description: Cache open file handles on client.
4600 Details : MDS now will return special lock along with openhandle, if
4601 requested and client is allowed to hold openhandle, even if unused,
4602 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4603 closing files for every read/write openration.
4605 Severity : Enhancement
4607 Description: Cache open negative dentries on client when possible.
4608 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4609 negative dentries on lock revocation.
4614 Description: Remounting a client read-only wasn't possible with a zconf mount
4615 Details : It wasn't possible to remount a client read-only with llmount.
4617 Severity : enhancement
4618 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4619 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4620 MPICH is included to add Lustre-specific ADIO interfaces.
4621 This is based closely on the UFS ADIO layer and only differs
4622 in file creation, in order to allow the OST striping to be set.
4623 This is user-contributed code and not supported by CFS.
4628 Description: extended inode attributes (immutable, append-only) work improperly
4629 when 2.4 and 2.6 kernels are used on client/server or vice versa
4630 Details : Introduce kernel-independent values for these flags.
4632 Severity : enhancement
4635 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4636 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4637 granularity. Now, Fractional MB granularity is be supported,
4638 this is very useful for low memory system.
4640 Severity : enhancement
4642 Description: Getattr by fid
4643 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4644 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4645 have locks and avoids getattr rpc requests for it on every stat.
4648 Frequency : Always, for filesystems larger than 2TB
4650 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4651 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4652 filesystems larger than 2TB to be mounted with mballoc enabled.
4655 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4657 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4658 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4659 filesystems larger than 2TB is mounted then the kernel will
4660 silently corrupt the start of the filesystem. CONFIG_LBD is
4661 enabled for all CFS-supported kernels, but the possibility of
4662 this happening with a modified kernel config exists.
4664 Severity : enhancement
4666 Description: add client O_DIRECT support for 2.6 kernels
4667 Details : It is now possible to do O_DIRECT reads and writes to files
4668 in the Lustre client mountpoint on 2.6 kernel clients.
4670 Severity : enhancement
4672 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4673 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4674 parallel, not waiting for response from every OST before sending
4675 a rpc to the next OST.
4680 Description: setattr vs write race when updating file timestamps
4681 Details : Client processes that update a file timestamp into the past
4682 right after writing to the file (e.g. tar) it is possible that
4683 the updated file modification time can be reset to the current
4684 time due to a race between processing the setattr and write RPC.
4686 Severity : enhancement
4688 Description: Bring 'lfs find' closer in line with regular Linux find.
4689 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4690 -print0 options and obtains all the needed info through the lustre
4693 Severity : enhancement
4695 Description: support up to 1024 configured devices on one node
4696 Details : change obd_dev array from statically allocated to dynamically
4697 allocated structs as they are first used to reduce memory usage
4702 Description: Flush dirty partially truncated pages during truncate
4703 Details : Immediatelly flush partially truncated pages in filter_setattr,
4704 this way we completely avoid having any pages in page cache on OST
4705 and can retire ugly workarounds during writes to flush such pages.
4710 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4711 Details : protect inode from truncation within vfs_unlink() context
4712 just take a reference before calling vfs_unlink() and release it
4713 when parent's i_sem is free.
4718 Description: last_id value checked outside lock on OST caused LASSERT failure
4719 Details : If there were multiple MDS->OST object precreate requests in
4720 flight, it was possible that the OST's last object id was checked
4721 outside a lock and incorrectly tripped an assertion. Move checks
4722 inside locks, and discard old precreate requests.
4725 Frequency : always, if extents are used on OSTs
4727 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4728 Details : OSTs using the extents format would not zero the high 16 bits of
4729 the index physical block number. This is not a problem for any
4730 OST filesystems smaller than 16TB, and no kernels support ext3
4731 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4732 new/modified files) and can be fixed for existing filesystems
4733 with e2fsprogs-1.39-cfs1.
4738 Description: import connection selection may be incorrect if timer wraps
4739 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4740 import connections to be ignored if the 32-bit jiffies counter
4741 wraps. Use a 64-bit jiffies counter.
4744 Frequency : very large clusters immediately after boot
4746 Description: LNET request buffers exhausted under heavy short-term load
4747 Details : If a large number of client requests are generated on a service
4748 that has previously never seen so many requests it is possible
4749 that the request buffer growth cannot keep up with the spike in
4750 demand. Instead of dropping incoming requests, they are held in
4751 the LND until the RPC service can accept more requests.
4754 Frequency : Sometimes during replay
4756 Description: Assertion failure in ll_local_open after replay.
4757 Details : If replay happened on an open request reply before we were able
4758 to set replay handler, reply will become not swabbed tripping the
4759 assertion in ll_local_open. Now we set the handler right after
4760 recognising of open request
4763 Frequency : very rare
4765 Description: kernel reports "badness in vsnprintf"
4766 Details : Reading from the "recovery_status" /proc file in small chunks
4767 may cause a negative length in lprocfs_obd_rd_recovery_status()
4768 call to vsnprintf() (which is otherwise harmless). Exit early
4769 if there is no more space in the output buffer.
4771 Severity : enhancement
4773 Description: clear OBD RPC statistics by writing to them
4774 Details : It is now possible to clear the OBD RPC statistics by writing
4775 to the "stats" file.
4780 Description: Client mtime is not the same on different clients after utimes
4781 Details : In some cases, the client was using the utimes() syscall on
4782 a file cached on another node. The clients now validate the
4783 ctime from the MDS + OSTs to determine which one is right.
4788 Description: Inability to activate failout mode
4789 Details : lconf script incorrectly assumed that in pythong string's numeric
4790 value is used in comparisons.
4793 Frequency : always with multiple stripes per file
4795 Description: Inefficient object allocation for mutli-stripe files
4796 Details : When selecting which OSTs to stripe files over, for files with
4797 a stripe count that divides evenly into the number of OSTs,
4798 the MDS is always picking the same starting OST for each file.
4799 Return the OST selection heuristic to the original design.
4804 Description: mount failures may take full timeout to return an error
4805 Details : Under some heavy load conditions it is possible that a
4806 failed mount can wait for the full obd_timeout interval,
4807 possibly several minutes, before reporting an error.
4808 Instead return an error as soon as the status is known.
4810 Frequency : quota enabled and large files being deleted
4812 Description: releasing more than 4GB of quota at once hangs OST
4813 Details : If a user deletes more than 4GB of files on a single OST it
4814 will cause the OST to spin in an infinite loop. Release
4815 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4820 Description: statfs data retrieved from /proc may be stale or zero
4821 Details : When reading per-device statfs data from /proc, in the
4822 {kbytes,files}_{total,free,avail} files, it may appear
4823 as zero or be out of date.
4826 Frequency : systems with MD RAID1 external journal devices
4828 Description: lconf's call to blkid is confused by RAID1 journal devices
4829 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4830 of returning all block devices that match the journal UUID.
4833 Frequency : always, for aggregate stripe size over 4GB
4835 Description: assertion fails when trying to use 4GB stripe size
4836 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4837 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4840 Frequency : always on ppc64
4842 Description: the first write on an ext3 filesystem with mballoc got stuck
4843 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4844 perform endianness conversion.
4846 ------------------------------------------------------------------------------
4848 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4850 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4851 INTEROPERATE with older versions automatically. Please read the
4852 user documentation before upgrading any part of a live system.
4853 * WARNING: Lustre networking configuration changes are required with
4854 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4857 * Support for kernels:
4858 2.6.9-22.0.2.EL (RHEL 4)
4859 2.6.5-7.244 (SLES 9)
4860 2.6.12.6 vanilla (kernel.org)
4863 Severity : enhancement
4864 Bugzilla : 7981/8208
4865 Description: Introduced Lustre Networking (LNET)
4866 Details : LNET is new networking infrastructure for Lustre, it includes
4867 a reorganized network configuration mode (see the user
4868 documentation for full details) as well as support for routing
4869 between different network fabrics. Lustre Networking Devices
4870 (LNDs) for the supported network fabrics have also been
4871 created for this new infrastructure.
4873 Severity : enhancement
4874 Description: Introduced Access control lists
4875 Details : clients can set ACLs on files and directories in order to have
4876 more fine-grained permissions than the standard Unix UGO+RWX.
4877 The MDS must be started with the "-o acl" mount option.
4879 Severity : enhancement
4880 Description: Introduced filesystem quotas
4881 Details : Administrators may now establish per-user quotas on the
4884 Severity : enhancement
4886 Description: Configuration change for the XT3
4887 The PTLLND is now used to run Lustre over Portals on the XT3
4888 The configure option(s) --with-cray-portals are no longer used.
4889 Rather --with-portals=<path-to-portals-includes> is used to
4890 enable building on the XT3. In addition to enable XT3 specific
4891 features the option --enable-cray-xt3 must be used.
4896 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4897 Details : A race between allocating a new llcd and re-getting the llcd_lock
4898 allowed another thread to grab newly-allocated llcd.
4900 Severity : enhancement
4902 Description: 2.6 OST async journal commit and locking fix to improve performance
4903 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4904 2.6 kernels are now async as they already were in 2.4 kernels so
4905 that they can commit concurrently with the network bulk transfer.
4906 For block-allocated files the filter allocation semaphore is held
4907 to avoid filesystem fragmentation during allocation. BKL lock
4908 removed for 2.6 xattr operations where it is no longer needed.
4913 Description: lconf incorrectly determined whether two IP networks could talk
4914 Details : In some more complicated routing and multiple-network
4915 configurations, lconf will avoid trying to make a network
4916 connection to a disjoint part of the IP space. It was doing the
4917 math incorrectly for one set of cases.
4922 Description: Fix for potential infinite loop processing records in an llog.
4923 Details : If an llog record is corrupted/zeroed, it is possible to loop
4924 forever in llog_process(). Validate the llog record length
4925 and skip the remainder of the block on error.
4928 Frequency : occasional (liblustre only)
4930 Description: liblustre could not open files whose last component is a symlink
4931 Details : sysio_path_walk() would incorrectly pass the open intent to
4932 intermediate path components.
4935 Frequency : rare (liblustre only with non-standard tuning)
4936 Bugzilla : 7201 (7350)
4937 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4938 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4939 after it has been released, only do this for referenced locks
4941 Severity : enhancement
4943 Description: specifying an (invalid) directory default stripe_size of -1
4944 would reset the directory default striping
4945 Details : stripe_size -1 was used internally to signal directory stripe
4946 removal, now use "all default" to signal dir stripe removal
4947 as a directory striping of "all default" is not useful
4950 Frequency : common for large clusters running liblustre clients
4952 Description: doing an ls when liblustre clients are running is slow
4953 Details : sending a glimpse AST to a liblustre client waits for every AST
4954 to time out, as liblustre clients will not respond. Since they
4955 cannot cache data we refresh the OST lock LVB from disk instead.
4957 Severity : enhancement
4959 Description: doing an ls at the same time as file IO can be slow
4960 Details : enqueue and other "small" requests can be blocked behind many
4961 large IO requests. Create a new OST IO portal for non-IO
4962 requests so they can be processed faster.
4965 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4967 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4968 created file with mode 000
4969 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4970 allow an ftruncate() as a non-root user to a file with mode 000.
4971 Lustre now respects this flag to disable mode checks when
4972 truncating a file owned by the user
4975 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4978 Description: Revoking locks from clients that went dead or catatonic might take
4980 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4981 cancellation of such locks instant on servers without waiting for
4982 any reply from clients. Clients drops these locks when cancel
4983 notification from server is received without replying.
4986 Frequency : liblustre-only, when liblustre client dies or becomes busy
4988 Description: Doing ls on Linux clients can take a long time with active
4990 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4991 granting such locks to it in the first place if possible. Locks
4992 are taken by proxy on the OST during the read or write and
4993 dropped immediately afterward. Add connect flags handling, do
4994 not grant locks to liblustre clients for glimpse ASTs.
4996 Severity : enhancement
4998 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4999 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5000 sequential reads and behaves badly with applications doing random
5001 reads. Improve it by reading ahead at least read region, and
5002 avoiding excessive large RPC for small reads.
5004 Severity : enhancement
5006 Description: Creating more than 1000 files for a single job may cause a load
5007 imbalance on the OSTs if there are also a large number of OSTs.
5008 Details : qos_prep_create() uses an OST index reseed value that is an
5009 even multiple of the number of available OSTs so that if the
5010 reseed happens in the middle of the object allocation it will
5011 still utilize the OSTs as uniformly as possible.
5016 Description: OST or MDS may oops in ping_evictor_main()
5017 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5018 but doesn't restart at beginning of obd_exports_timed list
5021 Severity : enhancement
5023 Description: improve by-nid export eviction on the MDS and OST
5024 Details : allow multiple exports with the same NID to be evicted at one
5025 time without re-searching the exports list.
5028 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5030 Description: MDS may oops in groups_free()
5031 Details : in rare race conditions a newly allocated group_info struct is
5032 freed again, and this can be NULL. The 2.4 compatibility code
5033 for groups_free() checked for a NULL pointer, but 2.6 did not.
5036 Frequency : common for liblustre clients doing little filesystem IO
5037 Bugzilla : 9352, 7313
5038 Description: server may evict liblustre clients accessing contended locks
5039 Details : if a client is granted a lock or receives a completion AST
5040 with a blocking AST already set it would not reply to the AST
5041 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5044 Frequency : lfs setstripe, only systems with more than 160 OSTs
5046 Description: unable to set striping with a starting offset beyond OST 160
5047 Details : llapi_create_file() incorrectly limited the starting stripe
5048 index to the maximum single-file stripe count.
5051 Frequency : LDAP users only
5053 Description: lconf did not handle in-kernel recovery with LDAP properly
5054 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5056 Severity : enhancement
5058 Description: bind OST threads to NUMA nodes to improve performance
5059 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5060 node and do their allocations there to localize memory access
5062 Severity : enhancement
5064 Description: llmount can determine client NID directly from Myrinet (GM)
5065 Details : the client NID code from gmnalnid was moved directly into
5066 llmount, removing the need to use this or specifying the
5067 client NID explicitly when mounting GM clients with zeroconf
5070 Frequency : if client is started with down MDS
5072 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5073 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5074 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5075 ptlrpc_import_delayed_req() because we didn't check if the
5076 request was interrupted, and we also didn't break out of the
5077 event loop if there was no timeout
5082 Description: data loss during non-page-aligned writes to a single file from
5083 both multiple nodes and multiple threads on one node at same time
5084 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5085 inconsistency led to false short-reads, that were cached and later
5086 used by ->prepare_write() to fill in partially written page,
5087 leading to data loss.
5090 Frequency : always, if lconf --abort_recovery used
5092 Description: lconf --abort_recovery fails with 'Operation not supported'
5093 Details : lconf was attempting to abort recovery on the MDT device and not
5096 Severity : enhancement
5098 Description: remove cleanup logs
5099 Details : replace lconf-generated cleanup logs with lustre internal
5100 cleanup routines. Eliminates the need for client-cleanup and
5103 Severity : enhancement
5105 Description: add support for EAs (user and system) on lustre filesystems
5106 Details : it is now possible to store extended attributes in the Lustre
5107 client filesystem, and with the user_xattr mount option it
5108 is possible to allow users to store EAs on their files also
5110 Severity : enhancement
5112 Description: Add possibility (config option) to show minimal available OST free
5114 Details : When compiled with --enable-mindf configure option, statfs(2)
5115 (and so, df) will return least minimal free space available from
5116 all OSTs as amount of free space on FS, instead of summary of
5117 free spaces of all OSTs.
5119 Severity : enhancement
5121 Description: do not expand extent locks acquired on OST-side
5122 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5123 by server: they are not cached anyway.
5126 Frequency : when mmap is used/binaries executed from Lustre
5128 Description: Unmmap pages before throwing them away from read cache.
5129 Details : llap_shrink cache now attempts to unmap pages before discarding
5130 them (if unmapping failed - do not discard). SLES9 kernel has
5131 extra checks that trigger if this unmapping is not done first.
5136 Description: lconf didn't resolve symlinks before checking to see whether a
5137 given mountpoint was already in use
5140 Frequency : when migrating failover services
5141 Bugzilla : 6395, 9514
5142 Description: When migrating a subset of services from a node (e.g. failback
5143 from a failover service node) the remaining services would
5144 time out and evict clients.
5145 Details : lconf --force (implied by --failover) sets the global obd_timeout
5146 to 5 seconds in order to quickly disconnect, but this caused
5147 other RPCs to time out too quickly. Do not change the global
5148 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5150 Severity : enhancement
5151 Frequency : if MDS is started with down OST
5152 Bugzilla : 9439,5706
5153 Description: Allow startup/shutdown of an MDS without depending on the
5154 availability of the OSTs.
5155 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5156 Add appropriate locking and lov-osc refcounts for safe
5157 cleaning. Add osc abort_inflight calls in case the
5158 synchronize never started.
5161 Frequency : occasional (Cray XT3 only)
5163 Description: root not authorized to access files in CRAY_PORTALS environment
5164 Details : The client process capabilities were not honoured on the MDS in
5165 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5166 been accessed by an authorized user then root was able to access
5167 the file on the local client also. The root user capabilities
5168 are now allowed on the MDS, as this environment has secure UID.
5171 Frequency : occasional
5173 Description: ldiskfs "too long searching" message happens too often
5174 Details : A debugging message (otherwise harmless) prints too often on
5175 the OST console. This has been reduced to only happen when
5176 there are fragmentation problems on the filesystem.
5181 Description: Division by zero in statfs when all OSCs are inactive
5182 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5183 lov_statfs divides by value returned by lov_get_stripecnt().
5187 Bugzilla : 9489, 3273
5188 Description: First write from each client to each OST was only 4kB in size,
5189 to initialize client writeback cache, which caused sub-optimal
5190 RPCs and poor layout on disk for the first written file.
5191 Details : Clients now request an initial cache grant at (re)connect time
5192 and so that they can start streaming writes to the cache right
5193 away and always do full-sized RPCs if there is enough data.
5194 If the OST is rebooted the client also re-establishes its grant
5195 so that client cached writes will be honoured under the grant.
5200 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5201 Details : Now I/O RPCs go to different portal number and (presumably) fast
5202 lock requests (and glimses) and other RPCs get their own service
5203 threads pool that should be able to service those RPCs
5206 Severity : enhancement
5208 Description: Ability to exchange lustre version between client and servers and
5209 issue warnings at client side if client is too old. Also for
5210 liblustre clients there is ability to refuse connection of too old
5212 Details : New 'version' field is added to connect data structure that is
5213 filled with version info. That info is later checked by server and
5217 Frequency : rare, liblustre only.
5218 Bugzilla : 9296, 9581
5219 Description: Two simultaneous writes from liblustre at offset within same page
5220 might proceed at the same time overwriting eachother with stale
5222 Details : I/O lock withing llu_file_prwv was released too early, before data
5223 actually was hitting the wire. Extended lock-holding time until
5224 server acknowledges receiving data.
5227 Frequency : extremely rare. Never observed in practice.
5229 Description: avoid generating lustre_handle cookie of 0.
5230 Details : class_handle_hash() generates handle cookies by incrementing
5231 global counter, and can hit 0 occasionaly (this is unlikely, but
5232 not impossible, because initial value of cookie counter is
5233 selected randonly). Value of 0 is used as a sentinel meaning
5234 "unassigned handle" --- avoid it. Also coalesce two critical
5235 sections in this function into one.
5237 Severity : enhancement
5239 Description: allow liblustre clients to delegate truncate locking to OST
5240 Details : To avoid overhead of locking, liblustre client instructs OST to
5241 take extent lock in ost_punch() on client's behalf. New connection
5242 flag is added to handle backward compatibility.
5244 Severity : enhancement
5245 Bugzilla : 4928, 7341, 9758
5246 Description: allow number of OST service threads to be specified
5247 Details : a module parameter allows the number of OST service threads
5248 to be specified via "options ost ost_num_threads={N}" in the
5249 OSS's /etc/modules.conf or /etc/modprobe.conf.
5253 Bugzilla : 6146, 9635, 9895
5254 Description: servers crash with bad pointer in target_handle_connect()
5255 Details : In rare cases when a client is reconnecting it was possible that
5256 the connection request was the last reference for that export.
5257 We would temporarily drop the export reference and get a new
5258 one, but this may have been the last reference and the export
5259 was just destroyed. Get new reference before dropping old one.
5261 Severity : enhancement
5262 Frequency : if client is started with failover MDS
5264 Description: Allow multiple MDS hostnames in the mount command
5265 Details : Try to read the configuration from all specified MDS
5266 hostnames during a client mount in case the "primary"
5269 Severity : enhancement
5271 Description: Stop sending data to evicted clients as soon as possible.
5272 Details : Check if the client we are about to send or are sending data to
5273 was evicted already. (Check is done every second of waiting,
5274 for which l_wait_event interface was extended to allow checking
5275 of exit condition at specified intervals).
5278 Frequency : rare, normally only when NFS exporting is done from client
5280 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5282 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5285 Frequency : always, if srand() is called before liblustre initialization
5287 Description: Liblustre uses system PRNG disturbing its usage by user application
5288 Details : Introduce internal to lustre fast and high-quality PRNG for
5289 lustre usage and make liblustre and some other places in generic
5290 lustre code to use it.
5292 Severity : enhancement
5293 Bugzilla : 9477, 9557, 9870
5294 Description: Verify that the MDS configuration logs are updated when xml is
5295 Details : Check if the .xml configuration logs are newer than the config
5296 logs stored on the MDS and report an error if this is the case.
5297 Request --write-conf, or allow starting with --old_conf.
5299 Severity : enhancement
5301 Description: Handle symlinks in the path when checking if Lustre is mounted.
5302 Details : Resolve intermediate symlinks when checking if a client has
5303 mounted a filesystem to avoid duplicate client mounts.
5308 Description: lconf can hit an error exception but still return success.
5309 Details : The lconf command catches the Command error exception at the top
5310 level script context and will exit with the associated exit
5311 status, but doesn't ensure that this exit status is non-zero.
5316 Description: failure of ptlrpc thread startup can cause oops
5317 Details : Starting a ptlrpc service thread can fail if there are a large
5318 number of threads or the server memory is very fragmented.
5319 Handle this without oopsing.
5322 Frequency : always, only if liblustre and non-default acceptor port was used
5324 Description: liblustre cannot connect to servers with non-default acceptor port
5325 Details : tcpnal_set_default_params() was not called and was therefore
5326 ignoring the environment varaible TCPNAL_PORT, as well as other
5327 TCPNAL_ environment variables
5332 Description: two objects could be created on the same OST for a single file
5333 Details : If an OST is down, in some cases it was possible to create two
5334 objects on a single OST for a single file. No problems other
5335 than potential performance impact and spurious error messages.
5339 Bugzilla : 5681, 9562
5340 Description: Client may oops in ll_unhash_aliases
5341 Details : Client dcache may become inconsistent in race condition.
5342 In some cases "getcwd" can fail if the current directory is
5348 Description: Inode refcounting problems in NFS export code
5349 Details : link_raw functions used to call d_instantiate without obtaining
5350 extra inode reference first.
5354 Bugzilla : 9942, 9903
5355 Description: Referencing freed requests leading to crash, memleaks with NFS.
5356 Details : We used to require that call to ll_revalidate_it was always
5357 followed by ll_lookup_it. Also with revalidate_special() it is
5358 possible to call ll_revalidate_it() twice for the same dentry
5359 even if first occurence returned success. This fix changes semantic
5360 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5361 reference on a request referred from the intent.
5362 ll_intent_release() then releases such a request.
5365 Frequency : rare, normally benchmark loads only
5367 Description: unlinked inodes were kept in memory on the client
5368 Details : If a client is repeatedly creating and unlinking files it
5369 can accumulate a lot of stale inodes in the inode slab cache.
5370 If there is no other client load running this can cause the
5371 client node to run out of memory. Instead flush old inodes
5372 from client cache that have the same inode number as a new inode.
5375 Frequency : SLES9 2.6.5 kernel and long filenames only
5376 Bugzilla : 9969, 10379
5377 Description: utime reports stale NFS file handle
5378 Details : SLES9 uses out-of-dentry names in some cases, which confused
5379 the lustre dentry revalidation. Change it to always use the
5383 Frequency : rare, unless heavy write-truncate concurrency is continuous
5384 Bugzilla : 4180, 6984, 7171, 9963, 9331
5385 Description: OST becomes very slow and/or deadlocked during object unlink
5386 Details : filter_destroy() was holding onto the parent directory lock
5387 while truncating+unlinking objects. For very large objects this
5388 may block other threads for a long time and slow overall OST
5389 responsiveness. It may also be possible to get a lock ordering
5390 deadlock in this case, or run out of journal credits because of
5391 the combined truncate+unlink. Solution is to do object truncate
5392 first in one transaction without parent lock, and then do the
5393 final unlink in a new transaction with the parent lock. This
5394 reduces the lock hold time dramatically.
5397 Frequency : rare, 2.4 kernels only
5399 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5400 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5401 before unmounting their filesystem in order to avoid deadlock.
5402 The kernel_locked() function in 2.4 kernels only checks whether
5403 the kernel lock is held, not whether it is this process that is
5404 holding it as 2.6 kernels do.
5409 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5410 Details : The client ptlrpc code may be trying to reconnect to a down
5411 server before a previous connection attempt has timed out.
5412 Increase the reconnect interval to be longer than the connection
5413 timeout interval to avoid sending duplicate connections to
5417 Frequency : echo_client brw_test command
5419 Description: fix echo_client to work with OST preallocated code
5420 Details : OST preallocation code (5137) didn't take echo_client IO path
5421 into account: echo_client calls filter methods outside of any
5422 OST thread and, hence, there is no per-thread preallocated
5423 pages and buffers to use. Solution: hijack pga pages for IO. As
5424 a byproduct, this avoids unnecessary data copying.
5428 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5429 Description: Client can oops in mdc_commit_close() after open replay
5430 Details : It was possible for the MDS to return an open request with no
5431 transaction number in mds_finish_transno() if the client was
5432 evicted, but without actually returning an error. Clients
5433 would later try to replay that open and may trip an assertion
5434 Simplify the client close codepath, and always return an error
5435 from the MDS in case the open is not successful.
5438 Frequency : rare, 2.6 OSTs only
5440 Description: OST may deadlock under high load on fragmented files
5441 Details : If there was a heavy load and highly-fragmented OST filesystems
5442 it was possible to have all the OST threads deadlock waiting on
5443 allocation of biovecs, because the biovecs were not released
5444 until the entire RPC IO was completed. Instead, release biovecs
5445 as soon as they are complete to ensure forward IO progress.
5447 Severity : enhancement
5449 Description: Support for specifying external journal device at mount
5450 Details : If an OST or MDS device is formatted with an external journal
5451 device, this device major/minor is stored in the ext3 superblock
5452 and may not be valid for failover. Allow detecting and
5453 specifying the external journal at mount time.
5458 Description: Mounting an MDS with pending unlinked files may cause oops
5459 Details : target_finish_recovery() calls mds_postrecov() which returned
5460 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5461 considers this an error and immediately begins cleaning up the
5462 lov, just after starting the mds_lov process
5464 Severity : enhancement
5466 Description: Implement 'lfs df' to report actual free space on per-OST basis
5467 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5468 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5469 usage of objects; '-h' to report in human readable format.
5471 ------------------------------------------------------------------------------
5473 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5480 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5481 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5482 at startup during orphan unlinking. Instead, skip buddy update
5483 before setup, as it will be regenerated after recovery is complete.
5488 Description: If an OST is inactive, its locks might reference stale inodes.
5489 Details : lov_change_cbdata() must iterate over all namespaces, even if
5490 they are inactive to clear inode references from the lock.
5492 Severity : enhancement
5493 Frequency : occasional, if non-standard max_dirty_mb used
5495 Description: Client will block write RPCs if not enough grant
5496 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5497 then the client will block writes while waiting for another RPC
5498 to complete instead of consuming its dirty limit. With change
5499 we get improved performance when max_dirty_mb is small.
5501 Severity : enhancement
5502 Bugzilla : 3389, 6253
5503 Description: Add support for supplementary groups on the MDS.
5504 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5505 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5506 lookups for user supplementary groups into a cache.
5510 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5511 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5512 for liblustre O_EXCL is enforced by MDS.
5516 Description: readdir never returns NULL in liblustre.
5517 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5518 dentry in struct dirent.
5522 Frequency : liblustre only, depends on application IO pattern
5523 Description: liblustre clients evicted if not contacting servers
5524 Details : Don't put liblustre clients into the ping_evictor list, so
5525 they will not be evicted by the pinger ever.
5527 Severity : enhancement
5529 Description: Add ability to evict clients by NID from MDS.
5530 Details : By echoing "nid:$NID" string into
5531 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5532 $NID would be instantly evicted from this MDS and from all active
5533 OSTs connected to it.
5537 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5538 Details : lookup_it_finish() used to query file size from OSTs that was not
5543 Description: service threads change working directory to that of init
5544 Details : Starting lustre service threads may pin the working directory
5545 of the parent thread, making that filesystem busy. Threads
5546 now change to the working directory of init to avoid this.
5550 Frequency : during shutdown only
5551 Description: shutdown with a failed MDS or OST can cause unmount to hang
5552 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5557 Frequency : 2.6 only
5558 Description: chmod/chown may include an extra supplementary group
5559 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5560 supplementary group and if none is specified this is used.
5563 Bugzilla : 5479 (6816)
5565 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5566 Details : If lookup is in progress on a file that is unlinked we might try
5567 to revalidate the inode and fail in revalidate after lookup is
5568 complete and ll_file_open() enqueues the open again but
5569 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5572 Frequency : always, if lconf --abort_recovery used
5574 Description: lconf --abort_recovery fails with 'Operation not supported'
5575 Details : lconf was attempting to abort recovery on the MDT device and not
5578 ------------------------------------------------------------------------------
5580 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5585 Frequency : rare (only unsupported configurations with a node running as an
5587 Bugzilla : 6514, 5137
5588 Description: Mounting a Lustre file system on a node running as an OST could
5590 Details : OSTs now preallocates memory needed to write out data at
5591 startup, instead of when needed, to avoid having to
5592 allocate memory in possibly low memory situations.
5593 Specifically, if the file system is mounted on on OST,
5594 memory pressure could force it to try to write out data,
5595 which it needed to allocate memory to do. Due to the low
5596 memory, it would be unable to do so and the node would
5597 become unresponsive.
5599 Severity : enhancement
5601 Description: Addition of lconf --service command line option
5602 Details : lconf now accepts a '--service <arg>' option, which is
5603 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5605 Severity : enhancement
5607 Description: Failover mode is now the default for OSTs.
5608 Details : By default, OSTs will now run in failover mode. To return to
5609 the old behaviour, add '--failout' to the lmc line for OSTs.
5611 Severity : enhancement
5613 Description: Health checks are now provided for MDS and OSTs
5614 Details : Additional detailed health check information on MSD and OSTs
5615 is now provided through the procfs health_check value.
5618 Frequency : occasional, depends on IO load
5620 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5621 after numerous create/delete cycles
5622 Details : The ext3 inode allocation policy would not allocate new inodes
5623 very well on the OSTs because there are no new directories
5624 being created. Instead we look for groups with free space if
5625 the parent directories are nearly full.
5630 Description: Network or server problems during mount may cause partially
5631 mounted clients instead of returning an error.
5632 Details : The config llog parsing code may overwrite the error return
5633 code during mount error handling, returning success instead
5639 Description: MDS can fail to allocate large reply buffers
5640 Details : After long uptimes the MDS can fail to allocate large reply
5641 buffers (e.g. zconf client mount config records) due to memory
5642 fragmentation or consumption by the buffer cache. Preallocate
5643 some large reply buffers so that these replies can be sent even
5644 under memory pressure.
5648 Frequency : rare (liblustre)
5649 Description: fsx running with liblustre complained that using truncate() to
5650 extend the file doesn't work. This patch corrects that issue.
5651 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5652 ATTR_SIZE and lsm use in llu_setattr_raw.
5656 Frequency : rare, only 2.6 kernels
5657 Description: Unusual file access patterns on the MDS may result in inode
5658 data being lost in very rare circumstances.
5659 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5660 code in the 2.6 kernel caused the inode and/or EA data not to
5661 be read from disk, causing single-file corruption.
5665 Frequency : rare, only 2.6 filesystems using extents
5666 Description: Heavy concurrent write and delete load may cause data corruption.
5667 Details : It was possible under high-load situations to have an extent
5668 metadata block in the block device cache from a just-unlinked
5669 file overwrite a newly-allocated data block. We now unmap any
5670 metadata buffers that alias just-allocated data blocks.
5674 Frequency : filesystems with default stripe_count larger than 77
5675 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5676 Details : lconf specifies an inode size of 4096 bytes when the default
5677 stripe_count is larger than 77. This conflicts with the default
5678 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5679 this case to avoid pinning too much memory for large EAs.
5681 ------------------------------------------------------------------------------
5683 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5688 Frequency : rare (extremely heavy IO load with hundreds of clients)
5690 Description: Client is evicted, gets IO error writing to file
5691 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5692 caused clients to be evicted for AST timeouts. The fixes in
5693 bug 5192 mean we no longer need to have such short AST timeouts
5694 so ldlm_timeout has been increased.
5697 Frequency : occasional during --force or --failover shutdown under load
5698 Bugzilla : 5949, 4834
5699 Description: Server oops/LBUG if stopped with --force or --failover under load
5700 Details : a collection of import/export refcount and cleanup ordering
5701 issues fixed for safer force cleanup
5704 Frequency : only filesystems larger than 120 OSTs
5705 Bugzilla : 5990, 6223
5706 Description: lfs getstripe would oops on a very large filesystem
5707 Details : lov_getconfig used kfree on vmalloc'd memory
5710 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5711 Bugzilla : 6242, 6243
5712 Description: reading from files that had been truncated to a non-zero size
5713 but never opened returned no data
5714 Details : ll_file_read() reads zeros from no-object files to EOF
5719 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5720 Details : There's little harm in aborting MDS/OSS recovery and letting it
5721 try again, so I removed the LASSERT and return an error instead.
5723 Severity : enhancement
5725 Description: New debugging infrastructure for tracking down data corruption
5726 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5727 (b) cover more of the client-side code path, and (c) try to narrow
5728 down where problems occurred
5732 Bugzilla : 3819, 4364, 4397, 6313
5733 Description: Racing close and eviction MDS could cause assertion in mds_close
5734 Details : It was possible to get multiple mfd references during close and
5735 client eviction, leading to one thread referencing a freed mfd.
5737 Severity: : enhancement
5738 Bugzilla : 3262, 6359
5739 Description: Attempts to reconnect to servers are now more aggressive.
5740 Details : This builds on the enhanced upcall-less recovery that was added
5741 in 1.4.2. When trying to reconnect to servers, clients will
5742 now try each server in the failover group every 10 seconds. By
5743 default, clients would previously try one server every 25 seconds.
5748 Description: After recovery, certain operations trigger a failed
5749 assertion on a client.
5750 Details : Failing over an mds, using lconf -d --failover, while a
5751 client was doing a readdir() call would cause the client to
5752 LBUG after recovery completed and the readdir() was resent.
5754 Severity : enhancement
5756 Description: Default groups are now added by lconf
5757 Details : You can now run lconf --group <servicename> without having to
5758 manually add groups with lmc.
5761 Frequency : occasional
5763 Description: Nodes with an elan id of 0 trigger a failed assertion
5766 Frequency : always when accessing e.g. tty/console device nodes
5768 Description: tty and some other devices nodes cannot be used on lustre
5769 Details : file's private_data field is used by device data and lustre
5770 values in there got lost. New field was added to struct file to
5771 store fs-specific private data.
5774 Frequency : when exporting Lustre via NFS
5776 Description: NFSD failed occasionally when looking up a path component
5777 Details : NFSD is looking up ".." which was broken in ext3 directories
5778 that had grown large enough to become hashed.
5781 Frequency : Clusters with multiple interfaces not on the same subnet
5783 Description: Nodes will repeatedly try to reconnect to an interface which it
5784 cannot reach and report an error to the log.
5785 Details : Extra peer list entries will be created by lconf with some peers
5786 unreachable. lconf now validates the peer before adding it.
5789 Frequency : Only if a default stripe is set on the filesystem root.
5791 Description: Setting a default stripe on the filesystem root prevented the
5792 filesystem from being remounted.
5793 Details : The client was sending extra request flags in the root getattr
5794 request and did not allocate a reply buffer for the dir EA.
5797 Frequency : occasional, higher if lots of files are accessed by one client
5798 Bugzilla : 6159, 6097
5799 Description: Client trips assertion regarding lsm mismatch/magic
5800 Details : While revalidating inodes the VFS looks up inodes with ifind()
5801 and in rare cases can find an inode that is being freed.
5802 The ll_test_inode() code will free the lsm during ifind()
5803 when it finds an existing inode and then the VFS later attaches
5804 this free lsm to a new inode.
5808 Bugzilla : 6422, 7030
5809 Description: MDS deadlock between mkdir and client eviction
5810 Details : Creating a new file via mkdir or mknod (starting a transaction
5811 and getting the ns lock) can deadlock with client eviction
5812 (gets ns lock and trying to finish a synchronous transaction).
5815 Frequency : occasional
5816 Description: While starting a server, the fsfilt_ext3 module could not be
5818 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5819 kernels. Previously, lconf would still use the ext3 name
5820 when trying to load modules. Now, it will correctly use
5821 ext3 on 2.4 and ldiskfs on 2.6.
5823 Severity : enhancement
5824 Description: The default stripe count has been changed to 1
5825 Details : The interpretation of the default stripe count (0, to lfs
5826 or lmc) has been changed to mean striping across a single
5827 OST, rather than all available. For general usage we have
5828 found a stripe count of 1 or 2 works best.
5830 Severity : enhancement
5831 Description: Add support for compiling against Cray portals.
5832 Details : Conditional compiling for some areas that are different
5836 Frequency : occasional
5837 Bugzilla : 6409, 6834
5838 Description: Creating files with an explicit stripe count may lead to
5839 a failed assertion on the MDS
5840 Details : If some OSTs are full or unavailable, creating files may
5841 trigger a failed assertion on the MDS. Now, Lustre will
5842 try to use other servers or return an error to the
5846 Frequency : occasional
5848 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5849 caused lock timeout during readahead (since 1.4.2).
5850 Details : Processes doing readahead might match a lock that hasn't been
5851 granted yet if there are overlapping and conflicting lock
5852 requests. The readahead process waits on ungranted lock
5853 (original lock is CBPENDING), while OST waits for that process
5854 to cancel CBPENDING read lock and eventually evicts client.
5856 Severity : enhancement
5858 Description: Initial enabling of flock support for clients
5859 Details : Implements fcntl advisory locking and file status functions.
5860 This feature is provided as an optional mount flag (default
5861 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5862 locking are implemented yet, and those that are are not guaranteed
5863 to be completely correct in production environments.
5864 mount -t lustre -o [flock|noflock] ...
5867 Frequency : occasional
5869 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5870 assertion in the ext3 JBD (journaling) layer.
5871 Details : The b_committed_data struct is protected by the big kernel lock
5872 in 2.4 kernels, serializing journal_commit_transaction() and
5873 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5874 access to this struct. In 2.6 kernels there is finer grained
5875 locking to improve SMP performance of the JBD layer.
5879 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5884 Description: merge and cleanup kernel patches.
5887 -----------------------------------------------------------------------------
5889 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5891 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5892 versions of Lustre. Please update all servers and clients to
5893 version 1.4.2 or later at the same time. You must also run
5894 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5895 to update the configuration logs.
5897 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5898 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5899 - swab llog records of type '0' so we get proper header size/idx (5861)
5900 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5901 - fix rename of one directory over another leaking an inode (5953)
5902 - avoid SetPageDirty on 2.6 (5981)
5903 - don't re-add just-being-destroyed locks to the waiting list (5653)
5904 - when creating new directories, inherit the parent's custom
5905 striping settings if present parent (3048)
5906 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5907 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5908 - don't hold client locks on temporary worklist from l_lru (5666)
5909 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5910 - automatically evict dead clients (5921)
5911 - Update file size properly in create+truncate+fstat case (6196)
5912 - Do not unhash mountpoint dentries, do not allow removal of
5914 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5915 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5916 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5917 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5918 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5919 - Correctly handle reads of files with no objects (6243)
5920 - lctl recover will also mark a device active if deactivate used (5933)
5922 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5923 - allow --write-conf on an MDS with different nettype than client (5619)
5924 - don't write config llogs to MDS for mounts not from that MDS (5617)
5925 - lconf should create multiple TCP connections from a client (5201)
5926 - init scripts are now turned off by default; run chkconfig --on
5927 lustre and chkconfig --on lustrefs to use them
5928 - upcalls are no longer needed for clients to recover to failover
5930 - add --abort-recovery option to lconf to abort recovery on device
5932 - add support for an arbitrary number of OSTs (3026)
5933 - Quota support protocol changes.
5934 - forward compatibility changes to wire structs (6007)
5935 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5936 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5937 - improved "lustre status" script
5938 - initialize blocksize for non-regular files (6062)
5939 - added --disable-server and --disable-client configure options (5782)
5940 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5941 - Vanilla 2.4.29 support
5942 - increase maximum number of obd devices to 520 (6242)
5943 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5944 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5945 SLES 9 2.6.5 kernels
5946 - limit stripes per file to 160 (the maximum EA size) (6093)
5948 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5951 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5952 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5953 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5954 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5955 - small ext3 extents cleanups and fixes (5733)
5956 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5957 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5958 - avoid needless client->OST connect, fix handle mismatch (5317)
5959 - fix DLM error path that led to out-of-sync client, long delays (5779)
5960 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5961 - fix several locking issues related to i_size (5492,5624,5654,5672)
5962 - don't move pending lock onto export if it is already evicted (5683)
5963 - fix kernel oops when creating .foo in unlinked directory (5548)
5964 - fix deadlock in obdfilter statistics vs. object create (5811)
5965 - use time_{before,after} to avoid timer jiffies wrap (5882)
5966 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5967 - Do not leak request if server was not able to process it (5154)
5968 - If mds_open unable to find parent dir, make that negative lookup(5154)
5969 - don't create new directories with extent-mapping (5909, 5936)
5971 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5972 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5973 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5974 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5975 - support for RHEL4 kernel on i686 (5773)
5976 - provide error messages when incompatible logs are encountered (5898)
5978 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5979 * version 1.4.0.10 (1.4.1 release candidate 1)
5981 - don't keep a lock reference when lock is not granted (4238)
5982 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5983 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5984 - drop import inflight refcount on signal_completed_replay error (5255)
5985 - unlock page after async write error during send (3677)
5986 - handle missing objects in filter_preprw_read properly (5265)
5987 - no transno return for symlink open, don't save no-trasno open (3440)
5988 - don't try to complete elan receive that already failed (4012)
5989 - free RPC server reply state on error (5406)
5990 - clean up thread from ptlrpc_start_thread() on error (5160)
5991 - readahead could read extra page into cache that wasn't ejected (5388)
5992 - prevent races in class_attach/setup/cleanup/detach (5260)
5993 - don't dereference de->d_inode after l_dput of de (5458)
5994 - use "int" for stripe value returned from lock_to_stripe (5544)
5995 - mballoc allocation and error-checking fixes in 2.6 (5504)
5996 - block device patches to fix I/O request sizes in 2.6 (5482)
5997 - look up hostnames for IB nals (5602)
5998 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5999 - don't start multiple acceptors for the same port (5277)
6000 - fix incorrect LASSERT in mds_getattr_name (5635)
6001 - export a proc file for general "ping" checking (5628)
6002 - fix "lfs check" to not block when the MDS is down (5628)
6004 - service request history (4965)
6005 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6006 - create an "evict_client" /proc entry on OSTs, like the MDS has
6007 - fix mount usage message, return errors per mount(8) (5168)
6008 - change grep [] to grep "[]" in tests so they work in more UMLs
6009 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6010 - remove ancient LOV_MAGIC_V0 EA support (5047)
6011 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6012 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6013 - allow readahead allocations to fail when low on memory (5383)
6014 - mmap locking landed again, after considerable improvement (2828)
6015 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6017 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6020 - send OST transaction number in read/write reply to free req (4966)
6021 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6022 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6023 - the watchdog thread now runs as interruptible (5246)
6024 - flock/lockf fixes (but it's still disabled, pending 5135)
6025 - don't use EXT3 constants in llite code (5094)
6026 - memory shortage at startup could cause assertion (5176)
6028 - reorganization of lov code
6029 - single portals codebase
6031 - add extents/mballoc support (5025)
6032 - direct I/O reads in the obdfilter (4048)
6033 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6035 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6038 - send OST transaction number in read/write reply to free req (4966)
6039 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6040 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6041 - the watchdog thread now runs as interruptible (5246)
6042 - handle missing objects in filter_preprw_read properly (5265)
6043 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6044 - drop import inflight refcount on signal_completed_replay error (5255)
6045 - unlock page after async write error during send (3677)
6046 - return original error code on reconstructed replies (3761)
6047 - no transno return for symlink open, don't save no-trasno open (3440)
6049 - add pid to ldlm debugging output (4922)
6050 - bump the watchdog timeouts -- we can't handle 30sec yet
6051 - extra debugging for orphan dentry/inode bug (5259)
6053 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6056 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6057 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6058 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6059 - fix NULL dereference in /proc/sys/portals/routes (4827)
6060 - allow failed mdc_close() operations to be interrupted (4561)
6061 - stop precreate on OST before MDS would time out on it (4778)
6062 - don't send partial-page writes before EOF from client (4410)
6063 - discard client grant for sub-page writes on large-page clients (4520)
6064 - don't free dentries not owned by NFS code, check generation (4806)
6065 - fix lsm leak if mds_create_objects() fails (4801)
6066 - limit debug_daemon file size, always print CERROR messages (4789)
6067 - use transno after validating reply (3892)
6068 - process timed out requests if import state changes (3754)
6069 - update mtime on OST during writes, return in glimpse (4829)
6070 - add mkfsoptions to LDAP (4679)
6071 - use ->max_readahead method instead of zapping global ra (5039)
6072 - don't interrupt __l_wait_event() during strace
6074 - add software watchdogs to catch hung threads quickly (4941)
6075 - make lustrefs init script start after nfs is mounted
6076 - fix CWARN/ERROR duplication (4930)
6077 - return async write errors to application if possible (2248)
6078 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6079 - print NAL number in %x format (4645)
6080 - update barely-supported suse-2.4.21-171 series (4842)
6081 - support for sles 9 %post scripts
6082 - support for building 2.6 kernel-source packages
6083 - support for sles km_* packages
6085 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6088 - ignore -ENOENT errors in osc_destroy (3639)
6089 - notify osc create thread that OSC is being cleaned up (4600)
6090 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6091 - reconstruct ost_handle() like mds_handle() (4657)
6092 - create a new thread to do import eviction to avoid deadlock (3969)
6093 - let lconf resolve symlinked-to devices (4629)
6094 - don't unlink "objects" from directory with default EA (4554)
6095 - hold socknal file ref over connect in case target is down (4394)
6096 - allow more than 32000 subdirectories in a single directory (3244)
6097 - fix blocks count for O_DIRECT writes (3751)
6098 - OST returns ENOSPC from object create when no space left (4539)
6099 - don't send truncate RPC if file size isn't changing (4410)
6100 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6101 - bind to privileged port in socknal and tcpnal (3689)
6103 - rate limit CERROR/CWARN console message to avoid overload (4519)
6104 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6105 - basic mmap support (3918)
6106 - kernel patch series update from b1_4 (4711)
6108 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6111 - avoid crash during MDS cleanup with OST shut down (2775)
6112 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6113 - don't use bad inodes on the MDS (3744)
6114 - dynamic object preallocation to improve recovery speed (4236)
6115 - don't hold spinlock over lock dumping or change debug flags (4401)
6116 - don't zero obd_dev when it is force cleaned (3651)
6117 - print grants to console if they go negative (4431)
6118 - "lctl deactivate" will stop automatic recovery attempts (3406)
6119 - look for existing locks in ldlm_handle_enqueue() (3764)
6120 - don't resolve lock handle twice in recovery avoiding race (4401)
6121 - revalidate should check working dir is a directory (4134)
6123 - don't always mark "slow" obdfilter messages as errors (4418)
6125 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6128 - don't close LustreDB during write_conf until it is done (3860)
6129 - fix typo in lconf for_each_profile (3821)
6130 - allow dumping logs from multiple threads at one time (3820)
6131 - don't allow multiple threads in OSC recovery (3812)
6132 - fix debug_size parameters (3864)
6133 - fix mds_postrecov to initialize import for llog ctxt (3121)
6134 - replace config semaphore with spinlock (3306)
6135 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6136 - don't allow enqueue to complete on a destroyed export (3822)
6137 - down write_lock before checking llog header bitmap (3825)
6138 - recover from lock replay timeout (3764)
6139 - up llog sem before sending rpc (3652)
6140 - reduce ns lock hold times when setting kms (3267)
6141 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6142 - fix NULL deref and obd_dev leak on setup error (3312)
6143 - replace some LBUG about llog ops with error handling (3841)
6144 - don't match INVALID dentries from d_lookup and spin (3784)
6145 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6146 - fix invalid assertion in ptlrpc_set_wait (3880)
6148 - add libwrap support for the TCP acceptor (3996)
6149 - add /proc/sys/portals/routes for non-root route listing (3994)
6150 - allow setting MDS UUID in .xml (2580)
6151 - print the stack of a process that LBUGs (4228)
6153 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6156 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6157 - make sure to unset replay flag from failed open requests (3440)
6158 - if default stripe count is 0, use OST count for inode size (3636)
6159 - update parent mtime/ctime on client for create/unlink (2611)
6160 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6161 - free recovery state on server during a forced cleanup (3571)
6162 - unregister_reply for resent reqs (3063)
6163 - loop back devices mounting and status check on 2.6 (3563)
6164 - fix resource-creation race that can provoke i_size == 0 (3513)
6165 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6166 - more debugging for page-accounting assertion (3746)
6167 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6168 - avoid deadlock after precreation failure (3758)
6169 - fix race and lock order deadlock in orphan handling (3450, 3750)
6170 - add validity checks when grabbing inodes from l_ast_data (3599)
6172 - add /proc/.../recovery_status to obdfilter (3428)
6173 - lightweight CDEBUG infrastructure, debug daemon (3668)
6174 - change default OSC RPC parameters to be better on small clusters
6175 - turn off OST read cache for files smaller than 32MB
6176 - install man pages and include them in rpms (3100)
6177 - add new init script for (un)mounting lustre filesystems (2593)
6178 - run chkconfig in %post for init scripts (3701)
6179 - drop scimac NAL (unmaintained)
6181 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6184 - clean kiobufs before and after use (3485)
6185 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6186 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6187 - change init script to fail more clearly if not run as root (1528)
6188 - allow clients to reconnect during replay (1742)
6189 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6190 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6191 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6192 - fix cond_resched (3554)
6193 - only evict unfinished clients after recovery (3515)
6194 - allow bulk resend, prevent data loss (3570)
6195 - dynamic ptlrpc request buffer allocation (2102)
6196 - don't allow unlinking open directory if it isn't empty (2904)
6197 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6198 - remove extraneous obd dereference causing LASSERT failure (3334)
6199 - don't use get_cycles() when creating temp. files on the mds (3156)
6200 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6201 - handle EEXIST for set-stripe, set proper directory name (3336)
6203 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6204 - fix ksocknal_fmb_callback() error messages (2918)
6206 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6209 - don't copy lvb into (possibly NULL) reply on error (2983)
6210 - don't deref dentry after dput, don't free lvb on error (2922)
6211 - use the kms to determine writeback rpc length (2947)
6212 - increment oti_logcookies when osc is inactive (2948)
6213 - update client's i_blocks count via lvb messages (2543)
6214 - handle intent open/close of special files properly (1557)
6215 - mount MDS with errors=remount-ro, like obdfilter (2009)
6216 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6217 - don't use cancelling-locks' kms values (2947)
6218 - use highest lock extent for kms, not last one (2925)
6219 - don't dereference ERR_PTR() dentry in error handling path (3107)
6220 - fix thread race in portals_debug_dumplog() (3122)
6221 - create lprocfs device entries at setup instead of at attach (1519)
6222 - common AST error handler, don't evict client on completion race (3145)
6223 - zero nameidata in detach_mnt in 2.6 (3118)
6224 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6225 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6226 - initialize RPC timeout timer earlier for 2.6 (3219)
6227 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6228 - print nal/nid for unknown nid (3258)
6229 - additional checks for oscc recovery before doing precreate (3284)
6230 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6231 - don't crash in mdc_close for bad permissions on open (3285)
6232 - zero i_rdev for non-device files (3147)
6233 - clear page->private before handing to FS, better assertion (3119)
6234 - tune the read pipeline (3236)
6235 - fix incorrect decref of invalidated dentry (2350)
6236 - provide read-ahead stats and refine rpc in flight stats (3328)
6237 - don't hold journal transaction open across create RPC (3313)
6238 - update atime on MDS at close time (3265)
6239 - close LDAP connection when recovering to avoid server load (3315)
6240 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6241 - don't leak open file on MDS after open resend (3325)
6242 - serialize filter_precreate and filter_destroy_precreated (3329)
6243 - loop device shouldn't call sync_dev() for nul device (3092)
6244 - clear page cache after eviction (2766)
6245 - resynchronize MDS->OST in background (2824)
6246 - refuse to mount the same filesystem twice on same mountpoint (3394)
6247 - allow llmount to create routes for mounting behind routers (3320)
6248 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6249 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6250 - fix rare problem with rename on htree directories (3417)
6252 - allow default OST striping configuration per directory (1414)
6253 - fix compilation for qswnal for 2.6 kernels (3125)
6254 - increase maximum number of MDS request buffers for large systems
6255 - change liblustreapi to be useful for external progs like lfsck (3098)
6256 - increase local configuration timeout for slow disks (3353)
6257 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6259 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6262 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6263 - don't overwrite extent policy data in reply if lock was blocked (2901)
6264 - drop filter export grants atomically with removal from device (2663)
6265 - del obd_self_export from work_list in class_disconnect_exports (2908)
6266 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6267 - swab reply message in mdc_close, other PPC fixes (2464)
6268 - fix destroying of named logs (2325)
6269 - overwrite old logs when running lconf --write_conf (2264)
6270 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6271 - fix race in target_handle_connect (2898)
6272 - mds_reint_create() should take same inode create lock (2926)
6273 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6274 - don't close files for self_export to avoid uninitialized obd (2936)
6275 - allow MDS with the same name as client node (2939)
6276 - hold dentry reference for closed log files for unlink (2325)
6277 - reserve space for all logs during transactions (2059)
6278 - don't evict page beyond end of stripe extent (2925)
6279 - don't oops on a deleted current working directory (2399)
6280 - handle hard links to targets without a parent properly (2517)
6281 - don't dereference NULL lock when racing during eviction (2867)
6282 - don't grow lock extents when lots of conflicting locks (2919)
6284 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6287 - account for cache space usage on clients to avoid data loss (974)
6288 - lfsck support in lustre kernel code (2349)
6289 - reduce journal credits needed for BRW writes (2370)
6290 - orphan handling to avoid losing space on client/server crashes
6291 - ptlrpcd can be blocked, stopping ALL progress (2477)
6292 - use lock value blocks to assist in proper KMS, faster stat (1021)
6293 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6294 - recovery for initial connections (2355)
6295 - fixes for mds_cleanup_orphans (1934)
6296 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6297 - block all file creations until orphan recovery completes (1901)
6298 - client remove rq_connection from request struct (2423)
6299 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6300 - recovery timer race (2670)
6301 - mdc_close recovey bug (2532)
6302 - ptlrpc cleanup bug (2710)
6303 - mds timeout on local locks (2588)
6304 - namespace lock held during RPCs (2431)
6305 - handle interrupted sync write properly (2503)
6306 - don't try to handle a message that hasn't been replied to (2699)
6307 - client assert failure during cleanup after abort recovery (2701)
6308 - leak mdc device after failed mount (2712)
6309 - ptlrpc_check_set allows timedout requests to complete (2714)
6310 - wait for inflight reqs when ptlrpcd finishes (2710)
6311 - make sure unregistered services are removed from the srv_list
6312 - reset bulk XID's when resending them (caught by 1138 test)
6313 - unregister_bulk after timeout
6314 - fix lconf error (2694)
6315 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6316 - readahead locks pages, leaves pending causing memory pressure (2673)
6317 - increase OST request buffers to 4096 on large machines (2729)
6318 - fix up permission of existing directories in simple_mkdir (2661)
6319 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6320 - don't assign transno to errored transactions (2742)
6321 - don't delete objects on OST if given a bogus objid from MDS (2751)
6322 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6323 - if rq_no_resend, then timeout request after recovery (2432)
6324 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6325 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6326 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6327 - MDS recovery shouldn't delete objects that it has given out (2730)
6328 - if enqueue arrives after completion, don't clobber LVB (2819)
6329 - don't unlock pages twice when trigger_group_io returns error (2814)
6330 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6331 - don't write pages to disk if there was an error (1450)
6332 - don't ping imports that have recovery disabled (2676)
6333 - take buffered bytes into account when balancing socknal conn (2817)
6334 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6335 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6336 - remove little-endian swabbing of llog records (1987)
6337 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6338 - retry reposting request buffers if they fail (1191)
6339 - grow extent at grant time to avoid granting a revoked lock (2809)
6340 - lock revoke doesn't evict page if covered by a second lock (2765)
6341 - disable VM readahead to avoid reading outside lock extents (2805)
6343 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6344 - updated kernel patches for hp-2.4.20 kernel (2681)
6346 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6349 - fix truncated write corruption (2366)
6350 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6352 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6353 - ensure len(uuid) < 37 in lmc (1171)
6354 - fix ia64 OOPS in llog_test (2255)
6355 - zero end of page at obdfilter for partial page writes (2648)
6356 - don't leave stale dentries around after renames (bug 2428)
6357 - fix timeouts when evicting a client with a single lock held (2642)
6358 - set deadline for the initial HELLO message to drain (2634)
6359 - print out dotted-quad IP addresses in the socknal (2302)
6361 - additional debugging for MDS client eviction problem (2443)
6362 - fix mkfsoptions support for osts (2603, 2604)
6364 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6367 - add series for the vanilla 2.6.0 kernel
6368 - add series for the vanilla 2.4.24 kernel
6369 - add series for a cray x86/64 UL kernel drop
6370 - fix xattr patches for the vanilla 2.4.19 series
6372 - generate true UUIDs in lmc (1171)
6373 - have portals stack dumping break in UML (2466)
6374 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6375 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6376 - wake callback waiting threads on client eviction (2460)
6377 - Add --ptldebug and --subsystem to lmc (1719)
6378 - update assertion to allow safe interrupt allocation
6379 - set rq_no_resend for cancel requests (2432)
6380 - recalculate ptlrpcd timeout after resend (2494)
6381 - call vfs_rmdir when removing pending directories (2368)
6382 - fix renaming a file to itself (2429)
6383 - lmc creates a default one-stripe lov (2454)
6384 - expand procfs space to handle large clusters (2326)
6385 - increase UML stack to avoid overflow
6386 - update lconf's list of debug and subsystem masks
6387 - fix lfs find --obd (2510)
6388 - /proc tunable for disabling filter read caching (2591)
6389 - stop rpm packages from altering slapd.conf (2301)
6390 - disable nagle in the socknal under 0conf (2578)
6391 - choose mds inode size based on stripe count (2572)
6392 - fix kernel-source rpm problems (2516)
6394 - add --disable-doc to avoid pdf generation (2421)
6395 - update documentation, tests, type-os, comments
6396 - avoid format warnings on ia64
6397 - remove the TOE NAL
6398 - tiny code cleanups by removing unused fields
6400 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6403 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6404 - remove the most blatant lies from BUILDING file (2371)
6405 - change default debug level to reasonable production setting
6406 - reduce client side cache size to reduce cache flush time
6407 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6408 - make TCP zerocopy and pinger support enabled by default (2476)
6409 - sync writes completed after process exits caused crashes (2319)
6410 - maintain correct mount count on the MDS (2356)
6411 - backout 1557, because 2316 wasn't really fixed
6412 - better file I/O statistics gathering in /proc
6413 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6414 - another kernel patch to fix zero-copy TCP function export
6415 - don't take duplicate lock when processing re-sent getattr (2420)
6416 - lctl uses obd_self_export instead of creating new conn (2353)
6417 - MDS/OST recovery case which requires object creation asserted (2425)
6418 - move lfs from /usr/sbin to /usr/bin in packages
6419 - fix race between mds_client_add and mds_client_free (2417)
6420 - use kmalloc instead of slabs in portals (2430)
6421 - don't create duplicate records when a failover MDS is present (2442)
6422 - remove unnecessary mount age check (2332)
6423 - don't remove directory inodes from locks prematurely (2451)
6424 - don't break if MDS service name is the same as hostname (2103)
6425 - fix races in client write RPC generation when cache full (2482)
6427 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6430 - remove now-unused request->rq_obd (278)
6431 - if an allocation fails, print out how much memory we've used (1933)
6432 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6433 - add the "configurable stack size" patch to most series files (1256)
6434 - ability to write large log records, for 100+ OST configs (2306)
6435 - fix NULL deref when filter_prep fails (2314)
6436 - fix operator precedence error in filter_sync
6437 - dynamic allocation of socknal TX descriptors (2315)
6438 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6439 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6440 - ensure that CWARN messages in Portals always get to the syslog
6441 - __init/__exit are not for prototype decls (ldlm_init/exit)
6442 - x86-64 compile warning fixes
6443 - fix gateway LMC keyword conflict (2318)
6444 - fix MDS lock inversions in getattr/reint paths (1844)
6445 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6446 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6447 - DLM race condition prevented some lock evictions (2328)
6448 - ENOMEM detection and retry on socknal sends (2230)
6449 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6450 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6451 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6452 - better error messages when a client is rejected during recovery (1505)
6453 - avoid cancelling locks which were never granted, after failure (2330)
6454 - fix i_sem/journal inversion in mds_client_add (2333)
6455 - fix truncate/getattr lock cycle deadlock (2334)
6456 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6457 - fix two rare exit paths which could leak an l_lock() ref (2321)
6458 - fencepost error in MDS/OST orphan recovery (2226)
6459 - make log record alignment 8 bytes (1988)
6460 - lstripe now fails when requested offset > ost_count (2237)
6461 - ensure that all kernel series have a complete list.h (1607)
6462 - fix crashes in special-file operations (2316)
6463 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6465 - allow configurable automake binary, for testing new versions
6466 - small update to the lfs documentation
6468 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6470 * fix negative export reference count in fsfilt_sync (2312)
6472 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6473 * release candidate 0.9.1
6475 - orphans are moved into the PENDING directory for possible recovery
6476 - replayed opens now open by fid for orphan/rename safety (1042)
6477 - last close of an orphan inode generates a transno (683)
6478 - chdir() and mount() now pin the directory entry (1020)
6479 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6480 - discard very old requests without processing them (1502)
6481 - remove some common, well-understood CERRORs (1505)
6482 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6483 - clear "grant" flags in OST replies until OST grant code lands (1644)
6484 - fix read performance by not clobbering i_blksize on client (1598)
6485 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6486 - make LRU size a /proc tunable, clears locks when reduced (707)
6487 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6488 - server locks take references on exports now (1558)
6489 - build fixes for 2.4.20-rh trees (1663)
6490 - return an error from lov_create if all OSCs are inactive (1751)
6491 - fix import levels when a reconnect happens without a timeout (1597)
6492 - exit early from mds_open if we get a lookup error (1749)
6493 - partial page read at EOF wouldn't wait for disk before sending (1642)
6494 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6495 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6496 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6497 - write/truncate lock inversion (1639)
6498 - Don't auto-load obdclass, portals modules during cleanup (1495)
6499 - fix timestamps from jumping to "now" (1763)
6500 - extra journal assertions (1648)
6501 - add an extra multiunlink test (1771)
6502 - fix read_record/write_record API (1776)
6503 - fix leak of offset_extent, possible incorrect i_size later (1772)
6504 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6505 - Debugging for the kqswnal_get_idle_tx problems (1820)
6506 - Allow recovery to be attempted multiple times (1536)
6507 - Write out MDS last_rcvd file after it is first created (1600)
6508 - Fix tx_descriptor leak in failed transmit situations (1827)
6509 - ext3 journaling fixes for assertion failure after IO error (1871)
6510 - class_export_put() on freed export after completion AST error (1896)
6511 - Fix revalidate looping in VFS (1322)
6512 - Don't access a freed export during MDS_REINT timeout (1521)
6513 - Add open-unlink recovery support on the MDS (1673,1764)
6514 - Return an error if no MDS data was read from last_rcvd (1946)
6515 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6516 - Don't setup a disk device that doesn't match exported UUID (317)
6517 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6518 - avoid committing NULL handle in force close
6519 - local.sh is now a one-stripe LOV configuration
6520 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6521 - fix ext3 htree duplicate directory entry corruption (1516)
6522 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6523 - update to new LOV EA format (2097)
6524 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6526 2003-06-15 Phil Schwan <phil@clusterfs.com>
6529 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6530 - per-import/export recovery handling (958, 931, 959)
6531 - multiple last-rcvd slots, for serving multiple FSes (949)
6532 - connections are again shared between multiple imp/exports (963, 964)
6533 - "umount -f" would hang if any requests needed to be sent (393, 978)
6534 - avoid pinning large req buffer by copying for queued messages (989)
6535 - add "uuid" to "lctl device" command to help upcalls (991)
6536 - "open" RPCs with transnos would confuse recovery counters (1037)
6537 - do proper endian conversion of all wire messages (288, 340, 891)
6538 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6539 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6540 - fix ptlrpc_connection leak in target_handle_connect (1174)
6541 - fix import refcounting bug in OST and MDS cleanup (1134)
6542 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6543 - fix very unlikely obd_types race condition (501)
6544 - remove osc_open hack for echo_client (1187)
6545 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6546 - a failure in read_inode2 leads to deadlock (1139)
6547 - cancel ack-locks as soon as transaction is committed (1072)
6548 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6549 - make sure to commitrw after any preprw to avoid deadlock (1162)
6550 - failing to execute a file in a lustre FS would lock inode (1203)
6551 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6552 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6553 - obd_uuid2tgt would walk off the end of the list (1255)
6554 - on IA64 the osc would give portals incorrect bulk size (1258)
6555 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6556 - fix lock inversion caused by new llite matching code (1282)
6557 - limit the number of dirty pages on a client to 10MB (1286)
6558 - timed out locks were not being corrected cancelled (1289)
6559 - fix O_DIRECT above 4GB on IA-32 (1292)
6560 * major user-visible changes
6561 - fail out/fail over policy now controlled by the upcall (993)
6563 - add OBD_PING to check server availability and failure (954)
6564 - lustre messages are now sent in sending host order (288, 340, 891)
6565 - add eadatalen to MDS getattr reply (340)
6566 - OST read replies may contain second buffer, with per-page status (593)
6568 2003-03-11 Phil Schwan <phil@clusterfs.com>
6571 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6572 - failed open()s could cause deadlock; fixed (867, 869)
6573 - stop cancelling OST locks when files are closed (481)
6574 - overlapping XID spaces caused network corruption (851, 853)
6575 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6576 - setattr_raw added, to do single-RPC, server-side setattrs
6577 - lmc/lconf syntax change for OST UUIDs
6578 - fix crashy race condition between ptlrpc_free_req and osc_close
6579 - don't use request in mdc_enqueue if we hit a timeout (889)
6580 - don't set the inode i_size for regular files from the MDS (896)
6581 - handle out of order completion AST (842)
6582 - don't LBUG if a lock request times out after receiving AST (913)
6583 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6584 - if a bad lock AST arrives, send an error instead of dropping entirely
6585 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6586 - fix leak in bulk IO when only partially completed (899, 900, 926)
6587 - fix O_DIRECT for ia64 (55)
6588 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6589 - C-z after timeout could hang a process forever; fixed (977)
6591 - client-side I/O cache (678, 924, 929, 941, 970)
6593 - READPAGE and SETATTRs which don't take server-side locks get
6596 2003-02-11 Phil Schwan <phil@clusterfs.com>
6599 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6600 - Fix fsfilt_extN_readpage() to read a full page of directory
6601 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6602 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6603 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6604 - Most intent operations were removed, in favour of a new RPC mode
6605 that does a single RPC to the server and bypasses most of the VFS
6606 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6607 - Aggressively cancel local locks on DLM servers
6608 - mds_reint_unlink sends EA to the client if it's the last nlink.
6609 client uses that EA to unlink OST objects.
6610 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6611 - recursive symlinks were fixed (439)
6612 - fixed NULL deref in DEBUG_REQ
6613 - filter_update_lastobjid no longer calls sync, which annoyed extN
6614 - fixed multi-client small-writes to a single file problem (445)
6615 - fixed mtime updates during file writes (607)
6616 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6617 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6618 - fixed Linux OST size reporting problem (444, 656)
6619 - OST now updates object mtime with writes or setattr (607, 619)
6620 - client verifies file size before zeroing page past EOF (445)
6621 - OST now writes last allocated objid to disk with allocation (108)
6622 - LOV on echo now works (409)
6624 - mds_reint_unlink sends a new buffer, with the EA included. this
6625 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6626 set if a regular file was being unlinked, and it was the last link
6627 - use PtlGet from the target for bulk writes (315)
6628 - OST now updates object mtime with writes or setattr (607, 619)
6629 - LDLM now has a grant-time callback to revalidate locked items, if
6631 - Many MDS operations were reorganized to combat race conditions
6633 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6634 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6636 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6639 - Fully reactivate OST imports after reconnection (512, others)
6640 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6641 - More graceful error handling for truncating on dead OST (515)
6642 - Don't error out unless we're actually accessing dead stripes (474)
6643 - Fix garbage sizes when stripes are missing (410)
6644 - LRU counters were broken, causing constant lock purge (433, 432)
6645 - garbage on read from stripes with failed OSTs (441)
6646 - mark OSCs as active before reconnecting during recovery (438)
6647 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6648 - lfind did not preserve OST order in output (443)
6649 - symlinks cause hung clients, incorrect data (439)
6650 - stop dereferencing request after dropping refcount (457)
6651 - don't LASSERT(spin_is_locked) on non-SMP (455)
6652 - fixes for many rename() bugs
6653 - fstat didn't correctly synchronize attributes (399)
6654 - server must handle lock cancellation during blocking AST prep (487)
6655 - bulk descriptors were free()d too soon (511)
6656 - fix paths in lconf, which would load incorrect modules (451, 507)
6657 - fix confusing lconf 'host not found' error message (386)
6658 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6659 - fix race condition in mdc_blocking_ast() for inode access (526)
6660 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6661 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6662 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6663 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6664 - fix lstripe in several ways (kernel side) (527)
6665 - fix request leak in ldlm_cli_enqueue (262)
6666 - incorrect OSC was marked inactive after OST failure
6667 - call mds_fs_cleanup before unmounting filesystem (524)
6668 - fix races between taking ns_lock and ldlm_lock_change_resource
6669 - fix races updating LOV export open file list
6670 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6671 - fix recovery NULL deref in ldlm_cli_cancel_unused
6672 - fix some DLM races by using new hash table for lock handles (419)
6673 - permit the client to specify desired inodes, at replay
6674 - duplicate requests when we queue them for replay reintegration
6675 - fix last_rcvd offset calculation
6676 - sync after each recovered transaction, so we always make progress
6677 - never, not always, ERESTART requests without transnos
6678 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6680 - skip replay if the MDS says that the client is already connected
6681 - don't check for a recovery-enabled export to match lctl's UUID
6682 - don't INC_USE_COUNT for phantom exports
6683 - don't crash when cleaning up phantom exports (567)
6684 - don't double-finish or set replay data for errored mdc_open requests
6685 - abort requests when they time out, so we don't get old replies
6686 - send/receive replies for AST messages again
6687 - if the client says that it doesn't have the lock, cancel it on the
6689 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6690 mark it as destroyed, it will all work out when decref is called
6691 - fix module use counts (22, 581)
6693 - ASTs now expect a reply (server cancels lock on error reply)
6695 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6698 - fix many simultaneous client startup (392)
6699 - fix dentry->d_it clobbering
6700 - credentials weren't being shipped for readdir/getattr operations
6701 - remove invalid assertions triggered during some concurrent MD
6703 - proper Lustre versions added (336, 389)
6704 - fix memory leak for create error case (398)
6705 - fix LOV locking bug that would get cli/srv out of sync
6706 - fix echo client over LOV (409)
6707 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6708 - fix double-O_EXCL intent crash (424)
6709 - avoid sending multiple lock CANCELs (352)
6711 - MDS can do multi-client recovery (modulo bugs in new code)
6713 - many updates, edits, cleanups
6715 2002-11-18 Phil Schwan <phil@clusterfs.com>
6718 - fix null d_it dereference (346)
6719 - fix full OST/dbench hang (333)
6720 - fix permission problem with file removal (286)
6721 - fix removal of OSCs from LOV when they fail
6722 - fix NULL deref during bulk timeout (214)
6723 - fix problems related to multiple filesystems on one MDS (241)
6724 - fixed serious subtle metadata locking bugs
6725 - free locks on clients when inodes are removed due to memory
6727 - fix inode pointer in lock data (285)
6728 - partial support for multiple MDS on a single host (241)
6729 - data locks weren't cancelled at clear_inode time (290, 311)
6730 - intent locks could lead to unbounded lock growth (205)
6731 - added a maximum lock count, an LRU list, and a flusher
6732 - fix multiple rename (365)
6733 - properly abstracted the echo client
6734 - OSC locked 1 byte too many; fixed
6735 - rewrote brw callback code:
6736 - fixed recovery bugs related to LOVs (306)
6737 - fixed too-many-pages-in-one-write crash (191)
6738 - fixed (again) crash in sync_io_timeout (214)
6739 - probably fixed callback-related race (385)
6741 - Add capability to MDS protocol
6742 - LDLM cancellations and callbacks on different portals
6744 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6747 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6748 - defer open object destruction to close time (601981, 138)
6749 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6750 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6751 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6752 - don't call obd_setattr in ll_file_release for destroyed objects
6753 * protocol change to lustre_msg: move |version| and add |flags|
6754 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6755 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6756 * added replay of create, unlink, link and rename operations during
6757 MDS failover; recovery should be much more robust now
6758 * remove failed OSCs from LOVs (only lov_create uses this so far)
6759 * the lustre-HOWTO was brought (more) up to date (582544)
6761 2002-10-23 Phil Schwan <phil@clusterfs.com>
6764 - in-use dentries weren't being reused properly (617851)
6765 - prevent multiple LDLM setup (599178)
6766 - fix LOV size calculations for truncate (617853)
6767 - fix client handling of MDS intent errors (POSIX)
6768 - fix permission bug in lovstripe.c test (624321)
6769 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6770 - truncate past end of file could corrupt data
6771 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6772 - a race in recovery could return ETIMEDOUT to apps (623947)
6773 - building outside the source directory was fixed
6774 * the lustre-HOWTO was brought (more) up to date (582544)
6775 * major progress was made on recovery functionality
6777 2002-10-10 Phil Schwan <phil@clusterfs.com>
6780 - recovery deadlock fix
6781 - rm -rf causes LBUG fix (617817)
6782 - file open by multiple tasks fix (618962)
6783 - directory permissions bugs (602707 and 620007)
6784 - journal_stop fixed with locking (611313)
6785 - O_APPEND failures resolved (618273, perhaps 614459)
6786 - lconf PATH fix (619770)
6787 - IA64 build fix (621450)
6788 - RPC buffer sizes scale with amount of memory
6790 2002-10-01 Phil Schwan <phil@clusterfs.com>
6793 - locks would be cancelled without throwing away data pages,
6794 resulting in inconsistent data (605627)
6795 - inode attributes were not always being refreshed (605627, 612449)
6796 - lconf now continues to cleanup after lctl reports an error
6797 - MDS now enforces user permissions (602707)
6798 - lprocfs cleanup fixed, but not yet enabled (614157)
6799 - fixed infinite server hang, should a client not respond to an AST
6800 - avoid going into recovery if user calls readlink() with a buffer
6801 that's too small (613941)
6802 - AST RPCs no longer require replies (614867) -- this may be changed
6803 - don't crash server if client sends an IOV that's too big (611336)
6804 - fixed lock conversion deadlock (611892)
6805 - fixed the following of symlinks (614622)
6806 * recovery: the server can remove locks from a client that dies, other
6807 clients can make progress
6808 * more extN patch fixes
6809 * compile-time configurable ptlrpc buffer allocations
6811 - collaborative read cache document
6812 - Lustre Lite Performance CDR document-in-progress
6814 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6817 - fix typo in patch-2.4.18
6819 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6822 - clear ptlrpc request each time in handle_incoming_request()
6823 - unlink of files now destroys the object on the OST
6825 2002-09-19 Peter Braam <braam@clusterfs.com>
6827 * add hard link support
6828 * change obdfile creation method
6829 * kernel patch changed
6831 2002-09-19 Peter Braam <braam@clusterfs.com>
6834 - stack overflow bug in extN fixed
6836 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6838 * documentation updates
6839 - add man pages for config tools
6840 - update tests/README to describe testing with new config tools
6841 - finish metadata API descriptions
6842 * bug fixes and cleanups
6843 - statfs workaround for 16TB limit
6844 - LOV stripe allocation improved, can stripe on subset of OSTs
6845 - LOV file size/IO offset was wrong for files > 4GB in size
6846 - object EA data was being dropped, caused files to be unreadable
6847 - memory overflow with non-LOV OST caused memory corruption
6848 - fixed regression tests to work with new config tools, obdfilter
6849 - fixed bug when directory size became larger than 1 block
6850 - fixed bug (for single client case) when PWD was deleted
6851 - invalidate local directory pages when doing intent-based ops
6852 - avoid LDLM oops when lock callback contained bad data
6854 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6856 * documentation updates
6857 * bug fixes and cleanups
6858 - configuration tools
6861 - 64-bit compile warnings
6862 - 64-bit internal statfs data
6864 * test_brw on persistent OST devices
6866 * lprocfs (disabled)
6868 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6870 * documentation updates
6871 * bug fixes and cleanups
6872 * configuration tools
6874 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6877 * many small fixes to 0_5_4
6878 * io/network handling
6879 * thinkos in MDS operations
6881 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6884 * crucial basic fixes to 0.5.3
6885 * IOR, Iozone work over Elan
6888 2002-08-07 Phil Schwan <phil@clusterfs.com>
6889 * version 0_5_3, our first alpha
6890 * we use the new Portals iovs
6891 * documentation updates
6892 * bug fixes and cleanups
6893 * small changes in the DLM wire protocol
6895 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6896 * version 0_5_1 with some initial stability,
6897 * locking on MD and file I/O.
6898 * documentation updates
6899 * several bug fixes since 0.5.0
6900 * small changes in wire protocol
6902 2002-07-18 Phil Schwan <phil@clusterfs.com>
6904 * delivered as Lustre Light Alpha
6905 * fixed a crash after handling invalid MDS requests
6906 * fixed directory pages for architectures with non-4k pages sizes
6908 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6909 * release version v0_4_4
6910 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6912 * Fixes a number of interruption problems with OST operations.
6913 * Update documentation for portals header changes
6914 * Move all wire protocol structs/defines to lustre_idl.h
6915 * Fixes symlink length bug.
6916 * Add tcpdump to repository.
6918 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6919 * release version v0_4_3
6920 * Fixes statfs for inodes on extN.
6921 * Fixes bug in runtests which would delete /etc/hosts.
6922 * Use 64-bit object IDs wherever possible (not into VFS though)
6923 Remove ost_get_info, which is unused by lustre, and out of date.
6925 2002-07-03 Peter Braam <braam@clusterfs.com>
6926 * release version v0_4_2 Fixes a lookup error (type not passed)
6927 * move forward to head of Portals
6928 * move forward to latest Lustre kernel
6930 2002-06-25 Peter Braam <braam@clusterfs.com>
6931 * release version v0_4_1. Hopefully stable on single node use.