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 3.10.0-1127.19.1.el7 (RHEL7.8)
9 3.10.0-1160.49.1.el7 (RHEL7.9)
10 * Server known to build on the following kernels (others may also work):
11 2.6.32-696.18.7.el6 (RHEL6.9)
12 3.10.0-862.14.4.el7 (RHEL7.5)
13 4.4.120-92.70 (SLES12 SP2)
14 4.4.180-94.97 (SLES12 SP3)
15 4.4.0-142 (Ubuntu 16.04)
16 4.15.0-45 (Ubuntu 18.04)
17 vanilla linux 4.6.7 (ZFS only)
18 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
19 * Client primary unpatched kernels built and tested during release cycle:
20 3.10.0-957.27.2.el7 (RHEL7.6)
21 3.10.0-1062.18.1.el7 (RHEL7.7)
22 3.10.0-1127.19.1.el7 (RHEL7.8)
23 3.10.0-1160.49.1.el7 (RHEL7.9)
24 4.18.0-80.11.2.el8 (RHEL8.0)
25 4.4.180-94.97 (SLES12 SP3)
26 4.12.14-95.54 (SLES12 SP4)
27 4.12.14-122.91 (SLES12 SP5)
28 4.15.0-45 (Ubuntu 18.04)
29 * Client known to build on unpatched kernels (others may also work):
30 2.6.32-696.18.7.el6 (RHEL6.9)
31 2.6.32-754.6.3.el6 (RHEL6.10)
32 3.10.0-862.14.4.el7 (RHEL7.5)
33 4.14.0-49.13.1.el7a (RHEL7.5)
34 4.14.0-115.2.2.el7a (RHEL7.6)
35 4.18.0-147.3.1.el8 (RHEL8.1)
36 4.18.0-193.28.1.el8 (RHEL8.2)
37 4.18.0-240.22.1.el8 (RHEL8.3)
38 4.18.0-305.19.1.el8 (RHEL8.4)
39 4.18.0-348.2.1.el8 (RHEL8.5)
40 4.18.0-372.9.1.el8 (RHEL8.6)
41 4.4.120-92.70 (SLES12 SP2)
42 4.4.0-142 (Ubuntu 16.04)
43 vanilla linux 4.15.0 (ZFS only)
44 * Recommended e2fsprogs version: 1.45.6.wc1 or newer
45 * Recommended ZFS / SPL version: 0.7.13
46 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
47 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
48 stack overflow. For more information, please refer to bugzilla 17630.
49 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
50 the re-exporting nodes to be mounted with "32bitapi" mount option
51 * Note for Ubuntu 14.04.5:
52 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
53 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
57 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
58 for currently supported client and server kernel versions.
59 * Server primary kernels built and tested during release cycle:
60 3.10.0-957.27.2.el7 (RHEL7.6)
61 3.10.0-1062.18.1.el7 (RHEL7.7)
62 3.10.0-1127.19.1.el7 (RHEL7.8)
63 3.10.0-1160.49.1.el7 (RHEL7.9)
64 * Server known to build on the following kernels (others may also work):
65 2.6.32-696.18.7.el6 (RHEL6.9)
66 3.10.0-862.14.4.el7 (RHEL7.5)
67 4.4.120-92.70 (SLES12 SP2)
68 4.4.180-94.97 (SLES12 SP3)
69 4.4.0-142 (Ubuntu 16.04)
70 4.15.0-45 (Ubuntu 18.04)
71 vanilla linux 4.6.7 (ZFS only)
72 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
73 * Client primary unpatched kernels built and tested during release cycle:
74 3.10.0-957.27.2.el7 (RHEL7.6)
75 3.10.0-1062.18.1.el7 (RHEL7.7)
76 3.10.0-1127.19.1.el7 (RHEL7.8)
77 3.10.0-1160.49.1.el7 (RHEL7.9)
78 4.18.0-80.11.2.el8 (RHEL8.0)
79 4.4.180-94.97 (SLES12 SP3)
80 4.12.14-95.54 (SLES12 SP4)
81 4.12.14-122.91 (SLES12 SP5)
82 4.15.0-45 (Ubuntu 18.04)
83 * Client known to build on unpatched kernels (others may also work):
84 2.6.32-696.18.7.el6 (RHEL6.9)
85 2.6.32-754.6.3.el6 (RHEL6.10)
86 3.10.0-862.14.4.el7 (RHEL7.5)
87 4.14.0-49.13.1.el7a (RHEL7.5)
88 4.14.0-115.2.2.el7a (RHEL7.6)
89 4.18.0-147.3.1.el8 (RHEL8.1)
90 4.18.0-193.28.1.el8 (RHEL8.2)
91 4.18.0-240.22.1.el8 (RHEL8.3)
92 4.18.0-305.19.1.el8 (RHEL8.4)
93 4.18.0-348.2.1.el8 (RHEL8.5)
94 4.4.120-92.70 (SLES12 SP2)
95 4.4.0-142 (Ubuntu 16.04)
96 vanilla linux 4.15.0 (ZFS only)
97 * Recommended e2fsprogs version: 1.45.6.wc1 or newer
98 * Recommended ZFS / SPL version: 0.7.13
99 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
100 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
101 stack overflow. For more information, please refer to bugzilla 17630.
102 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
103 the re-exporting nodes to be mounted with "32bitapi" mount option
104 * Note for Ubuntu 14.04.5:
105 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
106 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
110 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
111 for currently supported client and server kernel versions.
112 * Server primary kernels built and tested during release cycle:
113 3.10.0-957.27.2.el7 (RHEL7.6)
114 3.10.0-1062.18.1.el7 (RHEL7.7)
115 3.10.0-1127.19.1.el7 (RHEL7.8)
116 3.10.0-1160.45.1.el7 (RHEL7.9)
117 * Server known to build on the following kernels (others may also work):
118 2.6.32-696.18.7.el6 (RHEL6.9)
119 3.10.0-862.14.4.el7 (RHEL7.5)
120 4.4.120-92.70 (SLES12 SP2)
121 4.4.180-94.97 (SLES12 SP3)
122 4.4.0-142 (Ubuntu 16.04)
123 4.15.0-45 (Ubuntu 18.04)
124 vanilla linux 4.6.7 (ZFS only)
125 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
126 * Client primary unpatched kernels built and tested during release cycle:
127 3.10.0-957.27.2.el7 (RHEL7.6)
128 3.10.0-1062.18.1.el7 (RHEL7.7)
129 3.10.0-1127.19.1.el7 (RHEL7.8)
130 3.10.0-1160.45.1.el7 (RHEL7.9)
131 4.18.0-80.11.2.el8 (RHEL8.0)
132 4.4.180-94.97 (SLES12 SP3)
133 4.12.14-95.54 (SLES12 SP4)
134 4.12.14-122.91 (SLES12 SP5)
135 4.15.0-45 (Ubuntu 18.04)
136 * Client known to build on unpatched kernels (others may also work):
137 2.6.32-696.18.7.el6 (RHEL6.9)
138 2.6.32-754.6.3.el6 (RHEL6.10)
139 3.10.0-862.14.4.el7 (RHEL7.5)
140 4.14.0-49.13.1.el7a (RHEL7.5)
141 4.14.0-115.2.2.el7a (RHEL7.6)
142 4.18.0-147.3.1.el8 (RHEL8.1)
143 4.18.0-193.28.1.el8 (RHEL8.2)
144 4.18.0-240.22.1.el8 (RHEL8.3)
145 4.18.0-305.25.1.el8 (RHEL8.4)
146 4.18.0-348.2.1.el8 (RHEL8.5)
147 4.4.120-92.70 (SLES12 SP2)
148 4.4.0-142 (Ubuntu 16.04)
149 vanilla linux 4.15.0 (ZFS only)
150 * Recommended e2fsprogs version: 1.45.6.wc1 or newer
151 * Recommended ZFS / SPL version: 0.7.13
152 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
153 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
154 stack overflow. For more information, please refer to bugzilla 17630.
155 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
156 the re-exporting nodes to be mounted with "32bitapi" mount option
157 * Note for Ubuntu 14.04.5:
158 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
159 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
163 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
164 for currently supported client and server kernel versions.
165 * Server primary kernels built and tested during release cycle:
166 3.10.0-957.27.2.el7 (RHEL7.6)
167 3.10.0-1062.18.1.el7 (RHEL7.7)
168 3.10.0-1127.19.1.el7 (RHEL7.8)
169 3.10.0-1160.15.2.el7 (RHEL7.9)
170 * Server known to build on the following kernels (others may also work):
171 2.6.32-696.18.7.el6 (RHEL6.9)
172 3.10.0-862.14.4.el7 (RHEL7.5)
173 4.4.120-92.70 (SLES12 SP2)
174 4.4.180-94.97 (SLES12 SP3)
175 4.4.0-142 (Ubuntu 16.04)
176 4.15.0-45 (Ubuntu 18.04)
177 vanilla linux 4.6.7 (ZFS only)
178 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
179 * Client primary unpatched kernels built and tested during release cycle:
180 3.10.0-957.27.2.el7 (RHEL7.6)
181 3.10.0-1062.18.1.el7 (RHEL7.7)
182 3.10.0-1127.19.1.el7 (RHEL7.8)
183 3.10.0-1160.15.2.el7 (RHEL7.9)
184 4.18.0-80.11.2.el8 (RHEL8.0)
185 4.4.180-94.97 (SLES12 SP3)
186 4.12.14-95.54 (SLES12 SP4)
187 4.12.14-122.57 (SLES12 SP5)
188 4.15.0-45 (Ubuntu 18.04)
189 * Client known to build on unpatched kernels (others may also work):
190 2.6.32-696.18.7.el6 (RHEL6.9)
191 2.6.32-754.6.3.el6 (RHEL6.10)
192 3.10.0-862.14.4.el7 (RHEL7.5)
193 4.14.0-49.13.1.el7a (RHEL7.5)
194 4.14.0-115.2.2.el7a (RHEL7.6)
195 4.18.0-147.3.1.el8 (RHEL8.1)
196 4.18.0-193.28.1.el8 (RHEL8.2)
197 4.18.0-240.1.1.el8 (RHEL8.3)
198 4.4.120-92.70 (SLES12 SP2)
199 4.4.0-142 (Ubuntu 16.04)
200 vanilla linux 4.15.0 (ZFS only)
201 * Recommended e2fsprogs version: 1.45.6.wc1 or newer
202 * Recommended ZFS / SPL version: 0.7.13
203 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
204 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
205 stack overflow. For more information, please refer to bugzilla 17630.
206 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
207 the re-exporting nodes to be mounted with "32bitapi" mount option
208 * Note for Ubuntu 14.04.5:
209 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
210 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
214 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
215 for currently supported client and server kernel versions.
216 * Server primary kernels built and tested during release cycle:
217 3.10.0-957.27.2.el7 (RHEL7.6)
218 3.10.0-1062.18.1.el7 (RHEL7.7)
219 3.10.0-1127.8.2.el7 (RHEL7.8)
220 * Server known to build on the following kernels (others may also work):
221 2.6.32-696.18.7.el6 (RHEL6.9)
222 3.10.0-862.14.4.el7 (RHEL7.5)
223 4.4.120-92.70 (SLES12 SP2)
224 4.4.180-94.97 (SLES12 SP3)
225 4.4.0-142 (Ubuntu 16.04)
226 4.15.0-45 (Ubuntu 18.04)
227 vanilla linux 4.6.7 (ZFS only)
228 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
229 * Client primary unpatched kernels built and tested during release cycle:
230 3.10.0-957.27.2.el7 (RHEL7.6)
231 3.10.0-1062.18.1.el7 (RHEL7.7)
232 3.10.0-1127.8.2.el7 (RHEL7.8)
233 4.18.0-80.11.2.el8 (RHEL8.0)
234 4.4.180-94.97 (SLES12 SP3)
235 4.12.14-95.45 (SLES12 SP4)
236 4.12.14-122.20 (SLES12 SP5)
237 4.15.0-45 (Ubuntu 18.04)
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 4.14.0-49.13.1.el7a (RHEL7.5)
243 4.14.0-115.2.2.el7a (RHEL7.6)
244 4.18.0-147.8.1.el8 (RHEL8.1)
245 4.18.0-193.1.2.el8 (RHEL8.2)
246 4.4.120-92.70 (SLES12 SP2)
247 4.4.0-142 (Ubuntu 16.04)
248 vanilla linux 4.15.0 (ZFS only)
249 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
250 * Recommended ZFS / SPL version: 0.7.13
251 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
252 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
253 stack overflow. For more information, please refer to bugzilla 17630.
254 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
255 the re-exporting nodes to be mounted with "32bitapi" mount option
256 * Note for Ubuntu 14.04.5:
257 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
258 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
262 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
263 for currently supported client and server kernel versions.
264 * Server primary kernels built and tested during release cycle:
265 3.10.0-957.27.2.el7 (RHEL7.6)
266 3.10.0-1062.9.1.el7 (RHEL7.7)
267 * Server known to build on the following kernels (others may also work):
268 2.6.32-696.18.7.el6 (RHEL6.9)
269 3.10.0-862.14.4.el7 (RHEL7.5)
270 4.4.120-92.70 (SLES12 SP2)
271 4.4.180-94.97 (SLES12 SP3)
272 4.4.0-142 (Ubuntu 16.04)
273 4.15.0-45 (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 primary unpatched kernels built and tested during release cycle:
277 3.10.0-957.27.2.el7 (RHEL7.6)
278 3.10.0-1062.9.1.el7 (RHEL7.7)
279 4.18.0-80.11.2.el8 (RHEL8.0)
280 4.4.180-94.97 (SLES12 SP3)
281 4.12.14-95.29 (SLES12 SP4)
282 4.15.0-45 (Ubuntu 18.04)
283 * Client known to build on unpatched kernels (others may also work):
284 2.6.32-696.18.7.el6 (RHEL6.9)
285 2.6.32-754.6.3.el6 (RHEL6.10)
286 3.10.0-862.14.4.el7 (RHEL7.5)
287 4.14.0-49.13.1.el7a (RHEL7.5)
288 4.14.0-115.2.2.el7a (RHEL7.6)
289 4.18.0-147.3.1.el8 (RHEL8.1)
290 4.4.120-92.70 (SLES12 SP2)
291 4.4.0-142 (Ubuntu 16.04)
292 vanilla linux 4.15.0 (ZFS only)
293 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
294 * Recommended ZFS / SPL version: 0.7.13
295 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
296 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
297 stack overflow. For more information, please refer to bugzilla 17630.
298 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
299 the re-exporting nodes to be mounted with "32bitapi" mount option
300 * Note for Ubuntu 14.04.5:
301 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
302 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
306 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
307 for currently supported client and server kernel versions.
308 * Server primary kernels built and tested during release cycle:
309 3.10.0-957.27.2.el7 (RHEL7.6)
310 3.10.0-1062.1.1.el7 (RHEL7.7)
311 * Server known to build on the following kernels (others may also work):
312 2.6.32-696.18.7.el6 (RHEL6.9)
313 3.10.0-862.14.4.el7 (RHEL7.5)
314 4.4.120-92.70 (SLES12 SP2)
315 4.4.180-94.97 (SLES12 SP3)
316 4.4.0-142 (Ubuntu 16.04)
317 4.15.0-45 (Ubuntu 18.04)
318 vanilla linux 4.6.7 (ZFS only)
319 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
320 * Client primary unpatched kernels built and tested during release cycle:
321 3.10.0-957.27.2.el7 (RHEL7.6)
322 3.10.0-1062.1.1.el7 (RHEL7.7)
323 4.18.0-80.7.1.el8 (RHEL8.0)
324 4.4.180-94.97 (SLES12 SP3)
325 4.12.14-95.29 (SLES12 SP4)
326 4.15.0-45 (Ubuntu 18.04)
327 * Client known to build on unpatched kernels (others may also work):
328 2.6.32-696.18.7.el6 (RHEL6.9)
329 2.6.32-754.6.3.el6 (RHEL6.10)
330 3.10.0-862.14.4.el7 (RHEL7.5)
331 4.14.0-49.13.1.el7a (RHEL7.5)
332 4.14.0-115.2.2.el7a (RHEL7.6)
333 4.4.120-92.70 (SLES12 SP2)
334 4.4.0-142 (Ubuntu 16.04)
335 vanilla linux 4.15.0 (ZFS only)
336 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
337 * Recommended ZFS / SPL version: 0.7.13
338 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
339 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
340 stack overflow. For more information, please refer to bugzilla 17630.
341 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
342 the re-exporting nodes to be mounted with "32bitapi" mount option
343 * Note for Ubuntu 14.04.5:
344 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
345 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
349 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
350 for currently supported client and server kernel versions.
351 * Server known to build on the following kernels (others may also work):
352 2.6.32-696.18.7.el6 (RHEL6.9)
353 3.10.0-862.14.4.el7 (RHEL7.5)
354 3.10.0-957.10.1.el7 (RHEL7.6)
355 4.4.120-92.70 (SLES12 SP2)
356 4.4.162-94.69 (SLES12 SP3)
357 4.4.0-142 (Ubuntu 16.04)
358 4.15.0-45 (Ubuntu 18.04)
359 vanilla linux 4.6.7 (ZFS only)
360 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
361 * Client known to build on unpatched kernels (others may also work):
362 2.6.32-696.18.7.el6 (RHEL6.9)
363 2.6.32-754.6.3.el6 (RHEL6.10)
364 3.10.0-862.14.4.el7 (RHEL7.5)
365 3.10.0-957.10.1.el7 (RHEL7.6)
366 4.14.0-49.13.1.el7a (RHEL7.5)
367 4.14.0-115.2.2.el7a (RHEL7.6)
368 4.4.120-92.70 (SLES12 SP2)
369 4.4.162-94.69 (SLES12 SP3)
370 4.12.14-95.13 (SLES12 SP4)
371 4.4.0-142 (Ubuntu 16.04)
372 4.15.0-45 (Ubuntu 18.04)
373 vanilla linux 4.15.0 (ZFS only)
374 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
375 * Recommended ZFS / SPL version: 0.7.13
376 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
377 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
378 stack overflow. For more information, please refer to bugzilla 17630.
379 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
380 the re-exporting nodes to be mounted with "32bitapi" mount option
381 * Note for Ubuntu 14.04.5:
382 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
383 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
388 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
389 for currently supported client and server kernel versions.
390 * Server known to build on the following kernels (others may also work):
391 2.6.32-696.18.7.el6 (RHEL6.9)
392 3.10.0-862.14.4.el7 (RHEL7.5)
393 3.10.0-957.10.1.el7 (RHEL7.6)
394 4.4.120-92.70 (SLES12 SP2)
395 4.4.162-94.69 (SLES12 SP3)
396 4.4.0-142 (Ubuntu 16.04)
397 4.15.0-45 (Ubuntu 18.04)
398 vanilla linux 4.6.7 (ZFS only)
399 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
400 * Client known to build on unpatched kernels (others may also work):
401 2.6.32-696.18.7.el6 (RHEL6.9)
402 2.6.32-754.6.3.el6 (RHEL6.10)
403 3.10.0-862.14.4.el7 (RHEL7.5)
404 3.10.0-957.10.1.el7 (RHEL7.6)
405 4.14.0-49.13.1.el7a (RHEL7.5)
406 4.14.0-115.2.2.el7a (RHEL7.6)
407 4.4.120-92.70 (SLES12 SP2)
408 4.4.162-94.69 (SLES12 SP3)
409 4.12.14-95.13 (SLES12 SP4)
410 4.4.0-142 (Ubuntu 16.04)
411 4.15.0-45 (Ubuntu 18.04)
412 vanilla linux 4.15.0 (ZFS only)
413 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
414 * Recommended ZFS / SPL version: 0.7.13
415 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
416 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
417 stack overflow. For more information, please refer to bugzilla 17630.
418 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
419 the re-exporting nodes to be mounted with "32bitapi" mount option
420 * Note for Ubuntu 14.04.5:
421 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
422 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
426 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
427 for currently supported client and server kernel versions.
428 * Server known to build on the following kernels (others may also work):
429 2.6.32-696.18.7.el6 (RHEL6.9)
430 3.10.0-862.14.4.el7 (RHEL7.5)
431 3.10.0-957.el7 (RHEL7.6)
432 4.4.120-92.70 (SLES12 SP2)
433 4.4.162-94.69 (SLES12 SP3)
434 4.4.0-131 (Ubuntu 16.04)
435 4.15.0-32 (Ubuntu 18.04)
436 vanilla linux 4.6.7 (ZFS only)
437 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
438 * Client known to build on unpatched kernels (others may also work):
439 2.6.32-696.18.7.el6 (RHEL6.9)
440 2.6.32-754.6.3.el6 (RHEL6.10)
441 3.10.0-862.14.4.el7 (RHEL7.5)
442 3.10.0-957.el7 (RHEL7.6)
443 4.14.0-49.13.1.el7a (RHEL7.5)
444 4.14.0-115.2.2.el7a (RHEL7.6)
445 4.4.120-92.70 (SLES12 SP2)
446 4.4.162-94.69 (SLES12 SP3)
447 4.4.0-131 (Ubuntu 16.04)
448 4.15.0-32 (Ubuntu 18.04)
449 vanilla linux 4.15.0 (ZFS only)
450 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
451 * Recommended ZFS / SPL version: 0.7.9
452 * Tested with ZFS / SPL version: 0.7.9
453 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
454 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
455 stack overflow. For more information, please refer to bugzilla 17630.
456 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
457 the re-exporting nodes to be mounted with "32bitapi" mount option
458 * Note for Ubuntu 14.04.5:
459 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
460 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
462 --------------------------------------------------------------------------------
465 2018-04-03 Intel Corporation
467 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
468 for currently supported client and server kernel versions.
469 * Server known to build on patched kernels:
470 2.6.32-431.29.2.el6 (RHEL6.5)
471 2.6.32-504.30.3.el6 (RHEL6.6)
472 2.6.32-573.26.1.el6 (RHEL6.7)
473 2.6.32-642.15.1.el6 (RHEL6.8)
474 2.6.32-696.18.7.el6 (RHEL6.9)
475 3.10.0-693.21.1.el7 (RHEL7.4)
476 3.0.101-0.47.71 (SLES11 SP3)
477 3.0.101-107 (SLES11 SP4)
478 3.12.74-60.64.40 (SLES12 SP1)
479 4.4.114-92.64 (SLES12 SP2)
480 4.4.114-94.11 (SLES12 SP3)
481 3.13.0-101 (Ubuntu 14.04, ZFS only)
482 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
483 4.4.0-116 (Ubuntu 16.04)
484 vanilla linux 4.6.7 (ZFS only)
485 * Client known to build on unpatched kernels:
486 2.6.32-431.29.2.el6 (RHEL6.5)
487 2.6.32-504.30.3.el6 (RHEL6.6)
488 2.6.32-573.26.1.el6 (RHEL6.7)
489 2.6.32-642.15.1.el6 (RHEL6.8)
490 2.6.32-696.18.7.el6 (RHEL6.9)
491 3.10.0-693.21.1.el7 (RHEL7.4)
492 3.0.101-0.47.71 (SLES11 SP3)
493 3.0.101-107 (SLES11 SP4)
494 3.12.74-60.64.40 (SLES12 SP1)
495 4.4.114-92.64 (SLES12 SP2)
496 4.4.114-94.11 (SLES12 SP3)
497 3.13.0-101 (Ubuntu 14.04)
498 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
499 4.4.0-116 (Ubuntu 16.04)
501 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
502 * Recommended ZFS / SPL version: 0.7.6
503 * Tested with ZFS / SPL version: 0.7.6
504 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
505 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
506 stack overflow. For more information, please refer to bugzilla 17630.
507 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
508 the re-exporting nodes to be mounted with "32bitapi" mount option
509 * Note for Ubuntu 14.04.5:
510 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
511 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
513 --------------------------------------------------------------------------------
515 2017-07-05 Intel Corporation
517 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
518 for currently supported client and server kernel versions.
519 * Server known to build on patched kernels:
520 2.6.32-431.29.2.el6 (RHEL6.5)
521 2.6.32-504.30.3.el6 (RHEL6.6)
522 2.6.32-573.26.1.el6 (RHEL6.7)
523 2.6.32-642.15.1.el6 (RHEL6.8)
524 2.6.32-696.3.1.el6 (RHEL6.9)
525 3.10.0-514.21.1.el7 (RHEL7.3)
526 3.0.101-0.47.71 (SLES11 SP3)
527 3.0.101-100 (SLES11 SP4)
528 3.12.74-60.64.40 (SLES12 SP1)
529 4.4.59-92.17 (SLES12 SP2)
530 vanilla linux 4.6.7 (ZFS only)
531 * Client known to build on unpatched kernels:
532 2.6.32-431.29.2.el6 (RHEL6.5)
533 2.6.32-504.30.3.el6 (RHEL6.6)
534 2.6.32-573.26.1.el6 (RHEL6.7)
535 2.6.32-642.15.1.el6 (RHEL6.8)
536 2.6.32-696.3.1.el6 (RHEL6.9)
537 3.10.0-514.21.1.el7 (RHEL7.3)
538 3.0.101-0.47.71 (SLES11 SP3)
539 3.0.101-100 (SLES11 SP4)
540 3.12.74-60.64.40 (SLES12 SP1)
541 4.4.59-92.17 (SLES12 SP2)
543 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
544 * Recommended ZFS / SPL version: 0.6.5.9
545 * Tested with ZFS / SPL version: 0.6.5.9
546 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
547 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
548 stack overflow. For more information, please refer to bugzilla 17630.
549 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
550 the re-exporting nodes to be mounted with "32bitapi" mount option
552 --------------------------------------------------------------------------------
554 12-07-2016 Intel Corporation
556 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
557 for currently supported client and server kernel versions.
558 * Server known to build on patched kernels:
559 2.6.32-431.29.2.el6 (RHEL6.5)
560 2.6.32-504.30.3.el6 (RHEL6.6)
561 2.6.32-573.26.1.el6 (RHEL6.7)
562 2.6.32-642.11.1.el6 (RHEL6.8)
563 3.10.0-514.el7 (RHEL7.3)
564 3.0.101-0.47.71 (SLES11 SP3)
565 3.0.101-80 (SLES11 SP4)
566 3.12.62-60.64.8 (SLES12 SP1)
567 vanilla linux 4.5.7 (ZFS only)
568 * Client known to build on unpatched kernels:
569 2.6.32-431.29.2.el6 (RHEL6.5)
570 2.6.32-504.30.3.el6 (RHEL6.6)
571 2.6.32-573.26.1.el6 (RHEL6.7)
572 2.6.32-642.11.1.el6 (RHEL6.8)
573 3.10.0-514.el7 (RHEL7.3)
574 3.0.101-0.47.71 (SLES11 SP3)
575 3.0.101-80 (SLES11 SP4)
576 3.12.62-60.64.8 (SLES12 SP1)
578 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
579 * Recommended ZFS / SPL version: 0.6.5.8
580 * Tested with ZFS / SPL version: 0.6.5.8
581 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
582 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
583 stack overflow. For more information, please refer to bugzilla 17630.
584 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
585 the re-exporting nodes to be mounted with "32bitapi" mount option
587 --------------------------------------------------------------------------------
589 02-29-2016 Intel Corporation
591 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
592 for currently supported client and server kernel versions.
593 * Server known to build on patched kernels:
594 2.6.32-431.29.2.el6 (RHEL6.5)
595 2.6.32-504.30.3.el6 (RHEL6.6)
596 2.6.32-573.12.1.el6 (RHEL6.7)
597 3.10.0-327.3.1.el7 (RHEL7.2)
598 3.0.101-0.47.71 (SLES11 SP3)
599 3.0.101-68 (SLES11 SP4)
600 vanilla linux 4.2.1 (ZFS only)
601 * Client known to build on unpatched kernels:
602 2.6.32-431.29.2.el6 (RHEL6.5)
603 2.6.32-504.30.3.el6 (RHEL6.6)
604 2.6.32-573.12.1.el6 (RHEL6.7)
605 3.10.0-327.3.1.el7 (RHEL7.2)
606 3.0.101-0.47.71 (SLES11 SP3)
607 3.0.101-68 (SLES11 SP4)
610 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
611 * Recommended ZFS / SPL version: 0.6.4.2
612 * Tested with ZFS / SPL version: 0.6.4.2
613 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
614 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
615 stack overflow. For more information, please refer to bugzilla 17630.
616 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
617 the re-exporting nodes to be mounted with "32bitapi" mount option
619 --------------------------------------------------------------------------------
621 03-10-2015 Intel Corporation
623 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
624 for currently supported client and server kernel versions.
625 * Server known to build on patched kernels:
626 2.6.32-431.29.2.el6 (RHEL6.5)
627 2.6.32-504.8.1.el6 (RHEL6.6)
628 3.0.101-0.46 (SLES11 SP3)
629 * Client known to build on unpatched kernels:
630 2.6.32-431.29.2.el6 (RHEL6.5)
631 2.6.32-504.8.1.el6 (RHEL6.6)
632 3.10.0-123.20.1.el7 (RHEL7)
633 3.0.101-0.46 (SLES11 SP3)
634 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
635 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
636 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
637 stack overflow. For more information, please refer to bugzilla 17630.
638 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
639 the re-exporting nodes to be mounted with "32bitapi" mount option
641 Severity : enhancement
643 Description: control OST-index in IDIF via ROCOMPAT flag.
644 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
645 last_rcvd file. For new formatted OST device, it will be auto set;
646 for the case of upgrading from old OST device, you can enable it
647 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
648 enabled, for new created OST-object, its IDIF-in-LMA will contain
649 the OST-index; for the existing OST-object, the OSD will convert
650 old format IDIF as new format IDIF with OST-index stored in the
651 LMA EA when accessing such OST-object or via OI scrub. Once such
652 flag is enabled, it cannot be reverted back, so the system cannot
653 be downgraded to the orignal incompatible version.
655 --------------------------------------------------------------------------------
657 07-30-2014 Intel Corporation
659 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
660 for currently supported client and server kernel versions.
661 * Server known to build on patched kernels:
662 2.6.32-431.20.3.el6 (RHEL6.5)
663 3.0.101-0.31 (SLES11 SP3)
664 * Client known to build on unpatched kernels:
665 2.6.32-431.20.3.el6 (RHEL6.5)
666 3.0.101-0.31 (SLES11 SP3)
669 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
670 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
671 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
672 stack overflow. For more information, please refer to bugzilla 17630.
673 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
674 the re-exporting nodes to be mounted with "32bitapi" mount option
676 --------------------------------------------------------------------------------
678 10-31-2013 Intel Corporation
680 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
681 * Server builds for kernels:
682 2.6.32-358.18.1.el6 (RHEL6.4)
683 * Client builds for unpatched kernels:
684 2.6.32-358.18.1.el6 (RHEL6.4)
685 3.0.92-0.8 (SLES11 SP3)
687 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
688 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
689 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
690 stack overflow. For more information, please refer to bugzilla 17630.
691 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
692 the re-exporting nodes to be mounted with "32bitapi" mount option
694 --------------------------------------------------------------------------------
696 05-30-2013 Intel Corporation
698 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
699 * Server builds for kernels:
700 2.6.32-358.6.2.el6 (RHEL6.4)
701 * Client builds for unpatched kernels:
702 2.6.32-358.6.2.el6 (RHEL6.4)
703 3.0.74-0.6.6 (SLES11 SP2)
705 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
706 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
707 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
708 stack overflow. For more information, please refer to bugzilla 17630.
709 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
710 the re-exporting nodes to be mounted with "32bitapi" mount option
712 --------------------------------------------------------------------------------
714 9-30-2012 Whamcloud, Inc.
716 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
717 * Server support for kernels:
718 2.6.32-279.5.1.el6 (RHEL6)
719 * Client support for unpatched kernels:
720 2.6.32-279.5.1.el6 (RHEL6)
722 2.6.18-238.19.1.el5 (RHEL5)
723 2.6.16.60-0.69.1 (SLES10)
724 * Recommended e2fsprogs version: 1.41.90.wc4
725 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
726 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
727 stack overflow. For more information, please refer to bugzilla 17630.
728 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
729 the re-exporting nodes to be mounted with "32bitapi" mount option
731 Severity : enhancement
733 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
735 Severity : enhancement
737 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
739 Severity : enhancement
741 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
743 Severity : enhancement
745 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
747 Severity : enhancement
749 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
751 --------------------------------------------------------------------------------
753 04-30-2012 Whamcloud, Inc.
755 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
756 * Server support for kernels:
757 2.6.32-220.4.2.el6 (RHEL6)
758 * Client support for unpatched kernels:
759 2.6.32-220.4.2.el6 (RHEL6)
761 2.6.18-238.19.1.el5 (RHEL5)
762 2.6.16.60-0.69.1 (SLES10)
763 * Recommended e2fsprogs version: 1.41.90.wc4
764 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
765 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
766 stack overflow. For more information, please refer to bugzilla 17630.
767 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
768 the re-exporting nodes to be mounted with "32bitapi" mount option
770 Severity : enhancement
772 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
774 Severity : enhancement
776 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
778 Severity : enhancement
780 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
782 Severity : enhancement
784 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
786 --------------------------------------------------------------------------------
788 09-30-2011 Whamcloud, Inc.
790 * Server support for kernels:
791 2.6.18-238.19.1.el5 (RHEL5)
792 2.6.32-131.6.1.el6 (RHEL6)
793 * Client support for unpatched kernels:
794 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
795 2.6.18-238.19.1.el5 (RHEL5)
796 2.6.32-131.6.1.el6 (RHEL6)
797 2.6.29.4-167.fc11 (FC11)
798 2.6.16.60-0.69.1 (SLES10)
800 * Recommended e2fsprogs version: 1.41.90.wc3
801 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
802 Lustre filesystem with 4K stack may cause a stack overflow. For more
803 information, please refer to bugzilla 17630.
804 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
805 mounted with "32bitapi" mount option
807 Severity : enhancement
808 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
810 Severity : enhancement
811 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
813 Severity : enhancement
814 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
817 Frequency : only with SLES10
819 Description: Use OFED "KMP" provided by Novell
820 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
821 Lustre is now built against this package. That means you need to
822 install the ofed-kmp package from Novell for the patchless client.
823 Note that the ofed-kmp that Novell ships may not exactly match the
824 kernel version but should still be compatible.
826 Severity : enhancement
828 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
829 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
831 --------------------------------------------------------------------------------
833 2010-07-15 Oracle, Inc.
835 * Server support for kernels:
836 2.6.18-164.11.1.el5 (RHEL 5)
837 2.6.18-164.11.1.0.1.el5 (OEL 5)
838 * Client support for unpatched kernels:
839 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
840 2.6.18-164.11.1.el5 (RHEL 5),
841 2.6.18-164.11.1.0.1.el5 (OEL 5)
842 2.6.16.60-0.42.8 (SLES 10),
844 2.6.29.4-167.fc11 (FC11)
845 * Recommended e2fsprogs version: 1.41.10-sun2
847 Severity : enhancement
849 Description: Distribute lfs_migrate script with lustre
850 Details : The lfs_migrate mentioned in the manual is now included in the
851 lustre tree in lustre/scripts and will distributed in lustre RPMs.
853 Severity : enhancement
855 Description: Failover nid list added to import proc file.
857 Severity : enhancement
859 Description: "som_preview" mount option required to use SOM
860 Details : The clients will need to use som_preview mount option
864 Description: Lustre does not do 1MB IOs to HW RAID
865 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
867 Severity : enhancement
869 Description: lfs quota output is not very convenient for awk/sed-parsing
870 Details : Some positions in lfs quota output table could be empty or
871 non-empty which made it hard to parse it with scripts, now
872 a dash is put instead of space where there is not supposed
877 Description: Update to OFED 1.5.1
878 Details : OFED 1.5.1 will be used in environments that don't use
879 vendor supplied OFED.
883 Description: Remove lnet. prefix from lctl params display
884 Details : Also renamed lnet.memused to lnet_memused to
889 Description: Use conf_param -d to remove permanant settings.
893 Description: Add list_param -R
894 Details : lctl list_param now supports recursive mode.
897 Bugzilla : 21829, 22312
898 Description: Fix llobdstat.
899 Details : 1. Now updating the write_bytes counter
900 2. Port script fixes from 1.8.
904 Description: decrease the usage of memory on clients.
905 Details : 1. On clients, recycle dentries and inodes unused.
906 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
911 Description: llobdstat fix and enhancement
912 Details : add a counter to set a limit to how many samples will be returned
913 fix a wildcard in the path to limit to obdfilter stats only
917 Description: lfs check only allowed for root
918 Details : Code cleanups and added a regresion test of non-root use of
923 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
924 Details : Fix hr->hr_index overload.
928 Description: Fix FIEMAP functionality.
932 Description: FC11 patchless client support added
936 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
937 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
941 Description: New acc-sm script and various test-framework enhancments.
945 Description: Previously landed patch was reverted.
949 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
953 Description: Add ldiskfs maxdirsize mount option.
957 Description: Fix mkfs.lustre support for 16TB LUNs.
961 Description: Print mmp_check_interval
962 Details : print mmp_check_interval and make it possible to abort
963 mount operation in case it takes too long.
967 Description: Add cascading_rw to lustre mpi tests.
971 Description: Store SOM data in EA.
972 Details : Also several SOM related cleanups.
976 Description: Add refcount to md_open_data
980 Description: Use refcount on lov_request_set to prevent memory corruption.
984 Description: Statahead: use parent dir's i_mutex for synchronization.
988 Description: loadgen improvements.
992 Description: Enable peer health detection in ksocklnd and ko2iblnd.
996 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
1000 Description: Fix race on obd_devcie:md_stats
1004 Description: Fix SLES 11 client builds.
1005 Details : Resolved uninitalized pointer gcc warnings.
1009 Description: Allow non-root access for "lfs check".
1010 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
1014 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
1016 Severity : enhancement
1018 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
1019 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
1023 Description: Sanity test_27p fails.
1024 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
1025 Sanity tests 27 were cleaned up and improved.
1029 Description: Unify req format on client/servers
1030 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
1031 buffers and swabbers approach..
1036 Description: drop join file stuff.
1038 Severity : enhancement
1040 Description: Add 2.6.27 and SLES11 (patchless client) support
1044 Description: Fix in ptlrpc_expire_one_request() to print the signed time
1047 Severity : enhancement
1049 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
1052 Severity : enhancement
1054 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
1056 Severity : enhancement
1058 Description: correctly handle big reply message.
1059 Details : send LNet event if reply is bigger than buffer and adjust this buffer
1064 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
1066 Severity : enhancement
1068 Description: Add support for OEL5.
1072 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
1074 Severity : enhancement
1076 Description: "weak-modules" support
1077 Details : Implement "weak-modules" support which enables kernel modules
1078 to be used with any kernel that implements the same kABI. In
1079 order to achieve this modules are now installed in
1080 /lib/modules/$(uname -r)/updates/kernel on all distributions.
1084 Description: client could not reconnect to OST because of an active request.
1085 Details : The client now retries to reconnect to the same server, if a
1086 connect request failed with EBUSY or -EAGAIN. Test case 35b
1087 conf-sanity.sh has been created.
1091 Description: Don't allow make backward step on assignin osc next id.
1092 Details : race between allocation next id and ll_sync thread can be cause
1093 of set wrong osc next id and can be kill valid ost objects.
1097 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
1101 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
1105 Description: Error handling in osc_statfs_interpret() has been improved.
1106 Details : Check in osc_statfs_interpret() for EBADR.
1110 Description: Do not send statfs() requests to OSTs disabled by administrator.
1111 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
1113 Severity : enhancement
1115 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
1117 Severity : enhancement
1119 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
1121 Severity : enhancement
1123 Description: Enhancement for lfs(1) command to use numeric uid/gid.
1127 Description: Improved cleanup procedure for conf-sanity test_46a.
1131 Description: Do not start bulk transfer, if client tried to reconnect.
1132 Details : Do not start bulk transfer, if client tried to reconnect,
1133 but failed due to active rpcs. Check for exp_abort_active_req
1134 has been implemented.
1138 Description: Prevent inconsistences between linux and lustre mount structures.
1139 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
1140 Make the sleep interruptible.
1142 Severity : enhancement
1144 Description: provide server to client comms path
1145 Details : server to client communications path via new LDLM_SET_INFO rpc
1149 Description: truncate starts GFP_FS allocation under transaction causing deadlock
1150 Details : ldiskfs_truncate calls grab_cache_page which may start page
1151 allocation under an open transaction. This may lead to
1152 calling prune_icache with consequent lustre reentrance.
1156 Description: send correctly lsm on open replay
1157 Details : MDS is trust to LSM size on replay open, but client can set wrong size
1162 Description: enable client interface failover
1163 Details : When a child reconnects from another NID, properly update export
1164 nid hash position and ldlm reverse import.
1166 Severity : enhancement
1168 Description: Communicate OST degraded/readonly state via statfs to MDS
1169 Details : Flags in the statfs returned from OSTs indicate whether the
1170 OST is in a degraded RAID state, or if the filesystem has
1171 turned read-only after a filesystem error is detected.
1176 Description: don't panic if EPROTO was hit when reading symlink
1177 Details : correctly handling request reference in error cases.
1179 Severity : enhancement
1181 Description: Add LustreNetLink, a kernel-userspace communcation path.
1183 Severity : enhancement
1185 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1190 Description: access to llog context before init.
1191 Details : move handling CATALOGS file at osc layer and forbid access to llog
1192 context before init.
1195 Frequency : rare, only when client is writing to OST on local node
1197 Description: Avoid deadlock for local client writes
1198 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1199 memory freeing context. This allows OST threads to set the
1200 PF_MEMALLOC flag on task structures in order to allocate memory
1201 from reserved pools and complete IO.
1202 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1203 so that the OST threads generate memory pressure and allow
1204 inactive pages to be reclaimed.
1206 Severity : enhancement
1208 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1213 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1214 Details : move ldlm namespace creation in setup phase to avoid grab
1215 _lprocfs_lock with cli_sem held.
1219 Description: Temporarily disable grant shrink.
1220 Details : Disable the feature for debugging.
1223 Frequency : only when formatting test filesystems
1225 Description: Unable to run several mkfs.lustre on loop devices at the same time
1226 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1227 formatting. The solution is to proper handle the error.
1229 Severity : enhancement
1231 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1233 Severity : enhancement
1235 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1240 Description: resolve race between obd_disconnect and class_disconnect_exports
1241 Details : if obd_disconnect will be called to already disconnected export he
1242 forget release one reference and osc module can't unloaded.
1244 Severity : enhancement
1246 Description: Allow tuning service thread via /proc
1247 Details : For each service a new
1248 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1249 created that can be used to set min/max thread counts, and get the
1250 current number of running threads.
1255 Descriptoin: don't return error if have partially created objects for file.
1256 Details : lov_update_create_set uses set->set_success as index for created
1257 objects, so if some requests will be failed, they will have hole
1258 at end of array and qos_shrink_lsm for allocate correct lsm.
1260 Severity : enhancement
1262 Description: Update OFED support to 1.4.1
1267 Description: Reduce small size read RPC
1268 Details : Set read-ahead limit for every file and only do read-ahead when
1269 available read-ahead pages are bigger than 1M to avoid small size
1272 Severity : enhancement
1274 Description: move AT tunable parameters for more consistent usage
1275 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1277 Severity : enhancement
1279 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1280 Details : allow skip disconnected ost for send statfs request and hide error
1286 Description: fix racy locking of mballoc block bitmaps causing BUG
1287 Details : The locking of the mballoc buddy bitmap and the in-memory
1288 block bitmap was using two different spin locks in some
1289 cases. This made it possible to incorrectly access the
1290 mballoc bitmap while another process was modifying it,
1291 causing a sanity assertion to fail. While no on-disk corruption
1292 was reported, there was some risk of this happening.
1295 Frequency : rare, on llog test 6
1297 Descriptoin: don't allow connect to already connected import
1298 Details : allowing connect to already connected import is hide connecting problem.
1303 Descriptoin: allow kill process which wait statahead result
1304 Details : for some reasons 'ls' can stick in waiting result from statahead,
1305 in this case need way for kill this process.
1307 Severity : enhancement
1309 Description: Add state history info file, enhance import info file
1310 Details : Track import connection state changes in a new osc/mdc proc file;
1311 add overview-type data to the osc/mdc import proc file.
1313 Severity : enhancement
1315 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1317 Severity : enhancement
1319 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1324 Descriptoin: OST grant too much space to client even there are not enough space.
1325 Details : Client will shrink its grant cache to OST if there are no write
1326 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1327 this grant cache if there are already not enough avaible space
1328 (left_space < total_clients * 32M).
1331 Frequency : start MDS on uncleanly shutdowned MDS device
1333 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1334 Details : stay in waiting mds<>ost recovery finished produce random bugs
1335 due race between two ll_sync thread for one lov target. send
1336 ACTIVATE event only if connect realy finished and import have
1340 Frequency : rare, connect and disconnect target at same time
1342 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1343 Details : don't call obd_disconnect under lov_lock. this long time
1344 operation and can block ptlrpcd which answer to connect request.
1349 Descriptoin: don't lose wakeup for imp_recovery_waitq
1350 Details : recover_import_no_retry or invalidate_import and import_close can
1351 both sleep on imp_recovery_waitq, but we was send only one wakeup
1355 Frequency : always with long access acl
1357 Descriptoin: mds can't pack reply with long acl.
1358 Details : mds don't control size of acl but they limited by reint/getattr
1361 Severity : enhancement
1363 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1365 Severity : enhancement
1367 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1370 Frequency : start MDS on uncleanly shutdowned MDS device
1372 Descriptoin: aborting recovery hang on MDS
1373 Details : don't throttle destroy RPCs for the MDT.
1376 Frequency : on remount
1378 Description: external journal device not working after the remount
1379 Details : clear dev_rdonly flag for external journal devices in
1385 Description: shutdown vs evict race
1386 Details : client_disconnect_export vs connect request race.
1387 if client will evicted at this time - we start invalidate
1388 thread without referece to import and import can be freed
1392 Frequency : rare, need acl's on inode.
1394 Description: client can't handle ost additional correctly
1395 Details : if ost was added after client connected to mds client can have
1396 hit lnet_try_match_md ... to big messages to wide striped files.
1397 in this case need teach client to handle config events about add
1398 lov target and update client max ea size at that event.
1400 Severity : enhancement
1402 Description: Changelogs
1403 Details : Changelogs are a lightweight mechanism to track filesystem
1404 metadata and namespace changes. The changelog is recorded
1405 permanently on the MDTs, and is periodically "consumed" / purged
1406 when records are no longer needed.
1408 Severity : enhancement
1410 Description: compact fld format with extents
1411 Details : Store range of seq rather than every seq in FLD. Seq
1412 controller update FLD rather than clients. In Case of CMD, mdt0
1413 has FLD, all other metadata server act as non persistent proxy
1414 for FLD queries and cache fld entries in fld cache.
1419 Description: don't skip ost target if they assigned to file
1420 Details : Drop slow OSCs if we can, but not for requested start idx.
1421 This means "if OSC is slow and it is not the requested
1422 start OST, then it can be skipped, otherwise skip it only
1423 if it is inactive/recovering/out-of-space.
1427 Description: more cleanup in mds_lov
1428 Details : not send LOV EA under replay, we can't know about they size at this
1429 time. Don't allow client connect to mds before any ost connected,
1430 for avoid problems with LOV EA size and returning EIO to client.
1432 Severity : enhancement
1434 Description: Interoperability at server side (Disk interoperability)
1436 Severity : enhancement
1438 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1440 Severity : enhancement
1442 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1444 Severity : enhancement
1446 Description: New client IO stack (CLIO).
1448 Severity : enhancement
1450 Description: Commit on sharing. Eliminate inter-client dependencies between
1451 uncommitted transactions by doing transaction commits.
1452 Thereby clients may recovery independently.
1455 Frequency : Create a symlink file with a very long name
1457 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1458 Details : If there is no extra space in the request for early cancels,
1459 ldlm_req_handles_avail() returns 0 instead of a negative value.
1461 Severity : enhancement
1463 Description: Add /proc entry for import status
1464 Details : The mdc, osc, and mgc import directories now have
1465 an import directory that contains useful import data for debugging
1466 connection problems.
1468 Severity : enhancement
1470 Description: Re-disable certain /proc logging
1471 Details : Enable and disable client's offset_stats, extents_stats and
1472 extents_stats_per_process stats logging on the fly.
1475 Frequency : Only on FC kernels 2.6.22+
1477 Description: oops in statahead
1478 Details : Do not drop reference count for the dentry from VFS when lookup,
1479 VFS will do that by itself.
1481 Severity : enhancement
1483 Description: Generic /proc file permissions
1484 Details : Set /Proc file permissions in a more generic way to enable non-
1485 root users operate on some /proc files.
1489 Description: Hitting mdc_commit_close() ASSERTION
1490 Details : Properly handle request reference release in
1491 ll_release_openhandle().
1495 Description: quota recovery deadlock during mds failover
1496 Details : This patch includes att18982, att18236, att18237 in bz14840.
1498 1. fix osts hang when mds does failover with quotaon
1499 2. prevent watchdog storm when osts threads wait for the
1504 Frequency : only patchless client
1505 Description: add workaround for race between add/remove dentry from hash
1507 Severity : enhancement
1509 Description: Allow OST glimpses to return PW locks
1513 Description: LBUG when llog conf file is full
1514 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1518 Description: Prevent import from entering FULL state when server in recovery
1522 Description: service mount cannot take device name with ":"
1523 Details : Only when device name contains ":/" will mount treat it as
1529 Description: replace ptlrpcd with the statahead thread to interpret the async
1530 statahead RPC callback
1534 Frequency : on recovery
1535 Description: I/O failures after umount during fail back
1536 Details : if client reconnected to restarted server we need join to recovery
1537 instead of find server handler is changed and process self
1538 eviction with cancel all locks.
1540 Severity : enhancement
1542 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1544 Severity : enhancement
1546 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1548 Severity : enhancement
1550 Description: Upcall on Lustre log has been dumped
1551 Details : Allow for a user mode script to be called once a Lustre log has
1552 been dumped. It passes the filename of the dumped log to the
1553 script, the location of the script can be specified via
1554 /proc/sys/lnet/debug_log_upcall.
1559 Description: avoid idr_remove called for id which is not allocated.
1560 Details : Move assigment s_dev for clustered nfs to end of initialization,
1561 to avoid problem with error handling.
1566 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1567 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1568 into CONN_USED_HASH and this prodice warning when put connection
1569 again in unused hash.
1572 Severity : enhancement
1574 Description: File striping can now be set to use an arbitrary pool of OSTs.
1576 Severity : enhancement
1578 Description: Export bytes_read/bytes_write count on OSC/OST.
1582 Description: Early reply size mismatch, MGC loses connection
1583 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1584 the connect flags are properly negotiated.
1589 Description: quotas are not honored with O_DIRECT
1590 Details : all writes with the flag O_DIRECT will use grants which leads to
1591 this problem. Now using OBD_BRW_SYNC to guard this.
1595 Description: add quota statistics
1596 Details : 1. sort out quota proc entries and proc code.
1597 2. add quota statistics
1599 Severity : enhancement
1601 Description: enable quota support for HEAD.
1605 Description: Properly propagate oinfo flags from lov to osc for statfs
1606 Details : restore missing copy oi_flags to lov requests.
1608 Severity : enhancement
1610 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1611 : l_getgroups(8), lst(8), routerstat(8)
1612 Details : included man pages for llobdstat(8), llstat(8),
1613 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1615 Severity : enhancement
1617 Description: configure's --enable-quota should check the
1618 : kernel .config for CONFIG_QUOTA
1619 Details : configure is terminated if --enable-quota is passed but
1620 : no quota support is in kernel
1624 Description: Remove portals compatibility
1625 Details : Remove portals compatibility, not interoperable with releases
1630 Description: Resolve device initialization race
1631 Details : Prevent proc handler from accessing devices added to the
1632 obd_devs array but yet be intialized.
1634 Severity : enhancement
1636 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1638 Severity : enhancement
1640 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1645 Description: Using wrong pointer in osc_brw_prep_request
1646 Details : Access to array[-1] can produce panic if kernel compiled with
1647 CONFIG_PAGE_ALLOC enabled
1651 Description: Client runs out of low memory
1652 Details : Consider only lowmem when counting initial number of llap pages
1656 Description: *optional* service tags registration
1657 Details : if the "service tags" package is installed on a Lustre node
1658 When the filesystem is mounted, a local-node service tag will
1659 be created. See http://inventory.sun.com/ for more information
1660 about the Service Tags asset management system.
1664 Description: Kernel BUG tries to release flock
1665 Details : Lustre does not destroy flock lock before last reference goes
1666 away. So always drop flock locks when client is evicted and
1667 perform unlock regardless of successfulness of speaking to MDS.
1671 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1675 Description: sanity test 65a fails if stripecount of -1 is set
1676 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1681 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1682 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1685 Severity : enhancement
1687 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1691 Description: do not process already freed flock
1692 Details : flock can possibly be freed by another thread before it reaches
1693 to ldlm_flock_completion_ast.
1697 Description: LBUG during stress test
1698 Details : Need properly lock accesses the flock deadlock detection list.
1702 Description: oops in page fault handler
1703 Details : kernel page fault handler can return two special 'pages' in error
1704 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1708 Description: timeout with invalidate import.
1709 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1710 should be handled by ptlrpcd. This produce long age waiting and
1711 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1713 Severity : enhancement
1715 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1719 Description: enable MGS and MDT services start separately
1720 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1721 a co-located MGS without starting the MGS, which is a complement
1722 to 'nosvc' mount option.
1726 Frequency : after recovery
1727 Description: precreate to many object's after del orphan.
1728 Details : del orphan st in oscc last_id == next_id and this triger growing
1729 count of precreated objects. Set flag LOW to skip increase count
1730 of precreated objects.
1734 Frequency : rare, on clear nid stats
1735 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1736 Details : when clean nid stats sometimes try destroy live entry,
1737 and this produce panic in free.
1741 Description: Stack overflow during MDS log replay
1742 ease stack pressure by using a thread dealing llog_process.
1746 Description: wait until IO finished before start new when do lock cancel.
1747 Details : VM protocol want old IO finished before start new, in this case
1748 need wait until PG_writeback is cleared until check dirty flag and
1749 call writepages in lock cancel callback.
1751 Severity : enhancement
1753 Description: using special macro for print time and cleanup in includes.
1757 Description: mds_mfd_close() ASSERTION(rc == 0)
1758 Details : In mds_mfd_close(), we need protect inode's writecount change
1759 within its orphan write semaphore to prevent possible races.
1763 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1767 Frequency : rare, on shutdown ost
1768 Description: don't hit live lock with umount ost.
1769 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1770 use shrink_dcache_sb instead.
1774 Description: don't panic with use echo client
1775 Details : echo client pass NULL as client nid pointer and this produce null
1776 pointer dereference.
1780 Description: fix build on ppc32
1781 Details : compile code with -m64 flag produce wrong object file for ppc32.
1785 Description: add message levels for liblustreapi
1789 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1790 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1791 will result in an -EAGAIN and in the exit call path we call
1792 alloc_rr() which will with increasing aggressiveness attempt to
1793 aquire precreated objects on the minimum number of required OSCs.
1797 Description: Use old size assignment to avoid deadlock
1798 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1799 the scheduling while holding a spinlock. We do not need locking
1800 for size in ll_update_inode() because size is only updated from
1801 the MDS for directories or files without objects, so there is no
1802 other place to do the update, and concurrent access to such inodes
1803 are protected by the inode lock.
1807 Description: resolve "_IOWR redefined" build error on SLES10
1811 Description: dump the memory debugging after all modules are unloaded to
1812 suppress false negative in conf_sanity test 39
1814 Severity : enhancement
1816 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1821 Description: fix SLES kernel versioning
1822 Details : the kernel version for our SLES 10 kernel did not include a "-"
1823 before the "smp" at the end. while this was not a problem in
1824 general, it did mean that software trying to use the kernel
1825 version to try to detect a vendor specific kernel would fail.
1826 this was most evident by the OFED build scripts.
1830 Description: Don't update lov_desc members until making sure they are valid
1831 Details : When updating lov_desc members via proc fs, need fix their
1832 validities before doing the real update.
1836 Description: don't put request into delay list while invalidate in flight.
1837 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1838 import in flight. this produce timeout for invalidate and sometimes
1839 can cause stale data.
1843 Frequency : on ppc only
1844 Description: not convert ost objects for directory because it's not exist.
1845 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1849 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1851 Severity : enhancement
1853 Description: Update to support 2.6.22.14 vanilla kernel.
1857 Frequency : rare, on recovery
1858 Description: read procfs can produce deadlock in some situation
1859 Details : Holding lprocfs lock which send rpc can produce block for destroy
1860 obd objects and this also block reconnect with -EALREADY. This
1861 isn't fix all lprocfs bugs - but make it rare.
1863 Severity : enhancement
1865 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1868 Frequency : frequent on X2 node
1870 Description: mdc_set_open_replay_data LBUG
1871 Details : Set replay data for requests that are eligible for replay.
1875 Description: lustre_mgs: operation 101 on unconnected MGS
1876 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1877 MGC, and late on MGC cannot successfully connect to MGS and a lot
1878 of the error messages complaining that MGS is not connected.
1882 Frequency : on network error
1883 Description: panic with double free request if network error
1884 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1885 but it's true only for synchronus enqueue, for async enqueue
1886 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1889 Severity : enhancement
1891 Description: client-side metadata stat-ahead during readdir(directory readahead)
1892 Details : perform client-side metadata stat-ahead when the client detects
1893 readdir and sequential stat of dir entries therein
1896 Frequency : on start mds
1898 Description: Implement get_info(last_id) in obdfilter.
1901 Frequency : occasional
1903 Description: Correctly check stale fid, not start epoch if ost not support SOM
1904 Details : open with flag O_CREATE need set old fid in op_fid3 because
1905 op_fid2 was overwritten with new generated fid, but mds can answer
1906 with one of these two fids and both is not stale. Setattr
1907 incorrectly started an epoch and assume will be called
1908 done_writing, but without SOM done_writing ever being called.
1911 Frequency : rare, depends on device drivers and load
1913 Description: MDS or OSS nodes crash due to stack overflow
1914 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1915 In some cases, in conjunction with device drivers that use a lot
1916 of stack the MDS (or possibly OSS) service threads could overflow
1917 the stack. One change which was identified to consume additional
1918 stack has been reworked to avoid the extra stack usage.
1921 Frequency : occasional
1923 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1924 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1925 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1926 oscc_has_objects and see OSC has no objests and call
1927 oscc_internal_create to resend create request.
1929 Severity : enhancement
1931 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1933 Severity : enhancement
1935 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1938 Frequency : very rare
1940 Description: Fix replay if there is an un-replied request and open
1941 Details : In some cases, older replay request will revert the
1942 mcd->mcd_last_xid on MDS which is used to record the client's
1943 latest sent request.
1945 Severity : enhancement
1947 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1949 Severity : enhancement
1951 Description: Add rhel5 support to HEAD.
1953 Serverity : enhancement
1955 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1960 Description: Don't allow skipping OSTs if index has been specified.
1961 Details : Don't allow skipping OSTs if index has been specified, make
1962 locking in internal create lots better.
1966 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1967 Details : access to bitfield in structure is always rounded to long
1968 and this produce problem with not atomic change any bit.
1972 Description: Lustre make rpms failed.
1973 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1974 builds Lustre rpms from tarball.
1977 Frequency : rare on shutdown ost
1979 Description: If llog cancel was not send before clean_exports phase, this can
1980 produce deadlock in llog code.
1981 Details : If llog thread has last reference to obd and call class_import_put
1982 this produce deadlock because llog_cleanup_commit_master wait when
1983 last llog_commit_thread exited, but this never success because was
1984 called from llog_commit_thread.
1988 Description: allow userland application know is lost one of stripes.
1989 Details : fill lvb_blocks with error code on ost and return it to
1990 application if error flag found.
1994 Description: NULL lov_tgts causing MDS oops
1995 Details : more safe checks for NULL lov_tgts for avoid oops.
1997 Severity : enhancement
1999 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
2003 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
2004 Details : If an OST is down the MDS will hang indefinitely in
2005 obd_statfs_rqset() waiting for the statfs data. While for
2006 MDS QOS usage of statfs, it should not stuck in waiting.
2008 Severity : enhancement
2010 Description: remote_acl support
2011 Details : Support ACL-based permission check for remote user.
2012 Support setfacl/getfacl for remote user with the utils
2013 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
2014 the system "{s,g}etfacl" utils.
2016 Severity : enhancement
2018 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
2020 Severity : enhancement
2022 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
2026 Description: make mgs_setparam() handle fsname containing dash
2027 Details : fsname containing a dash does not work with lctl conf_param
2029 Severity : enhancement
2031 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
2033 Severity : enhancement
2035 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
2039 Frequency : rare, at shutdown
2040 Description: access already free / zero obd_namespace.
2041 Details : if client_disconnect_export was called without force flag set,
2042 and exist connect request in flight, this can produce access to
2043 NULL pointer (or already free pointer) when connect_interpret
2044 store ocd flags in obd_namespace.
2048 Frequency : only at startup
2049 Description: not alloc memory with spinlock held.
2050 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
2051 if any spinlock held.
2053 Severity : enhancement
2055 Description: make lustre randomly fail allocating memory
2056 Details : Make lustre randomly failed allocating memory for testing purpose.
2058 Severity : enhancement
2060 Description: lost problems with lov objid file
2061 Details : Fixes some scability and access to not inited memory problems
2062 in work with lov objdid file.
2067 Description: lfs find does not continue on file error
2068 Details : Continue other files processing when a file/dir is absent.
2072 Description: Inconsistent usage of lustre_pack_reply()
2073 Details : Standardize the usage of lustre_pack_reply() such that it
2074 always generate a CERROR on failure.
2077 Frequency : occasional
2079 Description: MDS hang or stay in waiting lock
2080 Details : If client receive lock with CBPENDING flag ldlm need send lock
2081 cancel as separate rpc, to avoid situation when cancel request
2082 can't processed due all i/o threads stay in wait lock.
2086 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
2090 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
2094 Description: Update to SLES9 kernel-2.6.5-7.287.3.
2098 Description: Update to RHEL4 latest kernel.
2100 Severity : enhancement
2102 Description: Build SLES10 patchless client fails
2103 Details : The configure was broken by run ./configure with
2104 --with-linux-obj=.... argument for patchless client. When the
2105 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
2106 can't search header adequately. Use absolute path such as
2107 -I($LINUX)/include instead.
2111 Description: interrupt oig_wait produce painc on resend.
2112 Details : brw_redo_request can be used for resend requests from ptlrpcd and
2113 private set, and this produce situation when rq_ptlrpcd_data not
2114 copyed to new allocated request and triggered LBUG on assert
2115 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
2116 ptlrpcd set if request is changed and can be safety changed to use
2121 Description: LASSERT_{REQ,REP}SWAB macros are buggy
2122 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
2123 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
2124 these macros and replace them with their difinitions instead.
2128 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
2129 Details : Update which_patch & target file for SLES10 latest kernel.
2132 Description: Remove client patches from SLES 10 kernel.
2133 Details : This causes SLES 10 clients to behave as patchless clients
2134 even on a Lustre-patched (server) kernel.
2136 Severity : enhancement
2138 Description: self-adjustable client's lru lists
2139 Details : use adaptive algorithm for managing client cached locks lru
2140 lists according to current server load, other client's work
2141 pattern, memory activities, etc. Both, server and client
2142 side namespaces provide number of proc tunables for controlling
2145 Severity : enhancement
2147 Description: light-weight GSS support
2148 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
2149 and reduce performance overhead.
2151 Severity : enhancement
2153 Description: Linux keyring support
2154 Details : Support using service of Linux keyring for Lustre GSS internal
2155 context refresh/cache mechanism.
2159 Description: Fix errors in lfs documentation
2160 Details : Fixes man pages
2164 Description: don't use GFP_* in generic Lustre code.
2165 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2169 Description: obdclass is limited by single OBD_ALLOC(idarray)
2170 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2174 Description: warnings with build patchless client with vanila 2.6.19 and up
2175 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2176 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2180 Frequency : Only for SLES
2181 Description: sanity_quota fail test_1
2182 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2183 file, which makes TSTID[2] inunique.
2187 Frequency : Only for SLES9
2188 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2189 Details : Update target/ChangeLog/which_patch .
2193 Frequency : Only for test_10 in sanity-quota.sh
2194 Description: a bug in quota test code
2195 Details : checking a test flag in a wrong place
2199 Frequency : Only for test_13 in sanity-quota.sh
2200 Description: a bug in quota test script
2201 Details : shouldn't check overfull allocated quota
2205 Description: jbd statistics
2206 Details : Port older jbd statistics patch for sles10
2208 Severity : enhancement
2210 Description: Add printing inode info into message about error in writepage.
2215 Description: reply_lock_interpret crash due to race with it and lock cancel.
2216 Details : Do not replay locks that are being cancelled. Do not reference
2217 locks by their address during replay, just by their handle.
2221 Frequency : When flocks are used.
2222 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2223 Details : Flock locks might destroy just granted lock if it could be merged
2224 with another existing flock, this is done in completion handler,
2225 so teach ldlm_cli_enquque_fini that this is a valid case for
2231 Description: Oops in read and write path when failing to allocate lock.
2232 Details : Check if lock allocation failed and return error back.
2236 Description: lstripe command fails for valid OST index
2237 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2238 instead of lov->desc.ld_active_tgt_count.
2242 Description: sanity.sh failed test 103
2243 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2248 Description: df doesn't work properly if diskfs blocksize != 4K
2249 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2253 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2254 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2256 Severity : enhancement
2258 Description: Adaptive timeouts
2259 Details : RPC timeouts adapt to changing server load and network
2260 conditions to reduce resend attempts and improve recovery time.
2264 Description: llapi_file_create() does not allow some changes
2265 Details : add llapi_file_open() that allows specifying the mode and
2266 open flags, and also returns an open file handle.
2270 Description: merge and cleanup kernel patches.
2271 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2275 Description: Add journal checksum support.(Kernel part)
2276 Details : The journal checksum feature adds two new flags i.e
2277 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2278 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2279 indicates that the commit block contains the checksum for
2280 the blocks described by the descriptor blocks. Now commit
2281 record can be sent to disk without waiting for descriptor
2282 blocks to be written to disk. This behavior is controlled
2283 using JBD2_FEATURE_ASYNC_COMMIT flag.
2287 Description: OSS needs mutliple precreate threads
2288 Details : Add ability to start more than one create thread per OSS.
2292 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2293 Details : execute lfs setstripe on client
2297 Description: mds_obd_create error creating tmp object
2298 Details : When the user sets quota on root, llog will be affected and can't
2299 create files and write files.
2302 Frequency : Always on ia64 patchless client, and possibly others.
2304 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2305 Details : This allows the patchless client to be loaded on architectures
2306 without this export.
2310 Description: RedHat Update kernel for RHEL5
2311 Details : Add the kernel config file for RHEL5.
2315 Description: RedHat Update kernel for RHEL5
2316 Details : Modify the kernel config file more closer RHEL5.
2320 Description: Build failure against Centos5 (RHEL5)
2321 Details : Define PAGE_SIZE when it isn't present.
2325 Description: test_7 on sanity-quota.sh can't be used on separate machines
2326 Details : add facet to handle it
2330 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2331 Details : with NFS, the anon dentry's parent was set to itself in
2332 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2333 corresponding dentry other than use rec->ur_name.
2335 Severity : enhancement
2337 Description: lfs setstripe enhancement
2338 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2342 Description: enable data checksumming by default
2343 Details : enable checksum by default, allow --disable-checksum
2344 configure option and "-o nochecksum" mount option. Checksums
2345 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2349 Description: lustre support for RHEL5
2350 Details : Add support for RHEL5.
2354 Description: Client eviction due to failover config
2355 Details : after a connection loss, the lustre client should attempt to
2356 reconnect to the last active server first before trying the
2357 other potential connections.
2361 Description: when mds and osts use different quota unit(32bit and 64bit),
2362 quota will be released repeatly.
2363 Details : void sending multiple quota reqs to mds, which will keep the
2364 status between the reqs.
2368 Description: osts not allocated evenly to files
2369 Details : change the condition to increase offset_idx
2372 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2374 Description: Data corruption for OSTs that are formatted larger than 2TB
2376 Details : When generating the bio request for lustre file writes the
2377 sector number would overflow a temporary variable before being
2378 used for the IO. The data reads correctly from Lustre (which
2379 will overflow in a similar manner) but other file data or
2380 filesystem metadata may be corrupted in some cases.
2384 Description: Tune the kernel for good SCSI performance.
2385 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2386 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2391 Description: rewrite ext2-derived code in obdclass/uuid.c
2392 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2393 from scratch preserving functionality.
2397 Description: "lfs find -obd UUID" prints directories
2398 Details : "lfs find -obd UUID" will return all directory names instead
2399 of just file names. It is incorrect because the directories
2400 do not reside on the OSTs.
2404 Description: use i_size_read and i_size_write in 2.6 port
2405 Details : replace inode->i_size access with i_size_read/write()
2408 Frequency : when removing large files
2410 Description: scheduling issue during removal of large Lustre files
2411 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2412 It causes scheduling issues when removing large files (17TB in the
2416 Frequency : only with liblustre clients on XT3
2418 Description: evictions taking too long
2419 Details : allow llrd to evict clients directly on OSTs
2422 Frequency : only on ppc
2424 Description: /proc/fs/lustre/devices broken on ppc
2425 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2426 We should make sure the type of 'index' is loff_t and then cast
2427 explicitly as needed below. Do not assign an explicitly cast
2431 Frequency : only for rhel5
2433 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2434 Details : Modify the target file & which_kernel.
2436 Severity : enhancement
2438 Description: omit set fsid for export NFS
2439 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2440 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2444 Frequency : Always, for kernels after 2.6.16
2445 Description: Fix warning idr_remove called for id=.. which is not allocated.
2446 Details : Last kernels save old s_dev before kill super and not allow
2447 to restore from callback - restore it before call kill_anon_super.
2451 Description: buffer overruns could theoretically occur
2452 Details : llapi_semantic_traverse() modifies the "path" argument by
2453 appending values to the end of the origin string, and a
2454 overrun may occur. Adding buffer overrun check in liblustreapi.
2458 Description: Fix error on 'ls .' at the top of the Lustre mount.
2459 Details : Don't revalidate dentry if it is a root dentry.
2463 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2464 Details : Modify vm-tunables-rhel4.patch.
2468 Description: Kernel config for 2.6.18-vanilla.
2469 Details : Modify targets/2.6-vanilla.target.in.
2470 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2471 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2472 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2473 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2477 Description: improve handling recoverable errors
2478 Details : if request processig with error which can be recoverable on server
2479 request should be resend, otherwise page released from cache and
2484 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2485 Details : Modify target file & which_patch.
2486 A flaw was found in the IA32 system call emulation provided
2487 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2488 value could be stored in the %RAX register, which could trigger an
2489 out-of-bounds system call table access. An untrusted local user
2490 could exploit this flaw to run code in the kernel
2491 (ie a root privilege escalation). (CVE-2007-4573).
2495 Description: O_DIRECT bypasses client statistics.
2496 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2497 read_ahead_stats were updated. copied the stats section from
2498 osc_send_oap_rpc() into async_internal().
2502 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2506 Description: change order of libsysio includes
2507 Details : '#include sysio.h' should always come before '#include xtio.h'
2511 Description: handle "serious error: objid * already exists" more gracefully
2512 Details : If LAST_ID value on disk is smaller than the objects existing in
2513 the O/0/d* directories, it indicates disk corruption and causes an
2514 LBUG(). If the object is 0-length, then we should use the existing
2515 object. This will help to avoid a full fsck in most cases.
2517 Severity : enhancement
2519 Description: adapt the lustre_config script to support the upgrade case
2520 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2521 server targets to 1.6 in parallel.
2525 Description: To avoid grant space > avaible space when the disk is almost
2526 full. Without this patch you might see the error "grant XXXX >
2527 available" or some LBUG about grant, when the disk is almost
2529 Details : In filter_check_grant, for non_grant cache write, we should
2530 check the left space by if (*left > ungranted + bytes), instead
2531 of (*left > ungranted), because only we are sure the left space
2532 is enough for another "bytes", then the ungrant space should be
2533 increase. In client, we should update cl_avail_grant only there
2534 is OBD_MD_FLGRANT in the reply.
2538 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2539 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2540 described in CVE-2007-4573. This problem could allow untrusted
2541 local users to gain root access.
2544 Frequency : when using O_DIRECT and quotas
2546 Description: Incorrect file ownership on O_DIRECT output files
2547 Details : block usage reported by 'lfs quota' does not take into account
2548 files that have been written with O_DIRECT.
2553 Description: touch file failed when fs is not full
2554 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2555 otherwise we can get ENOSP while fs is not full.
2559 Description: parallel lock callbacks
2560 Details : Instead of sending blocking and completion callbacks as separated
2561 requests, adding them to a set and sending in parallel.
2564 Frequency : only for Cray XT3
2565 Bugzilla : 12829/13455
2566 Description: Changing primary group doesn't change the group lustre assigns to
2568 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2569 overridden with the primary group provided by the group upcall,
2570 whereas the supplied fsgid can be trusted if it is in the list of
2571 supplementary groups returned by the group upcall.
2573 Severity : enhancement
2575 Description: Allow masking D_WARNING, D_ERROR messages from console
2576 Details : Console messages can now be disabled via lnet.printk.
2580 Description: User code with malformed file open parameter crashes client node
2581 Details : Before packing join_file req, all the related reference should be
2582 checked carefully in case some malformed flags cause fake
2583 join_file req on client.
2587 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2588 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2589 mode is equal to the granted mode. However, it should also check
2590 that LDLM_FL_CP_REQD is not set.
2594 Description: Heavy nfs access might result in deadlocks
2595 Details : After ELC code landed, it is now improper to enqueue any mds
2596 locks under och_sem, because enqueue might want to decide to
2597 cancel open locks for same inode we are holding och_sem for.
2601 Description: Client eviction while running blogbench
2602 Details : A lot of unlink operations with concurrent I/O can lead to a
2603 deadlock causing evictions. To address the problem, the number of
2604 oustanding OST_DESTROY requests is now throttled to
2605 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2606 callbacks are processed in priority.
2610 Description: enable ACLs on MDS by default
2611 Details : ACLs must be enabled on MDS by default.
2614 Frequency : PPC/PPC64 only
2616 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2617 Details : fix duplicated definitions between asm-ppc64/types.h and
2618 lustre_types.h on PPC.
2621 Frequency : PPC/PPC64 only
2623 Description: asm-ppc/segment.h does not exist
2624 Details : fix compile issue on PPC.
2628 Description: better handle error messages in extents code
2631 Frequency : RHEL4 only
2633 Description: mkfs is very slow on IA64/RHEL4
2634 Details : A performance regression has been discovered in the MPT Fusion
2635 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2636 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2637 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2639 Severity : enhancement
2641 Description: SNMP support enhancement
2642 Details : Adding total number of sampled request for an MDS node in snmp
2645 Severity : enhancement
2647 Description: Optimize ldlm waiting list processing for PR extent locks
2648 Details : When processing waiting list for read extent lock and meeting read
2649 lock that is same or wider to it that is not contended, skip
2650 processing rest of the list and immediatelly return current
2651 status of conflictness, since we are guaranteed there are no
2652 conflicting locks in the rest of the list.
2656 Description: Time out and refuse to reconnect
2657 Details : When the failover node is the primary node, it is possible
2658 to have two identical connections in imp_conn_list. We must
2659 compare not conn's pointers but NIDs, otherwise we can defeat
2660 connection throttling.
2664 Description: port llog fixes from b1_6 into HEAD
2665 Details : Port llog reference couting and some llog cleanups from b1_6
2666 (bug 10800) into HEAD, for protect from panic and access to already
2667 free llog structures.
2671 Description: Detect stride IO mode in read-ahead
2672 Details : When a client does stride read, read-ahead should detect that and
2673 read-ahead pages according to the detected stride pattern.
2677 Description: data checksumming impacts single node performance
2678 Details : add support for several checksum algorithm. Currently, only CRC32
2679 and Adler-32 are supported. The checksum type can be changed on
2680 the fly via /proc/fs/lustre/osc/*/checksum_type.
2684 Description: use adler32 for page checksums
2685 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2690 Description: build for x2 fails
2691 Details : fix compile issue on Cray systems.
2695 Description: Properly match for duplicate locks
2696 Details : Due to different lock order from skiplists code, we need to
2697 traverse entire list for now
2700 Frequency : only on PPC/SLES10
2702 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2703 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2704 be defined. Add a hack in mkfs_lustre.c to work around this compile
2709 Description: LASSERT on MDS when client holding flock lock dies
2710 Details : ldlm pool logic depends on number of granted locks equal to
2711 number of released locks which is not true for flock locks, so
2712 just exclude such locks from consideration.
2716 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2717 Details : Use fsfilt_sync() for both the whole filesystem sync and
2718 individual file sync to eliminate dangerous inode locking
2719 with I_LOCK that can lead to a deadlock.
2723 Description: performance in 1.6.3
2724 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2725 sure that 1MB requests are not fragmented by the block layer.
2727 Severity : enhancement
2729 Description: organize the server-side client stats on per-nid basis
2730 Details : Change the structure of stats under obdfilter and mds to
2738 The "uuid"s file would list the uuids of _active_ exports.
2739 And the clear entry is to clear all stats and stale nids.
2741 Severity : enhancement
2743 Description: eliminate client locks in face of contention
2744 Details : file contention detection and lockless i/o implementation
2745 for contended files.
2749 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2752 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2754 Description: MDS may lose file striping (and hence file data) in some cases
2755 Details : If there are additional extended attributes stored on the MDS,
2756 in particular ACLs, SELinux, or user attributes (if user_xattr
2757 is specified for the client mount options) then there is a risk
2758 of attribute loss. Additionally, the Lustre file striping
2759 needs to be larger than default (e.g. striped over all OSTs),
2760 and an additional attribute must be stored initially in the
2761 inode and then increase in size enough to be moved to the
2762 external attribute block (e.g. ACL growing in size) for file
2767 Description: skiplist implementation simplification
2768 Details : skiplists are used to group compatible locks on granted list
2769 that was implemented as tracking first and last lock of each
2770 lock group the patch changes that to using doubly linked lists
2774 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2775 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2776 because LASSERT on both the data supplied by a client, and the
2777 data on disk is dangerous and incorrect.
2779 Severity : enhancement
2781 Description: Slow truncate/writes to huge files at high offsets.
2782 Details : Directly associate cached pages to lock that protect those pages,
2783 this allows us to quickly find what pages to write and remove
2784 once lock callback is received.
2788 Description: more ldlm soft lockups
2789 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2790 starve other threads from the resource lock for a long time in
2791 case of long waiting queue, so change the debug level from
2792 D_OTHER to the less frequently used D_INFO.
2794 Severity : enhancement
2796 Description: add -gid, -group, -uid, -user options to lfs find
2800 Description: Hung threads in invalidate_inode_pages2_range
2801 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2802 once one is completed. As a result, some RPCs are stuck in the
2803 queue and are never sent.
2807 Description: filter threads hungs on waiting journal commit
2808 Details : Cleanup filter group llog code, then only filter group llog will
2809 be only created in the MDS/OST syncing process.
2813 Description: Procfs and llog threads access destoryed import sometimes.
2814 Details : Sync the import destoryed process with procfs and llog threads by
2815 the import refcount and semaphore.
2817 Severity : enhancement
2819 Description: openlock cache of b1_6 port to HEAD
2824 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2825 Details : If insertion of an extent fails, then discard the inode
2826 preallocation and free data blocks else it can lead to duplicate
2831 Description: don't always update ctime in ext3_xattr_set_handle()
2832 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2833 In some cases the ctime should not be updated, for example for
2834 2.0->1.8 compatibility it is necessary to delete an xattr and it
2835 should not update the ctime.
2839 Bugzilla : 15713/16362
2840 Description: Assertion in iopen_connect_dentry in 1.6.3
2841 Details : looking up an inode via iopen with the wrong generation number can
2842 populate the dcache with a disconneced dentry while the inode
2843 number is in the process of being reallocated. This causes an
2844 assertion failure in iopen since the inode's dentry list contains
2845 both a connected and disconnected dentry.
2849 Description: assertion failure in ldlm_handle2lock()
2850 Details : fix a race between class_handle_unhash() and class_handle2object()
2851 introduced in lustre 1.6.5 by bug 13622.
2856 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2857 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2862 Description: oops in sd_iostats_seq_show()
2863 Details : unloading/reloading the scsi low level driver triggers a kernel
2864 bug when trying to access the sd iostat file.
2869 Description: Kernel panics during QLogic driver reload
2870 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2871 patch, causing memory corruption.
2876 Description: journal_dev option does not work in b1_6
2877 Details : pass mount option during pre-mount.
2879 Severity : enhancement
2881 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2882 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2883 extent information of a file. It can be used to map logical blocks
2884 in a file to physical blocks in the block device.
2888 Description: LDLM soft lockups - improvement
2889 Details : It is be possible to send the lock handle along with each read
2890 or write request because the client is already doing a lock match
2891 itself so there isn't any reason the OST should have to re-do that
2895 Frequency : only on Cray X2
2897 Description: X2 build failures
2898 Details : fix build failures on Cray X2.
2902 Description: xid & resent requests
2903 Details : Initialize RPC XID from clock at startup (randomly if clock is
2906 Severity : enhancement
2908 Description: Add lustre_start utility to start or stop multiple Lustre servers
2913 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2914 Details : In case of memory pressure, list_del() can be called twice on
2915 req->rq_history_list, causing a kernel oops.
2919 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2920 Details : fix stack overflow in the distributed lock manager by defering
2921 export eviction after a failed AST to the elt thread instead of
2922 handling it in the dlm interpret routine.
2926 Description: Convert some comments to new format.
2927 Details : Update documenting comments to match doxygen conventions.
2931 Description: Grammar fixes.
2932 Details : A couple of trivial spelling fixes.
2936 Description: OSD_COUNTERS-mandatory
2937 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2938 to simplify the code.
2942 Description: simplify cmm_device freeing logic.
2943 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2947 Description: Add lockdep support to dt_object_operations locking interface.
2948 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2949 parameter indicating lock ordering. Update mdd code to use new
2954 Description: Introduce failloc constants for lockless IO tests.
2955 Details : Add two new failloc constants to test lockless IO. Only one of
2956 them in implemented---another is checked in yet to be landed
2961 Description: Add lockdep support for inode mutex.
2962 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2963 in the situations where Lustre has to lock more than one inode
2968 Description: Add optional invariants checking support.
2969 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2970 configure switch. This macro is to be used for consistency and
2971 sanity checks that are too expensive to be left in `production'
2976 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2977 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2982 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2983 Details : Use spin_lock_nested() in (the only) situation where more than
2984 one ldlm_lock is locked simultaneously. Also, fix possible
2985 dead-lock in ldlm_lock_change_resource() by enforcing particular
2990 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2991 Details : Instead of specifying each ldlm_lock call-back through separate
2992 parameter, wrap them into struct ldlm_callback_suite.
2996 Description: Kill join_lru obd method and its callers.
2997 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2998 in memory---a requirement implemented through ->o_join_lru() obd
2999 method in HEAD. Remove this method and its users.
3003 Description: Add asynchronous ldlm ENQUEUE completion handler.
3004 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
3005 case that stock ldlm_completion_ast() cannot handle as it waits
3006 until lock is granted. Introduce new ldlm_completion_ast_async()
3007 for this. Also comment ldlm_completion_ast().
3011 Description: ldlm_error <-> errno conversion.
3012 Details : Add functions to map (rather arbitrary) between LDLM error codes
3013 and standard errno values. CLIO needs this to prevent LDLM specific
3014 constants from escaping ldlm and osc.
3018 Description: Kill unused ldlm_handle2lock_ns() function.
3019 Details : Kill unused ldlm_handle2lock_ns() function.
3023 Description: Add lu_ref support to ldlm_lock
3024 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
3025 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
3026 and ldlm_resource. LDLM interface has to be changed, because code that
3027 releases a reference on a lock, has to "know" what reference this is.
3028 In the most frequent case
3030 lock = ldlm_handle2lock(handle);
3032 LDLM_LOCK_PUT(lock);
3034 no changes are required. When any other reference (received _not_ from
3035 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
3036 instead of LDLM_LOCK_PUT().
3038 Arguably, changes are pervasive, and interface requires some discipline
3039 for proper use. On the other hand, it was very instrumental in finding
3040 a few leaked lock references.
3044 Description: Add ldlm_lock_addref_try().
3045 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
3046 attempts to addref a lock that might be being canceled
3051 Description: Add ldlm_weigh_callback().
3052 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
3053 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
3054 hard-coded `number of pages' logic.
3058 Description: Add lockdep annotations to llog code.
3059 Details : Use appropriately tagged _nested() locking calls in the places
3060 where llog takes more than one ->lgh_lock lock.
3064 Description: Add loi_kms_set().
3065 Details : Wrap kms updates into a helper function.
3069 Description: Constify instances of struct lsm_operations.
3070 Details : Constify instances of struct lsm_operations.
3074 Description: lu_conf support.
3075 Details : On a server, a file system object is uniquely identified
3076 by a fid, which is sufficient to locate and load all object
3077 state (inode). On a client, on the other hand, more data are
3078 necessary instantiate an object. Change lu_object_find() and
3079 friends to take additional `lu_conf' argument describing object.
3080 Typically this includes layout information.
3084 Description: lu_context fixes.
3085 Details : Introduce new lu_context functions that are needed on the client
3086 side, where some system threads (ptlrpcd) are shared by multiple
3087 modules, and so cannot be stopped during module shutdown.
3091 Description: Add start and stop methods to lu_device_type_operations.
3092 Details : Introduce two new methods in lu_device_type_operations, that are
3093 invoked when first instance of a given type is created and last one
3094 is destroyed respectively. This is need by CLIO.
3098 Description: Add lu_ref support to struct lu_device.
3099 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
3100 track leaked references.
3104 Description: Introduce lu_kmem_descr.
3105 Details : lu_kmem_descr and its companion interface allow to create
3106 and destroy a number of kmem caches at once.
3110 Description: Fix lu_object finalization race.
3111 Details : Fix a race between lu_object_find() finding an object and its
3112 concurrent finalization. This race is (most likely) not possible
3113 on the server, but might happen on the client.
3117 Description: Introduce lu_ref interface.
3118 Details : lu_ref is a debugging module allowing to track references to
3119 a given object. It is quite cpu expensive, and has to be
3120 explicitly enabled with --enable-lu_ref. See usage description
3125 Description: Factor lu_site procfs stats into a separate function.
3126 Details : Separate lu_site stats printing code into a separate function
3127 that can be reused on a client.
3131 Description: Constify instances of struct {lu,dt,md}_device_operations.
3132 Details : Constify instances of struct {lu,dt,md}_device_operations.
3136 Description: Introduce struct md_site and move meta-data specific parts of
3137 struct lu_site here.
3138 Details : Move md-specific fields out of struct lu_site into special struct
3139 md_site, so that lu_site can be used on a client.
3143 Description: Kill mdd_lov_destroy().
3144 Details : Remove unused mdd code.
3148 Description: Add st_block checking to multistat.c.
3149 Details : Add st_block checking to multistat.c.
3153 Description: Add lu_ref support to struct obd_device.
3154 Details : Add lu_ref tracking to obd_device.
3158 Description: Kill obd_set_fail_loc().
3159 Details : Remove unused code.
3163 Description: Add special type for ptlrpc_request interpret functions.
3164 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
3165 there. Actual usage will be in CLIO.
3169 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
3170 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
3171 doesn't work with lockdep.
3175 Description: Add rwv.c test program.
3176 Details : New testing program exercising readv(2) and writev(2) (Qian).
3180 Description: Add sendfile.c test program.
3181 Details : New testing program exercising sendfile(2) (Jay).
3185 Description: Ratelimit a message that can be very frequent.
3186 Details : Ratelimit a memory allocation failure message that can
3191 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3192 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3196 Description: Liblustre build fixes.
3197 Details : Liblustre build fixes.
3201 Description: libcfs: add cfs_{need,cond}_resched() interface.
3202 Details : libcfs: add cfs_{need,cond}_resched() definition and
3203 implementations for Linux, NT, and liblustre.
3205 Severity : enhancement
3207 Description: More exported tunables for mballoc
3208 Details : Add support for tunable preallocation window and new tunables for
3209 large/small requests
3213 Description: Detect corruption of block bitmap and checking for preallocations
3214 Details : Checks validity of on-disk block bitmap. Also it does better
3215 checking of number of applied preallocations. When corruption is
3216 found, it turns filesystem readonly to prevent further corruptions.
3220 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3221 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3222 window to 0. If it does hit cache miss, and read-pattern is still
3223 stride-io mode, does not reset the stride window, but also does not
3224 increase the stride window length in this case.
3228 Frequency : only for big-endian servers
3229 Description: Check if system is big-endian while mounting fs with extents feature
3230 Details : Mounting a filesystem with extents feature will fail on big-endian
3231 systems since ext3-based ldiskfs is not supported on big-endian
3232 systems. This can be over-riden with "bigendian_extents" mount option.
3234 Severity : enhancement
3236 Description: The root squash functionality
3237 Details : A security feature, which is to prevent users from being able
3238 to mount lustre on their desktop, run as root, and delete
3239 all of the files in the filesystem. The goal is accomplished by
3240 remapping user id (UID) and group id (GID) of the root user to
3241 a UID and GID specified by the system administartor via Lustre
3242 configuration management server (MGS). The functionality also
3243 allows to specify sets of clients for which the remapping does
3248 Description: Excessive recovery window
3249 Details : With AT enabled, the recovery window can be excessively long (6000+
3250 seconds). To address this problem, we no longer use
3251 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3252 timeout no longer depends on the service time, it is set to
3253 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3254 time via pb_service_time.
3258 Description: Watchdog triggered on MDS failover
3259 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3260 the OSTs know that the MDS "UUID" can be reused for the same export
3261 from a different NID, so we do not need to wait for the export to be
3265 Frequency : rare, only if using MMP with Linux RAID
3267 Description: MMP doesn't work with Linux RAID
3268 Details : While using HA for Lustre servers with Linux RAID, it is possible
3269 that MMP will not detect multiple mounts. To make this work we
3270 need to unplug the device queue in RAID when the MMP block is being
3271 written. Also while reading the MMP block, we should read it from
3272 disk and not the cached one.
3274 Severity : enhancement
3276 Description: open file using fid
3277 Details : A file can be opened using just its fid, like
3278 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3281 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3284 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3285 Details : add proper locking for creation of jbd_4k slab cache
3289 Description: MMP check in ext3_remount() fails without displaying any error
3290 Details : When multiple mount protection fails during remount, proper error
3293 Severity : enhancement
3295 Description: Allow stripe size to be up to 4G-64k
3296 Details : Fix math logic to allow large stripe sizes.
3300 Description: add check for >8TB ldiskfs filesystems
3301 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3302 Don't allow >8TB ldiskfs filesystems to be mounted without
3303 force_over_8tb mount option
3308 Descriptoin: OSS complains "too many blocks" and crashes
3309 Details : In filter_commitrw_write():
3310 1. lquota_chkquota();
3311 2. obdo_from_inode();
3312 3. lquota_pending_commit();
3313 "2." has changed uid and gid in this case, then caused
3314 lquota_chkquota() and lquota_pending_commit() to use
3315 different uid and gid. That is the root of the bug.
3319 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3320 Details : Detect and print out the 'stripe count = -1' case properly.
3324 Description: Out or order replies might be lost on replay
3325 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3326 request with tid smaller than one currently being added, add it
3327 to the start, not end of the list.
3329 --------------------------------------------------------------------------------
3331 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3333 * Support for kernels:
3334 2.6.5-7.283 (SLES 9),
3335 2.6.9-55.EL (RHEL 4),
3336 2.6.16.46-0.14 (SLES 10),
3337 2.6.18.8 vanilla (kernel.org)
3338 * Client support for unpatched kernels:
3339 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3340 2.6.16 - 2.6.22 vanilla (kernel.org)
3341 * Due to recently discovered recovery problems, we do not recommend
3342 using patchless RHEL 4 clients with this or any earlier release.
3343 * Recommended e2fsprogs version: 1.39.cfs8
3344 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3345 * Starting with this release, the ldiskfs backing filesystem required
3346 by Lustre is now in its own package, lustre-ldiskfs. This package
3347 should be installed. It is versioned separately from Lustre and
3348 may be released separately in future.
3352 Description: block reactivating mgc import until all deactivates complete
3353 Details : Fix race when failing back MDT/MGS to itself (testing)
3356 Frequency : at statup only
3358 Description: mds_lov_synchronize race leads to various problems
3359 Details : simultaneous MDT->OST connections at startup can cause the
3360 sync to abort, leaving the OSC in a bad state.
3362 Severity : enhancement
3364 Description: add optional extra BUILD_VERSION info
3365 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3366 to override the lustre version.
3369 Frequency : 2.6.18 servers only
3371 Description: ll_kern_mount() doesn't release the module reference
3372 Details : The ldiskfs module reference count never drops down to 0
3373 because ll_kern_mount() doesn't release the module reference.
3378 Description: server LBUG when using old ost_num_threads parameter
3379 Details : Accept the old ost_num_threads parameter but warn that it
3380 is deprecated, and fix an off-by-one error that caused an LBUG.
3385 Description: Transient SCSI error results in persistent IO issue
3386 Details : iobuf->dr_error is not reinitialized to 0 between two
3390 Frequency : sometimes when underlying device returns I/O errors
3392 Description: OSTs not going read-only during write failures
3393 Details : OSTs are not remounted read-only when the journal commit threads
3394 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3395 instead of the ext3 wrappers.
3399 Description: poor connect scaling with increasing client count
3400 Details : Don't run filter_grant_sanity_check for more than 100 exports
3401 to improve scaling for large numbers of clients.
3404 Frequency : SLES10 only
3406 Description: sanity-quota.sh quotacheck failed: rc = -22
3407 Details : Quotas cannot be enabled on SLES10.
3410 Frequency : liblustre clients only
3412 Description: getdirentries does not give error when run on compute nodes
3413 Details : getdirentries does not fail when the size specified as an argument
3414 is too small to contain at least one entry
3416 Severity : enhancement
3418 Description: Add LNET router traceability for debug purposes
3419 Details : If a checksum failure occurs with a router as part of the
3420 IO path, the NID of the last router that forwarded the bulk data
3421 is printed so it can be identified.
3426 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3427 Details : Due to a race condition, liblustre clients were occasionally
3428 evicted incorrectly.
3430 Severity : enhancement
3432 Description: lfs setstripe use optional parameters instead of postional
3435 Severity : enhancement
3437 Description: Nanosecond timestamp support for ldiskfs
3438 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3439 resolution timestamps. There is not yet support for this at
3440 the Lustre filesystem level.
3443 Frequency : during server recovery
3445 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3446 Details : request with rq_no_resend flag not awake l_wait_event if they get
3450 Frequency : nfs export on patchless client
3452 Description: connectathon hang when test nfs export over patchless client
3453 Details : Disconnected dentry cannot be found with lookup, so we do not need
3454 to unhash it or make it invalid
3457 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3458 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3459 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3464 Description: open req refcounting wrong on reconnect
3465 Details : If reconnect happened between getting open reply from server and
3466 call to mdc_set_replay_data in ll_file_open, we will schedule
3467 replay for unreferenced request that we are about to free.
3468 Subsequent close will crash in variety of ways.
3469 Check that request is still eligible for replay in
3470 mdc_set_replay_data().
3475 Description: disable writes to filesystem when reading health_check file
3476 Details : the default for reading the health_check proc file has changed
3477 to NOT do a journal transaction and write to disk, because this
3478 can cause reads of the /proc file to hang and block HA state
3479 checking on a healthy but otherwise heavily loaded system. It
3480 is possible to return to the previous behaviour during configure
3481 with --enable-health-write.
3483 Severity : enhancement
3485 Description: 64-bit inode version
3486 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3487 to the inode. This will be required for version-based recovery.
3492 Description: MDS fails to start if a duplicate client export is detected
3493 Details : in some rare cases it was possible for a client to connect to
3494 an MDS multiple times. Upon recovery the MDS would detect this
3495 and fail during startup. Handle this more gracefully.
3497 Severity : enhancement
3499 Description: Add -o localflock option to simulate old noflock
3501 Details : This will achieve local-only flock/fcntl locks
3507 Description: log_commit_thread vs filter_destroy race leads to crash
3508 Details : Take import reference before releasing llog record semaphore
3513 Description: Wrong request locking in request set processing
3514 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3515 imp_delayed_list, in this place should be used imp_lock.
3518 Frequency : when reconnection
3520 Description: Grant Leak when osc reconnect to OST
3521 Details : When osc reconnect ost, OST(filter) should check whether it
3522 should grant more space to client by comparing fed_grant and
3523 cl_avail_grant, and return the granted space to client instead
3524 of "new granted" space, because client will call osc_init_grant
3525 to update the client grant space info.
3528 Frequency : when client reconnect to OST
3530 Description: Grant Leak when osc do resend and replay bulk write
3531 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3532 bulk write request, because the grant info while be sync between
3533 OSC and OST when reconnect, and we should ignore the grant info
3534 these of resend/replay write req.
3539 Description: Grant space more than available left space sometimes.
3540 Details : When then OST is about to be full, if two bulk writing from
3541 different clients came to OST. Accord the avaliable space of the
3542 OST, the first req should be permitted, and the second one
3543 should be denied by ENOSPC. But if the seconde arrived before
3544 the first one is committed. The OST might wrongly permit second
3545 writing, which will cause grant space > available space.
3548 Frequency : when client is evicted
3550 Description: Grant might be wrongly erased when osc is evicted by OST
3551 Details : when the import is evicted by server, it will fork another
3552 thread ptlrpc_invalidate_import_thread to invalidate the
3553 import, where the grant will be set to 0. While the original
3554 thread will update the grant it got when connecting. So if
3555 the former happened latter, the grant will be wrongly errased
3556 because of this race.
3561 Description: Checking Stale with correct fid
3562 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3563 to check whether it is stale, because sometimes, we want the
3564 enqueue happened anyway, and op_data.fid2 will not be initialized.
3566 Severity : enhancement
3568 Description: update patchless client
3569 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3572 Frequency : only with 2.4 kernel
3574 Description: random memory corruption
3575 Details : size of struct ll_inode_info is to big for union inode.u and this
3576 can be cause of random memory corruption.
3581 Description: Memory leak in recovery
3582 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3583 It should also check obd_fail before fsfilt_start, otherwise if
3584 fsfilt_start return -EROFS,(failover mds during mds recovery).
3585 then the req will return with repmsg->transno = 0 and rc = EROFS.
3586 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3587 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3588 be freed no matter whether fsfilt_commit success or not.
3591 Frequency : only with huge count clients
3593 Description: Prevents from taking the superblock lock in llap_from_page for
3595 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3596 for llap_from_page call in ll_removepage prevents from taking the
3597 superblock lock for a soon died page.
3602 Description: Not check open intent error before release open handle
3603 Details : in some rare cases, the open intent error is not checked before
3604 release open handle, which may cause
3605 ASSERTION(open_req->rq_transno != 0), because it tries to release
3606 the failed open handle.
3611 Description: Set cat log bitmap only after create log success.
3612 Details : in some rare cases, the cat log bitmap is set too early. and it
3613 should be set only after create log success.
3617 Description: Accessing a block bevice can re-enable I/O when Lustre is
3618 tearing down a device.
3619 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3624 Description: service threads may hog cpus when there are a lot of requests
3625 Details : Insert cond_resched to give other threads a chance to use some CPU
3630 Description: the cat log was not initialized in recovery
3631 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3632 unlink log on mds will not be initialized until mds post
3633 recovery. And also in mds post recovery, the unlink log will
3634 initialization will be done asynchronausly, so there will be race
3635 between add unlink log and unlink log initialization.
3639 Description: brw_stats were being printed incorrectly
3640 Details : brw_stats were being printed as log2 but all of them were not
3641 recorded as log2. Also remove some code duplication arising from
3642 filter_tally_{read,write}.
3646 Frequency : rare, only in recovery.
3647 Description: ASSERTION(req->rq_type != LI_POISON) failed
3648 Details : imp_lock should be held while iterating over imp_sending_list for
3649 prevent destroy request after get timeout in ptlrpc_queue_wait.
3653 Description: replay-single.sh test 52 fails
3654 Details : A lock's skiplist need to be cleanup when it being unlinked
3655 from its resource list.
3659 Description: Short directio read returns full requested size rather than
3661 Details : Direct I/O operations should return actual amount of bytes
3662 transferred rather than requested size.
3664 Severity : enhancement
3666 Description: metadata RPC reduction (e.g. for rm performance)
3667 Details : decrease the amount of synchronous RPC between clients and servers
3668 by canceling conflicing lock before the operation on the client
3669 and packing thier handles into the main operation RPC to server.
3671 Severity : enhancement
3673 Description: Async OSC create to avoid the blocking unnecessarily.
3674 Details : If an OST has no remaining object, system will block on the
3675 creation when it needs to create a new object on this OST. Now,
3676 ways use pre-created objects when available, instead of blocking on
3677 an empty osc while others are not empty. If we must block, we
3678 block for the shortest possible period of time.
3682 Description: improve handling recoverable errors
3683 Details : if request processig with error which can be recoverable on server
3684 request should be resend, otherwise page released from cache and
3687 Severity : enhancement
3689 Description: refine locking for avoid write wrong info into lov_objid file
3690 Details : fix possible races with add new target and write/update data in
3693 --------------------------------------------------------------------------------
3695 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3700 Frequency : on some architectures
3702 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3703 Details : Uninitialized flags sometimes cause configuration commands to
3707 Frequency : patchless clients only
3709 Description: missing __iget() symbol export
3710 Details : The __iget() symbol export is missing. To avoid the need for
3711 this on patchless clients the deathrow inode reaper is turned
3712 off, and we depend on the VM to clean up old inodes. This
3713 dependency was during via the fix for bug 12181.
3715 --------------------------------------------------------------------------------
3717 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3719 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3720 INTEROPERATE with older versions automatically. In many cases a
3721 special upgrade step is needed. Please read the
3722 user documentation before upgrading any part of a live system.
3723 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3724 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3725 * WARNING: Lustre configuration and startup changes are required with
3726 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3730 Severity : enhancement
3732 Description: MountConf
3733 Details : Lustre configuration is now managed via mkfs and mount
3734 commands instead of lmc and lconf. New obd types (MGS, MGC)
3735 are added for dynamic configuration management. See
3736 https://mail.clusterfs.com/wikis/lustre/MountConf for
3739 Severity : enhancement
3741 Description: dynamic OST addition
3742 Details : OSTs can now be added to a live filesystem
3744 Severity : enhancement
3746 Description: startup order invariance
3747 Details : MDTs and OSTs can be started in any order. Clients only
3748 require the MDT to complete startup.
3750 Severity : enhancement
3752 Description: parallel, asynchronous orphan cleanup
3753 Details : orphan cleanup is now performed in separate threads for each
3754 OST, allowing parallel non-blocking operation.
3756 Severity : enhancement
3758 Description: optimized stripe assignment
3759 Details : stripe assignments are now made based on ost space available,
3760 ost previous usage, and OSS previous usage, in order to try
3761 to optimize storage space and networking resources.
3763 Severity : enhancement
3765 Description: Permanently set tunables
3766 Details : All writable /proc/fs/lustre tunables can now be permanently
3767 set on a per-server basis, at mkfs time or on a live system.
3769 Severity : enhancement
3771 Description: Lustre message v2
3772 Details : Add lustre message format v2.
3774 Severity : enhancement
3776 Description: client OST exclusion list
3777 Details : Clients can be started with a list of OSTs that should be
3778 declared "inactive" for known non-responsive OSTs.
3782 Description: SPEC SFS validation failure on NFS v2 over lustre.
3783 Details : Changes the blocksize for regular files to be 2x RPC size,
3784 and not depend on stripe size.
3786 Severity : enhancement
3788 Description: Multiple MD RPCs in flight.
3789 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3790 To avoid overly-overloading MDS, introduce a limit on number of
3791 MDS RPCs in flight for a single client and add /proc controls
3792 to adjust this limit.
3794 Severity : enhancement
3796 Description: client read/write statistics
3797 Details : Add client read/write call usage stats for performance
3798 analysis of user processes.
3799 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3800 file access. extents_stats shows chunk size distribution.
3801 extents_stats_per_process show chunk size distribution per
3804 Severity : enhancement
3806 Description: mds statistics
3807 Details : Add detailed mds operations statistics in
3808 /proc/fs/lustre/mds/*/stats.
3812 Description: Failure of copying files with lustre special EAs.
3813 Details : Client side always return success for setxattr call for lustre
3814 special xattr (currently only "trusted.lov").
3819 Description: Refcount LNET uuids
3820 Details : The global LNET uuid list grew linearly with every startup;
3821 refcount repeated list entries instead of always adding to
3824 Severity : enhancement
3826 Description: Dynamic service threads
3827 Details : Within a small range, start extra service threads
3828 automatically when the request queue builds up.
3831 Frequency : mixed-endian client/server environments
3833 Description: mixed-endian crashes
3834 Details : The new msg_v2 system had some failures in mixed-endian
3837 Severity : enhancement
3839 Description: Easy OST removal
3840 Details : OSTs can be permanently deactivated with e.g. 'lctl
3841 conf_param lustre-OST0001.osc.active=0'
3843 Severity : enhancement
3845 Description: MGS proc entries
3846 Details : Added basic proc entries for the MGS showing what filesystems
3849 Severity : enhancement
3851 Description: provide MGS failover
3852 Details : Added config lock reacquisition after MGS server failover.
3854 Severity : enhancement
3856 Description: add Linux 2.4 support
3857 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3861 Description: a large application tries to do I/O to the same resource and dies
3862 in the middle of it.
3863 Details : Check the req->rq_arrival time after the call to
3864 ost_brw_lock_get(), but before we do anything about
3865 processing it & sending the BULK transfer request. This
3866 should help move old stale pending locks off the queue as
3867 quickly as obd_timeout.
3870 Frequency : when an incorrect nid is specified during startup
3872 Description: ptlrpc connect to non-existant node causes kernel crash
3873 Details : LNET can't be re-entered from an event callback, which
3874 happened when we expire a message after the export has been
3875 cleaned up. Instead, hand the zombie cleanup off to another
3878 Severity : enhancement
3880 Description: plain/inodebits lock performance improvement
3881 Details : Grouping plain/inodebits in granted list by their request modes
3882 and bits policy, thus improving the performance of search through
3886 Frequency : only if OST filesystem is corrupted
3888 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3889 Details : for a short time RPCs with bulk IO are in the replay list,
3890 but replay of bulk IOs is unimplemented. If the OST filesystem
3891 is corrupted due to disk cache incoherency and then replay is
3892 started it is possible to trip an assertion. Avoid putting
3893 committed RPCs into the replay list at all to avoid this issue.
3896 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3898 Description: System hang on startup
3899 Details : This bug allowed the liblustre (e.g. catamount) client to
3900 return to the app before handling all startup RPCs. This
3901 could leave the node unresponsive to lustre network traffic
3902 and manifested as a server ptllnd timeout.
3904 Severity : enhancement
3906 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3907 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3908 boolean to control whether to print peer debug info when a
3909 client's RPC times out.
3912 Frequency : only for kernels with patches from Lustre below 1.4.3
3914 Description: Remove old rdonly API
3915 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3918 Frequency : only for devices with external journals
3920 Description: Set external device read-only also
3921 Details : During a commanded failover stop, we set the disk device
3922 read-only while the server shuts down. We now also set any
3923 external journal device read-only at the same time.
3926 Frequency : when upgrading from 1.4 while trying to change parameters
3928 Description: The wrong (new) MDC name was used when setting parameters for
3929 upgraded MDT's. Also allows changing of OSC (and MDC)
3930 parameters if --writeconf is specified at tunefs upgrade time.
3933 Frequency : when setting specific ost indicies
3935 Description: QOS code breaks on skipped indicies
3936 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3937 created with --index need not be sequential.
3939 Severity : enhancement
3941 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3942 Details : The uninit_groups feature works in conjunction with the kernel
3943 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3944 up the pass1 processing of e2fsck. This is a read-only feature
3945 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3946 mount filesystems that have had this feature enabled.
3948 Severity : enhancement
3950 Description: Improve multi-block allocation algorithm to avoid fragmentation
3951 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3952 allocation locality and avoid filesystem fragmentation.
3954 ------------------------------------------------------------------------------
3956 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3958 * Support for kernels:
3959 2.4.21-47.0.1.EL (RHEL 3)
3960 2.6.5-7.283 (SLES 9)
3961 2.6.9-42.0.10.EL (RHEL 4)
3962 2.6.12.6 vanilla (kernel.org)
3963 2.6.16.27-0.9 (SLES 10)
3964 * Recommended e2fsprogs version: 1.39.cfs5
3966 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3970 Frequency : occasional, depends on client load and configuration
3971 Bugzilla : 12181, 12203
3972 Description: data loss for recently-modified files
3974 Details : In some cases it is possible that recently written or created
3975 files may not be written to disk in a timely manner (this should
3976 normally be within 30s unless client IO load is very high).
3977 The problem appears as zero-length files or files that are a
3978 multiple of 1MB in size after a client crash or client eviction
3979 that are missing data at the end of the file.
3981 This problem is more likely to be hit on clients where files are
3982 repeatedly created and unlinked in the same directory, clients
3983 have a large amount of RAM, have many CPUs, the filesystem has
3984 many OSTs, the clients are rebooted frequently, and/or the files
3985 are not accessed by other nodes after being written.
3987 The presence of the problem can be detected by looking at
3988 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3989 smaller than the second (nr_unused) then dirty files will not
3990 be flushed automatically to disk. "sync; sleep 10" should be
3991 run several times on the node before unmounting it to update
3992 Lustre (this is also safe to run on nodes without this problem).
3994 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3995 that can cause this same problem, so customers using that kernel
3996 also need to update the kernel in addition to Lustre. In order
3997 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3999 It is normal that files written just before a client crash (less
4000 than 30s) may not yet have been flushed to disk, even for local
4004 Frequency : frequent on thin XT3 nodes
4006 Description: UUID collision on thin XT3 Linux nodes
4007 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
4008 randomly, since we relied on an insufficiently-seeded PRNG.
4013 Description: OSS hangs after "All ost request buffers busy"
4014 Details : A deadlock between quota and journal operations caused OSS
4015 hangs after printing "All ost request buffers busy."
4018 Frequency : always on liblustre builds
4020 Description: Cleanup compiler warnings on liblustre
4023 Frequency : always on liblustre builds on XT3
4025 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
4029 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
4030 > 32000 subdirectories
4031 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
4032 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
4033 subdirectory count crosses 32000. This will aid e2fsck to
4034 correctly handle more than 32000 subdirectories.
4037 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
4039 Description: System hang on startup
4040 Details : This bug allowed the liblustre (e.g. catamount) client to
4041 return to the app before handling all startup RPCs. This
4042 could leave the node unresponsive to lustre network traffic
4043 and manifested as a server ptllnd timeout.
4045 Severity : enhancement
4047 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
4048 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
4049 boolean to control whether to print peer debug info when a
4050 client's RPC times out.
4055 Description: make O_SYNC working on 2.6 kernels
4056 Details : 2.6 kernels use different method for mark pages for write,
4057 so need add a code to lustre for O_SYNC work.
4062 Description: Failure to close file and release space on NFS
4063 Details : Put inode details into lock acquired in ll_intent_file_open.
4064 Use mdc_intent_lock in ll_intent_open to properly
4065 detect all kind of errors unhandled by mdc_enqueue.
4070 Description: proc file read during shutdown sometimes raced obd removal,
4072 Details : Add lock to prevent obd access after proc file removal.
4075 Frequency : Only for files larger than 4GB on 32-bit clients.
4077 Description: improperly doing page alignment of locks
4078 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
4079 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
4084 Description: RPCs being resent when they shouldn't be
4085 Details : Some RPCs that should not be resent are being resent. This
4086 can cause inconsistencies in the RPC state machine. Do not
4087 resend such requests.
4090 Frequency : rare, only with NFS export
4092 Description: Crash on NFS re-export node
4093 Details : under very unusual load conditions an assertion is hit in
4094 ll_intent_file_open()
4097 Frequency : only if OST filesystem is corrupted
4099 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4100 Details : for a short time RPCs with bulk IO are in the replay list,
4101 but replay of bulk IOs is unimplemented. If the OST filesystem
4102 is corrupted due to disk cache incoherency and then replay is
4103 started it is possible to trip an assertion. Avoid putting
4104 committed RPCs into the replay list at all to avoid this issue.
4109 Description: large O_DIRECT requests fail under memory pressure/fragmentation
4110 Details : Large single O_DIRECT read and write calls can fail to allocate
4111 a sufficiently large buffer to process the request. In case of
4112 allocation failure the allocation is retried with a smaller
4113 buffer and broken into smaller requests.
4115 Severity : enhancement
4117 Description: Add -o localflock option to simulate old noflock behaviour.
4118 Details : This will achieve local-only flock/fcntl locks coherentness.
4123 Description: versioning check is incomplete
4124 Details : Checking the version difference of client vs. server, report
4125 error if the gap is too big.
4130 Description: add support PG_writeback bit
4131 Details : add support for PG_writeback bit for Lustre, for more carefull
4132 work with page cache in 2.6 kernel. This also fix some deadlocks
4133 and remove hack for work O_SYNC with 2.6 kernel.
4135 Severity : enhancement
4137 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4138 Details : The uninit_groups feature works in conjunction with the kernel
4139 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4140 up the pass1 processing of e2fsck. This is a read-only feature
4141 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4142 mount filesystems that have had this feature enabled.
4144 Severity : enhancement
4146 Description: Improve multi-block allocation algorithm to avoid fragmentation
4147 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4148 allocation locality and avoid filesystem fragmentation.
4150 ------------------------------------------------------------------------------
4152 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
4154 * Support for kernels:
4155 2.6.9-42.0.3.EL (RHEL 4)
4156 2.6.5-7.276 (SLES 9)
4157 2.4.21-47.0.1.EL (RHEL 3)
4158 2.6.12.6 vanilla (kernel.org)
4159 2.6.16.21-0.8 (SLES10)
4160 * Recommended e2fsprogs version: 1.39.cfs2-0
4162 * The backwards-compatible /proc/sys/portals symlink has been removed
4163 in this release. Before upgrading, please ensure that you change
4164 any configuration scripts or /etc/sysctl.conf files that access
4165 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4166 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4167 in advance of the upgrade on any system running Lustre 1.4.6 or
4168 newer, since /proc/sys/lnet was added in that version.
4169 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4173 Frequency : only when quota is used
4175 Description: avoid scanning export list for quota master
4176 Details : Change the algorithms to avoid scanning export list in order
4177 to improve the efficiency.
4180 Frequency : MDS failover only, very rarely
4182 Description: "went back in time" messages on mds failover
4183 Details : The greatest transno may be lost when the current operation
4184 finishes with an error (transno==0) and the client's last_rcvd
4185 record is over-written. Save the greatest transno in the
4186 mds_last_transno for this case.
4189 Frequency : always for specific kernels and striping counts
4191 Description: client may get "Matching packet too big" without ACL support
4192 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4193 when trying to access files in certain configurations. The
4194 clients should in fact be denied when mounting because they do
4195 not understand ACLs.
4198 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4200 Description: many clients connecting with IO in progress causes connect timeouts
4201 Details : Avoid synchronous journal commits to avoid delays caused by many
4202 clients connecting/disconnecting when bulk IO is in progress.
4203 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4204 OST_IO_PORTAL to avoid delays behind potentially many pending
4208 Frequency : occasionally with multiple writers to a single file
4210 Description: shared writes to file may result in wrong size reported by stat()
4211 Details : Allow growing of kms when extent lock is cancelled
4214 Frequency : always with random mmap IO to multi-striped file
4216 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4217 Details : Only if the hole is at the end of OST object so that kms is too
4218 small. Fix is to increase kms accordingly in ll_nopage.
4221 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4223 Description: writes to a missing object would leak memory on the OST
4224 Details : If there is an inconsistency between the MDS and OST filesystems,
4225 such that the MDS references an object that doesn't exist, writes
4226 to that object will leak memory due to incorrect cleanup in the
4227 error handling path, eventually running out of memory on the OST.
4232 Description: Creating too long symlink causes lustre errors
4233 Details : Check symlink and name lengths before sending requests to MDS.
4236 Frequency : only if flock is enabled (not on by default)
4238 Description: posix locks not released on fd closure on 2.6.9+
4239 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4240 kernels, this caused such locks not to be released on fd close and
4241 then assertions on fs unmount about still used locks.
4244 Frequency : MDS failover only, very rarely
4246 Description: clients may get ASSERTION(granted_lock != NULL)
4247 Details : When request was taking a long time, and a client was resending
4248 a getattr by name lock request. The were multiple lock requests
4249 with the same client lock handle and
4250 mds_getattr_name->fixup_handle_for_resent_request found one of the
4251 lock handles but later failed with ASSERTION(granted_lock != NULL).
4256 Description: handle->h_buffer_credits > 0, assertion failure
4257 Details : h_buffer_credits is zero after truncate, causing assertion
4258 failure. This patch extends the transaction or creates a new
4262 Frequency : NFS re-export or patchless client
4263 Bugzilla : 11179, 10796
4264 Description: Crash on NFS re-export node (__d_move)
4265 Details : We do not want to hash the dentry if we don't have a lock.
4266 But if this dentry is later used in d_move, we'd hit uninitialised
4267 list head d_hash, so we just do this to init d_hash field but
4268 leave dentry unhashed.
4271 Frequency : NFS re-export or patchless client
4273 Description: NFS exports has problem with symbolic link
4274 Details : lustre client didn't properly install dentry when re-exported
4275 to NFS or running patchless client.
4278 Frequency : NFS re-export or patchless client
4280 Description: Various nfs/patchless fixes.
4281 Details : fixes reuse disconected alias for lookup process - this fixes
4282 warning "find_exported_dentry: npd != pd",
4283 fix permission error with open files at nfs.
4284 fix apply umask when do revalidate.
4287 Frequency : occasional
4289 Description: Crash on NFS re-export node
4290 Details : calling clear_page() on the wrong pointer triggered oops in
4291 generic_mapping_read().
4294 Frequency : rarely, using O_DIRECT IO
4296 Description: unaligned directio crashes client with LASSERT
4297 Details : check for unaligned buffers before trying any requests.
4300 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4302 Description: stale data returned from RAID cache
4303 Details : If only a small amount of IO is done to the RAID device before
4304 reading it again it is possible to get stale data from the RAID
4305 cache instead of reading it from disk.
4308 Frequency : always for sles10 kernel
4310 Description: sles10 support
4311 Details : ll_follow_link: compile fixes and using of nd_set_link
4312 under newer kernels.
4315 Frequency : depends on arch, kernel and compiler version, always on sles10
4318 Description: recursive or deep enough symlinks cause stack overflow
4319 Details : getting rid of large stack-allocated variable in
4323 Frequency : depends on hardware
4325 Description: lustre write performance loss in the SLES10 kernel
4326 Details : the performance loss is caused by using of write barriers in the
4327 ext3 code. The SLES10 kernel turns barrier support on by
4328 default. The fix is to undo that change for ldiskfs.
4330 ------------------------------------------------------------------------------
4332 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4334 * Support for kernels:
4335 2.6.9-42.0.3EL (RHEL 4)
4336 2.6.5-7.276 (SLES 9)
4337 2.4.21-47.0.1.EL (RHEL 3)
4338 2.6.12.6 vanilla (kernel.org)
4342 Frequency : quota enabled and large files being deleted
4344 Description: releasing more than 4GB of quota at once hangs OST
4345 Details : If a user deletes more than 4GB of files on a single OST it
4346 will cause the OST to spin in an infinite loop. Release
4347 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4352 Description: statfs data retrieved from /proc may be stale or zero
4353 Details : When reading per-device statfs data from /proc, in the
4354 {kbytes,files}_{total,free,avail} files, it may appear
4355 as zero or be out of date.
4358 Frequency : systems with MD RAID1 external journal devices
4360 Description: lconf's call to blkid is confused by RAID1 journal devices
4361 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4362 of returning all block devices that match the journal UUID.
4365 Frequency : always, for aggregate stripe size over 4GB
4367 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4368 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4369 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4372 Frequency : always if "lfs find" used on a local file/directory
4374 Description: "lfs find" segfaults if used on a local file/directory
4375 Details : The case where a directory component was not specified wasn't
4376 handled correctly. Handle this properly.
4379 Frequency : always on ppc64
4381 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4382 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4383 perform endianness conversion.
4386 Frequency : rarely (truncate to non-zero file size after write under load)
4387 Bugzilla : 10730, 10687
4388 Description: Files padded with zeros to next 4K multiple
4389 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4390 it is possible that files that are truncated to a non-zero size
4391 immediately after being written are filled with zero bytes beyond
4392 the truncated size. No file data is lost.
4394 Severity : enhancement
4396 Description: Allow recovery/failover for liblustre clients.
4397 Details : liblustre clients were unaware of failover configurations until
4400 Severity : enhancement
4402 Description: user file locks should fail when not mounting with flock option
4403 Details : Set up an error-returning stub in ll_file_operations.lock field
4404 to prevent incorrect behaviour when client is mounted without
4405 flock option. Also, set up properly f_op->flock field for
4409 Frequency : always on ia64
4411 Description: "lfs df" loops on printing out MDS statfs information
4412 Details : The obd_ioctl_data was not initialized and in some systems
4413 this caused a failure during the ioctl that did not return
4414 an error. Initialize the struct and return an error on failure.
4417 Frequency : SLES 9 only
4419 Description: Error of copying files with lustre special EAs as root
4420 Details : Client side always return success for setxattr call for lustre
4421 special xattr (currently only "trusted.lov").
4424 Frequency : rarely on clusters with both ia64+i386 clients
4426 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4427 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4428 clients are doing concurrent writes to the same file it is
4429 possible that the ia64 clients may LASSERT because the OST
4430 extent locks are not PAGE_SIZE aligned. Ensure that grown
4431 locks are always aligned on the request boundary.
4434 Frequency : specific use, occasional
4436 Description: Overwriting in use executable truncates on-disk binary image
4437 Details : If one node attempts to overwrite an executable in use by
4438 another node, we now correctly return ETXTBSY instead of
4439 truncating the file.
4441 Severity : enhancement
4443 Description: Async OSC create to avoid the blocking unnecessarily.
4444 Details : If an OST has no remaining object, system will block on the creating
4445 when need to create a new object on this OST. Now, ways use
4446 pre-created objects when available, instead of blocking on an
4447 empty osc while others are not empty. If we must block, we block
4448 for the shortest possible period of time.
4453 Description: chmod on Lustre root is propagated to other clients
4454 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4455 invalid by the follow_mount time.
4458 Frequency : liblustre clients only
4460 Description: Race in 'instant cancel' lock handling could lead to such locks
4461 never to be granted in case of SMP MDS
4462 Details : Do not destroy not yet granted but cbpending locks in
4466 Frequency : replay/resend of open
4468 Description: non null lock assetion failure in mds_intent_policy
4469 Details : Trying to replay/resend lockless open requests resulted in
4470 mds_open() returning 0 with no lock. Now it sets a flag if
4471 a lock is going to be returned.
4473 Severity : enhancement
4475 Description: Checksum enhancements
4476 Details : New checksum enhancements allow for resending RPCs that failed
4479 Severity : enhancement
4481 Description: Tunables on number of dirty pages in cacche
4482 Details : Allow to set limit on number of dirty pages cached.
4487 Description: client crash on unmount - lock still has references
4488 Details : In some error handling cases it was possible to leak a lock
4489 reference on a client while accessing a file. This was not
4490 harmful to the client during operation, but would cause the
4491 client to crash when the filesystem is unmounted.
4494 Frequency : specific case, rare
4496 Description: ETXTBSY on mds though file not in use by client
4497 Details : ETXTBSY is no longer incorrectly returned when attempting to
4498 chmod or chown a directory that the user previously tried to
4499 execute or a currently-executing binary.
4502 Frequency : extremely rare except on liblustre-based clients
4504 Description: Lustre space not freed when files are deleted
4505 Details : Clean up open-unlinked files after client eviction. Previously
4506 the unlink was skipped and the files remained as orphans.
4511 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4512 Details : In some cases it was possible to send a blocking callback to
4513 a client doing a glimpse, even though that client didn't get
4514 a lock granted. When the glimpse lock is cancelled on the OST
4515 the freed lock is left on the waiting list and corrupted the list.
4518 Frequency : all core dumps
4520 Description: Broke core dumps to lustre
4521 Details : Negative dentry may be unhashed if parent does not have UPDATE
4522 lock, but some callers, e.g. do_coredump, expect dentry to be
4523 hashed after successful create, hash it in ll_create_it.
4525 ------------------------------------------------------------------------------
4527 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4529 * Support for kernels:
4530 2.6.9-42.0.2.EL (RHEL 4)
4531 2.6.5-7.276 (SLES 9)
4532 2.4.21-40.EL (RHEL 3)
4533 2.6.12.6 vanilla (kernel.org)
4537 Frequency : always on RHEL 3
4539 Description: Number of open files grows over time
4540 Details : The number of open files grows over time, whether or not
4541 Lustre is started. This was due to a filp leak introduced
4542 by one of our kernel patches.
4544 ------------------------------------------------------------------------------
4546 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4548 * Support for kernels:
4549 2.6.9-42.EL (RHEL 4)
4550 2.6.5-7.276 (SLES 9)
4551 2.4.21-40.EL (RHEL 3)
4552 2.6.12.6 vanilla (kernel.org)
4557 Bugzilla : 5719, 9635, 9792, 9684
4558 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4559 Details : If a server is under heavy load and cannot reply to new
4560 connection requests before the client resends the (re)connect,
4561 the connection handling code can behave badly if two service
4562 threads are concurrently handing separate (re)connections from
4563 the same client. Add better locking to the connection handling
4564 code, and ensure that only a single connection will be processed
4565 for a given client UUID, even if the lock is dropped.
4567 Severity : enhancement
4569 Description: add TCP zero-copy support to kernel
4570 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4571 sends if the hardware supports scatter-gather and checksumming.
4572 This allows socklnd to do client-write and server-read more
4573 efficiently and reduce CPU utilization from skbuf copying.
4576 Frequency : only if NFS exporting from client
4578 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4579 Details : If setting attributes on a file created under NFS that had
4580 never been opened it would be possible to oops the client
4581 if the file had no objects.
4584 Frequency : always for liblustre
4586 Description: liblustre client does MDS+OSTs setattr RPC for each write
4587 Details : When doing a write from a liblustre client, the client
4588 incorrectly issued an RPC to the MDS and each OST the file was
4589 striped over in order to update the timestamps. When writing
4590 with small chunks and many clients this could overwhelm the MDS
4591 with RPCs. In all cases it would slow down the write because
4592 these RPCs are unnecessary.
4594 Severity : enhancement
4596 Description: allow number of MDS service threads to be changed at module load
4597 Details : It is now possible to change the number of MDS service threads
4598 running. Adding "options mds mds_num_threads={N}" to the MDS's
4599 /etc/modprobe.conf will set the number of threads for the next
4600 time Lustre is restarted (assuming the "mds" module is also
4601 reloaded at that time). The default number of threads will
4602 stay the same, 32 for most systems.
4607 Description: OST crash if filesystem is unformatted or corrupt
4608 Details : If an OST is started on a device that has never been formatted
4609 or if the filesystem is corrupt and cannot even mount then the
4610 error handling cleanup routines would dereference a NULL pointer.
4615 Description: NULL pointer deref in llap_from_page.
4616 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4617 incorrect) mapping if the page was truncated/reclaimed while it was
4618 searched for. Check for this condition and skip such pages when
4619 doing readahead. Introduce extra check to llap_from_page() to
4620 verify page->mapping->host is non-NULL (so page is not anonymous).
4623 Frequency : Sometimes when using sys_sendfile
4625 Description: "page not covered by a lock" warnings from ll_readpage
4626 Details : sendfile called ll_readpage without right page locks present.
4627 Now we introduced ll_file_sendfile that does necessary locking
4628 around call to generic_file_sendfile() much like we do in
4632 Frequency : with certain MDS communication failures at client mount time
4634 Description: NULL pointer deref after failed client mount
4635 Details : a client connection request may delayed by the network layer
4636 and not be sent until after the PTLRPC layer has timed out the
4637 request. If the client fails the mount immediately it will try
4638 to clean up before the network times out the request. Add a
4639 reference from the request import to the obd device and delay
4640 the cleanup until the network drops the request.
4643 Frequency : occasionally during client (re)connect
4645 Description: assertion failure during client (re)connect
4646 Details : processing a client connection request may be delayed by the
4647 client or server longer than the client connect timeout. This
4648 causes the client to resend the connection request. If the
4649 original connection request is replied in this interval, the
4650 client may trip an assertion failure in ptlrpc_connect_interpret()
4651 which thought it would be the only running connect process.
4654 Frequency : only with obd_echo servers and clients that are rebooted
4656 Description: kernel BUG accessing uninitialized data structure
4657 Details : When running an obd_echo server it did not start the ping_evictor
4658 thread, and when a client was evicted an uninitialized data
4659 structure was accessed. Start the ping_evictor in the RPC
4660 service startup instead of the OBD startup.
4662 Severity : enhancement
4663 Bugzilla : 10193 (patchless)
4664 Description: Remove dependency on various unexported kernel interfaces.
4665 Details : No longer need reparent_to_init, exit_mm, exit_files,
4666 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4669 Frequency : rare (only users of deprecated and unsupported LDAP config)
4671 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4672 Details : LDAP apparently contains 'lustreName' attributes instead of
4673 'name'. A simple remapping of the name is sufficient.
4676 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4678 Description: waiting_locks_callback trips kernel BUG if client is evicted
4679 Details : Running with the dump_on_timeout debug flag turned on makes
4680 it possible that the waiting_locks_callback() can try to dump
4681 the Lustre kernel debug logs from an interrupt handler. Defer
4682 this log dumping to the expired_lock_main() thread.
4684 Severity : enhancement
4686 Description: Support NFS exporting on 2.6 kernels.
4687 Details : Implement non-rawops metadata methods for NFS server to use without
4688 changing NFS server code.
4691 Frequency : very rare (synthetic metadata workload only)
4693 Description: two racing renames might cause an MDS thread to deadlock
4694 Details : Running the "racer" program may cause one MDS thread to rename
4695 a file from being the source of a rename to being the target of
4696 a rename at exactly the same time that another thread is doing
4697 so, and the second thread has already enqueued these locks after
4698 doing a lookup of the target and is trying to relock them in
4699 order. Ensure that we don't try to re-lock the same resource.
4702 Frequency : only very large systems with liblustre clients
4704 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4705 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4706 all OSTs in parallel. This allows the eviction of stale liblustre
4707 clients to proceed much faster than if they were done in series,
4708 and also offers similar improvements for other set_info RPCs.
4713 Description: excessive CPU usage during initial read phase on client
4714 Details : During the initial read phase on a client, it would agressively
4715 retry readahead on the file, consuming too much CPU and impacting
4716 performance (since 1.4.5.8). Improve the readahead algorithm
4717 to avoid this, and also improve some other common cases (read
4718 of small files in particular, where "small" is files smaller than
4719 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4724 Description: MDS crash when receiving packet with unknown intent.
4725 Details : Do not LBUG in unknown intent case, just return -EFAULT
4727 Severity : enhancement
4728 Bugzilla : 9293, 9385
4729 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4730 Details : Do not serialize getattr (non-intent version) and statfs.
4733 Frequency : occasional, when OST network is overloaded/intermittent
4735 Description: client evicted by OST after bulk IO timeout
4736 Details : If a client sends a bulk IO request (read or write) the OST
4737 may evict the client if it is unresposive to its data GET/PUT
4738 request. This is incorrect if the network is overloaded (takes
4739 too long to transfer the RPC data) or dropped the OST GET/PUT
4740 request. There is no need to evict the client at all, since
4741 the pinger and/or lock callbacks will handle this, and the
4742 client can restart the bulk request.
4745 Frequency : Always when mmapping file with no objects
4747 Description: client crashes when mmapping file with no objects
4748 Details : Check that we actually have objects in a file before doing any
4749 operations on objects in ll_vm_open, ll_vm_close and
4755 Description: Request leak when working with deleted CWD
4756 Details : Introduce advanced request refcount tracking for requests
4757 referenced from lustre intent.
4759 Severity : Enhancement
4761 Description: Cache open file handles on client.
4762 Details : MDS now will return special lock along with openhandle, if
4763 requested and client is allowed to hold openhandle, even if unused,
4764 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4765 closing files for every read/write openration.
4767 Severity : Enhancement
4769 Description: Cache open negative dentries on client when possible.
4770 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4771 negative dentries on lock revocation.
4776 Description: Remounting a client read-only wasn't possible with a zconf mount
4777 Details : It wasn't possible to remount a client read-only with llmount.
4779 Severity : enhancement
4780 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4781 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4782 MPICH is included to add Lustre-specific ADIO interfaces.
4783 This is based closely on the UFS ADIO layer and only differs
4784 in file creation, in order to allow the OST striping to be set.
4785 This is user-contributed code and not supported by CFS.
4790 Description: extended inode attributes (immutable, append-only) work improperly
4791 when 2.4 and 2.6 kernels are used on client/server or vice versa
4792 Details : Introduce kernel-independent values for these flags.
4794 Severity : enhancement
4797 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4798 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4799 granularity. Now, Fractional MB granularity is be supported,
4800 this is very useful for low memory system.
4802 Severity : enhancement
4804 Description: Getattr by fid
4805 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4806 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4807 have locks and avoids getattr rpc requests for it on every stat.
4810 Frequency : Always, for filesystems larger than 2TB
4812 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4813 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4814 filesystems larger than 2TB to be mounted with mballoc enabled.
4817 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4819 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4820 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4821 filesystems larger than 2TB is mounted then the kernel will
4822 silently corrupt the start of the filesystem. CONFIG_LBD is
4823 enabled for all CFS-supported kernels, but the possibility of
4824 this happening with a modified kernel config exists.
4826 Severity : enhancement
4828 Description: add client O_DIRECT support for 2.6 kernels
4829 Details : It is now possible to do O_DIRECT reads and writes to files
4830 in the Lustre client mountpoint on 2.6 kernel clients.
4832 Severity : enhancement
4834 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4835 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4836 parallel, not waiting for response from every OST before sending
4837 a rpc to the next OST.
4842 Description: setattr vs write race when updating file timestamps
4843 Details : Client processes that update a file timestamp into the past
4844 right after writing to the file (e.g. tar) it is possible that
4845 the updated file modification time can be reset to the current
4846 time due to a race between processing the setattr and write RPC.
4848 Severity : enhancement
4850 Description: Bring 'lfs find' closer in line with regular Linux find.
4851 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4852 -print0 options and obtains all the needed info through the lustre
4855 Severity : enhancement
4857 Description: support up to 1024 configured devices on one node
4858 Details : change obd_dev array from statically allocated to dynamically
4859 allocated structs as they are first used to reduce memory usage
4864 Description: Flush dirty partially truncated pages during truncate
4865 Details : Immediatelly flush partially truncated pages in filter_setattr,
4866 this way we completely avoid having any pages in page cache on OST
4867 and can retire ugly workarounds during writes to flush such pages.
4872 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4873 Details : protect inode from truncation within vfs_unlink() context
4874 just take a reference before calling vfs_unlink() and release it
4875 when parent's i_sem is free.
4880 Description: last_id value checked outside lock on OST caused LASSERT failure
4881 Details : If there were multiple MDS->OST object precreate requests in
4882 flight, it was possible that the OST's last object id was checked
4883 outside a lock and incorrectly tripped an assertion. Move checks
4884 inside locks, and discard old precreate requests.
4887 Frequency : always, if extents are used on OSTs
4889 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4890 Details : OSTs using the extents format would not zero the high 16 bits of
4891 the index physical block number. This is not a problem for any
4892 OST filesystems smaller than 16TB, and no kernels support ext3
4893 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4894 new/modified files) and can be fixed for existing filesystems
4895 with e2fsprogs-1.39-cfs1.
4900 Description: import connection selection may be incorrect if timer wraps
4901 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4902 import connections to be ignored if the 32-bit jiffies counter
4903 wraps. Use a 64-bit jiffies counter.
4906 Frequency : very large clusters immediately after boot
4908 Description: LNET request buffers exhausted under heavy short-term load
4909 Details : If a large number of client requests are generated on a service
4910 that has previously never seen so many requests it is possible
4911 that the request buffer growth cannot keep up with the spike in
4912 demand. Instead of dropping incoming requests, they are held in
4913 the LND until the RPC service can accept more requests.
4916 Frequency : Sometimes during replay
4918 Description: Assertion failure in ll_local_open after replay.
4919 Details : If replay happened on an open request reply before we were able
4920 to set replay handler, reply will become not swabbed tripping the
4921 assertion in ll_local_open. Now we set the handler right after
4922 recognising of open request
4925 Frequency : very rare
4927 Description: kernel reports "badness in vsnprintf"
4928 Details : Reading from the "recovery_status" /proc file in small chunks
4929 may cause a negative length in lprocfs_obd_rd_recovery_status()
4930 call to vsnprintf() (which is otherwise harmless). Exit early
4931 if there is no more space in the output buffer.
4933 Severity : enhancement
4935 Description: clear OBD RPC statistics by writing to them
4936 Details : It is now possible to clear the OBD RPC statistics by writing
4937 to the "stats" file.
4942 Description: Client mtime is not the same on different clients after utimes
4943 Details : In some cases, the client was using the utimes() syscall on
4944 a file cached on another node. The clients now validate the
4945 ctime from the MDS + OSTs to determine which one is right.
4950 Description: Inability to activate failout mode
4951 Details : lconf script incorrectly assumed that in pythong string's numeric
4952 value is used in comparisons.
4955 Frequency : always with multiple stripes per file
4957 Description: Inefficient object allocation for mutli-stripe files
4958 Details : When selecting which OSTs to stripe files over, for files with
4959 a stripe count that divides evenly into the number of OSTs,
4960 the MDS is always picking the same starting OST for each file.
4961 Return the OST selection heuristic to the original design.
4966 Description: mount failures may take full timeout to return an error
4967 Details : Under some heavy load conditions it is possible that a
4968 failed mount can wait for the full obd_timeout interval,
4969 possibly several minutes, before reporting an error.
4970 Instead return an error as soon as the status is known.
4972 Frequency : quota enabled and large files being deleted
4974 Description: releasing more than 4GB of quota at once hangs OST
4975 Details : If a user deletes more than 4GB of files on a single OST it
4976 will cause the OST to spin in an infinite loop. Release
4977 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4982 Description: statfs data retrieved from /proc may be stale or zero
4983 Details : When reading per-device statfs data from /proc, in the
4984 {kbytes,files}_{total,free,avail} files, it may appear
4985 as zero or be out of date.
4988 Frequency : systems with MD RAID1 external journal devices
4990 Description: lconf's call to blkid is confused by RAID1 journal devices
4991 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4992 of returning all block devices that match the journal UUID.
4995 Frequency : always, for aggregate stripe size over 4GB
4997 Description: assertion fails when trying to use 4GB stripe size
4998 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4999 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
5002 Frequency : always on ppc64
5004 Description: the first write on an ext3 filesystem with mballoc got stuck
5005 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
5006 perform endianness conversion.
5008 ------------------------------------------------------------------------------
5010 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
5012 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
5013 INTEROPERATE with older versions automatically. Please read the
5014 user documentation before upgrading any part of a live system.
5015 * WARNING: Lustre networking configuration changes are required with
5016 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
5019 * Support for kernels:
5020 2.6.9-22.0.2.EL (RHEL 4)
5021 2.6.5-7.244 (SLES 9)
5022 2.6.12.6 vanilla (kernel.org)
5025 Severity : enhancement
5026 Bugzilla : 7981/8208
5027 Description: Introduced Lustre Networking (LNET)
5028 Details : LNET is new networking infrastructure for Lustre, it includes
5029 a reorganized network configuration mode (see the user
5030 documentation for full details) as well as support for routing
5031 between different network fabrics. Lustre Networking Devices
5032 (LNDs) for the supported network fabrics have also been
5033 created for this new infrastructure.
5035 Severity : enhancement
5036 Description: Introduced Access control lists
5037 Details : clients can set ACLs on files and directories in order to have
5038 more fine-grained permissions than the standard Unix UGO+RWX.
5039 The MDS must be started with the "-o acl" mount option.
5041 Severity : enhancement
5042 Description: Introduced filesystem quotas
5043 Details : Administrators may now establish per-user quotas on the
5046 Severity : enhancement
5048 Description: Configuration change for the XT3
5049 The PTLLND is now used to run Lustre over Portals on the XT3
5050 The configure option(s) --with-cray-portals are no longer used.
5051 Rather --with-portals=<path-to-portals-includes> is used to
5052 enable building on the XT3. In addition to enable XT3 specific
5053 features the option --enable-cray-xt3 must be used.
5058 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
5059 Details : A race between allocating a new llcd and re-getting the llcd_lock
5060 allowed another thread to grab newly-allocated llcd.
5062 Severity : enhancement
5064 Description: 2.6 OST async journal commit and locking fix to improve performance
5065 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
5066 2.6 kernels are now async as they already were in 2.4 kernels so
5067 that they can commit concurrently with the network bulk transfer.
5068 For block-allocated files the filter allocation semaphore is held
5069 to avoid filesystem fragmentation during allocation. BKL lock
5070 removed for 2.6 xattr operations where it is no longer needed.
5075 Description: lconf incorrectly determined whether two IP networks could talk
5076 Details : In some more complicated routing and multiple-network
5077 configurations, lconf will avoid trying to make a network
5078 connection to a disjoint part of the IP space. It was doing the
5079 math incorrectly for one set of cases.
5084 Description: Fix for potential infinite loop processing records in an llog.
5085 Details : If an llog record is corrupted/zeroed, it is possible to loop
5086 forever in llog_process(). Validate the llog record length
5087 and skip the remainder of the block on error.
5090 Frequency : occasional (liblustre only)
5092 Description: liblustre could not open files whose last component is a symlink
5093 Details : sysio_path_walk() would incorrectly pass the open intent to
5094 intermediate path components.
5097 Frequency : rare (liblustre only with non-standard tuning)
5098 Bugzilla : 7201 (7350)
5099 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
5100 Details : llu_lookup_finish_locks() tries to set lock data on a lock
5101 after it has been released, only do this for referenced locks
5103 Severity : enhancement
5105 Description: specifying an (invalid) directory default stripe_size of -1
5106 would reset the directory default striping
5107 Details : stripe_size -1 was used internally to signal directory stripe
5108 removal, now use "all default" to signal dir stripe removal
5109 as a directory striping of "all default" is not useful
5112 Frequency : common for large clusters running liblustre clients
5114 Description: doing an ls when liblustre clients are running is slow
5115 Details : sending a glimpse AST to a liblustre client waits for every AST
5116 to time out, as liblustre clients will not respond. Since they
5117 cannot cache data we refresh the OST lock LVB from disk instead.
5119 Severity : enhancement
5121 Description: doing an ls at the same time as file IO can be slow
5122 Details : enqueue and other "small" requests can be blocked behind many
5123 large IO requests. Create a new OST IO portal for non-IO
5124 requests so they can be processed faster.
5127 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
5129 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
5130 created file with mode 000
5131 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
5132 allow an ftruncate() as a non-root user to a file with mode 000.
5133 Lustre now respects this flag to disable mode checks when
5134 truncating a file owned by the user
5137 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
5140 Description: Revoking locks from clients that went dead or catatonic might take
5142 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
5143 cancellation of such locks instant on servers without waiting for
5144 any reply from clients. Clients drops these locks when cancel
5145 notification from server is received without replying.
5148 Frequency : liblustre-only, when liblustre client dies or becomes busy
5150 Description: Doing ls on Linux clients can take a long time with active
5152 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
5153 granting such locks to it in the first place if possible. Locks
5154 are taken by proxy on the OST during the read or write and
5155 dropped immediately afterward. Add connect flags handling, do
5156 not grant locks to liblustre clients for glimpse ASTs.
5158 Severity : enhancement
5160 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
5161 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5162 sequential reads and behaves badly with applications doing random
5163 reads. Improve it by reading ahead at least read region, and
5164 avoiding excessive large RPC for small reads.
5166 Severity : enhancement
5168 Description: Creating more than 1000 files for a single job may cause a load
5169 imbalance on the OSTs if there are also a large number of OSTs.
5170 Details : qos_prep_create() uses an OST index reseed value that is an
5171 even multiple of the number of available OSTs so that if the
5172 reseed happens in the middle of the object allocation it will
5173 still utilize the OSTs as uniformly as possible.
5178 Description: OST or MDS may oops in ping_evictor_main()
5179 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5180 but doesn't restart at beginning of obd_exports_timed list
5183 Severity : enhancement
5185 Description: improve by-nid export eviction on the MDS and OST
5186 Details : allow multiple exports with the same NID to be evicted at one
5187 time without re-searching the exports list.
5190 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5192 Description: MDS may oops in groups_free()
5193 Details : in rare race conditions a newly allocated group_info struct is
5194 freed again, and this can be NULL. The 2.4 compatibility code
5195 for groups_free() checked for a NULL pointer, but 2.6 did not.
5198 Frequency : common for liblustre clients doing little filesystem IO
5199 Bugzilla : 9352, 7313
5200 Description: server may evict liblustre clients accessing contended locks
5201 Details : if a client is granted a lock or receives a completion AST
5202 with a blocking AST already set it would not reply to the AST
5203 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5206 Frequency : lfs setstripe, only systems with more than 160 OSTs
5208 Description: unable to set striping with a starting offset beyond OST 160
5209 Details : llapi_create_file() incorrectly limited the starting stripe
5210 index to the maximum single-file stripe count.
5213 Frequency : LDAP users only
5215 Description: lconf did not handle in-kernel recovery with LDAP properly
5216 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5218 Severity : enhancement
5220 Description: bind OST threads to NUMA nodes to improve performance
5221 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5222 node and do their allocations there to localize memory access
5224 Severity : enhancement
5226 Description: llmount can determine client NID directly from Myrinet (GM)
5227 Details : the client NID code from gmnalnid was moved directly into
5228 llmount, removing the need to use this or specifying the
5229 client NID explicitly when mounting GM clients with zeroconf
5232 Frequency : if client is started with down MDS
5234 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5235 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5236 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5237 ptlrpc_import_delayed_req() because we didn't check if the
5238 request was interrupted, and we also didn't break out of the
5239 event loop if there was no timeout
5244 Description: data loss during non-page-aligned writes to a single file from
5245 both multiple nodes and multiple threads on one node at same time
5246 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5247 inconsistency led to false short-reads, that were cached and later
5248 used by ->prepare_write() to fill in partially written page,
5249 leading to data loss.
5252 Frequency : always, if lconf --abort_recovery used
5254 Description: lconf --abort_recovery fails with 'Operation not supported'
5255 Details : lconf was attempting to abort recovery on the MDT device and not
5258 Severity : enhancement
5260 Description: remove cleanup logs
5261 Details : replace lconf-generated cleanup logs with lustre internal
5262 cleanup routines. Eliminates the need for client-cleanup and
5265 Severity : enhancement
5267 Description: add support for EAs (user and system) on lustre filesystems
5268 Details : it is now possible to store extended attributes in the Lustre
5269 client filesystem, and with the user_xattr mount option it
5270 is possible to allow users to store EAs on their files also
5272 Severity : enhancement
5274 Description: Add possibility (config option) to show minimal available OST free
5276 Details : When compiled with --enable-mindf configure option, statfs(2)
5277 (and so, df) will return least minimal free space available from
5278 all OSTs as amount of free space on FS, instead of summary of
5279 free spaces of all OSTs.
5281 Severity : enhancement
5283 Description: do not expand extent locks acquired on OST-side
5284 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5285 by server: they are not cached anyway.
5288 Frequency : when mmap is used/binaries executed from Lustre
5290 Description: Unmmap pages before throwing them away from read cache.
5291 Details : llap_shrink cache now attempts to unmap pages before discarding
5292 them (if unmapping failed - do not discard). SLES9 kernel has
5293 extra checks that trigger if this unmapping is not done first.
5298 Description: lconf didn't resolve symlinks before checking to see whether a
5299 given mountpoint was already in use
5302 Frequency : when migrating failover services
5303 Bugzilla : 6395, 9514
5304 Description: When migrating a subset of services from a node (e.g. failback
5305 from a failover service node) the remaining services would
5306 time out and evict clients.
5307 Details : lconf --force (implied by --failover) sets the global obd_timeout
5308 to 5 seconds in order to quickly disconnect, but this caused
5309 other RPCs to time out too quickly. Do not change the global
5310 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5312 Severity : enhancement
5313 Frequency : if MDS is started with down OST
5314 Bugzilla : 9439,5706
5315 Description: Allow startup/shutdown of an MDS without depending on the
5316 availability of the OSTs.
5317 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5318 Add appropriate locking and lov-osc refcounts for safe
5319 cleaning. Add osc abort_inflight calls in case the
5320 synchronize never started.
5323 Frequency : occasional (Cray XT3 only)
5325 Description: root not authorized to access files in CRAY_PORTALS environment
5326 Details : The client process capabilities were not honoured on the MDS in
5327 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5328 been accessed by an authorized user then root was able to access
5329 the file on the local client also. The root user capabilities
5330 are now allowed on the MDS, as this environment has secure UID.
5333 Frequency : occasional
5335 Description: ldiskfs "too long searching" message happens too often
5336 Details : A debugging message (otherwise harmless) prints too often on
5337 the OST console. This has been reduced to only happen when
5338 there are fragmentation problems on the filesystem.
5343 Description: Division by zero in statfs when all OSCs are inactive
5344 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5345 lov_statfs divides by value returned by lov_get_stripecnt().
5349 Bugzilla : 9489, 3273
5350 Description: First write from each client to each OST was only 4kB in size,
5351 to initialize client writeback cache, which caused sub-optimal
5352 RPCs and poor layout on disk for the first written file.
5353 Details : Clients now request an initial cache grant at (re)connect time
5354 and so that they can start streaming writes to the cache right
5355 away and always do full-sized RPCs if there is enough data.
5356 If the OST is rebooted the client also re-establishes its grant
5357 so that client cached writes will be honoured under the grant.
5362 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5363 Details : Now I/O RPCs go to different portal number and (presumably) fast
5364 lock requests (and glimses) and other RPCs get their own service
5365 threads pool that should be able to service those RPCs
5368 Severity : enhancement
5370 Description: Ability to exchange lustre version between client and servers and
5371 issue warnings at client side if client is too old. Also for
5372 liblustre clients there is ability to refuse connection of too old
5374 Details : New 'version' field is added to connect data structure that is
5375 filled with version info. That info is later checked by server and
5379 Frequency : rare, liblustre only.
5380 Bugzilla : 9296, 9581
5381 Description: Two simultaneous writes from liblustre at offset within same page
5382 might proceed at the same time overwriting eachother with stale
5384 Details : I/O lock withing llu_file_prwv was released too early, before data
5385 actually was hitting the wire. Extended lock-holding time until
5386 server acknowledges receiving data.
5389 Frequency : extremely rare. Never observed in practice.
5391 Description: avoid generating lustre_handle cookie of 0.
5392 Details : class_handle_hash() generates handle cookies by incrementing
5393 global counter, and can hit 0 occasionaly (this is unlikely, but
5394 not impossible, because initial value of cookie counter is
5395 selected randonly). Value of 0 is used as a sentinel meaning
5396 "unassigned handle" --- avoid it. Also coalesce two critical
5397 sections in this function into one.
5399 Severity : enhancement
5401 Description: allow liblustre clients to delegate truncate locking to OST
5402 Details : To avoid overhead of locking, liblustre client instructs OST to
5403 take extent lock in ost_punch() on client's behalf. New connection
5404 flag is added to handle backward compatibility.
5406 Severity : enhancement
5407 Bugzilla : 4928, 7341, 9758
5408 Description: allow number of OST service threads to be specified
5409 Details : a module parameter allows the number of OST service threads
5410 to be specified via "options ost ost_num_threads={N}" in the
5411 OSS's /etc/modules.conf or /etc/modprobe.conf.
5415 Bugzilla : 6146, 9635, 9895
5416 Description: servers crash with bad pointer in target_handle_connect()
5417 Details : In rare cases when a client is reconnecting it was possible that
5418 the connection request was the last reference for that export.
5419 We would temporarily drop the export reference and get a new
5420 one, but this may have been the last reference and the export
5421 was just destroyed. Get new reference before dropping old one.
5423 Severity : enhancement
5424 Frequency : if client is started with failover MDS
5426 Description: Allow multiple MDS hostnames in the mount command
5427 Details : Try to read the configuration from all specified MDS
5428 hostnames during a client mount in case the "primary"
5431 Severity : enhancement
5433 Description: Stop sending data to evicted clients as soon as possible.
5434 Details : Check if the client we are about to send or are sending data to
5435 was evicted already. (Check is done every second of waiting,
5436 for which l_wait_event interface was extended to allow checking
5437 of exit condition at specified intervals).
5440 Frequency : rare, normally only when NFS exporting is done from client
5442 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5444 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5447 Frequency : always, if srand() is called before liblustre initialization
5449 Description: Liblustre uses system PRNG disturbing its usage by user application
5450 Details : Introduce internal to lustre fast and high-quality PRNG for
5451 lustre usage and make liblustre and some other places in generic
5452 lustre code to use it.
5454 Severity : enhancement
5455 Bugzilla : 9477, 9557, 9870
5456 Description: Verify that the MDS configuration logs are updated when xml is
5457 Details : Check if the .xml configuration logs are newer than the config
5458 logs stored on the MDS and report an error if this is the case.
5459 Request --write-conf, or allow starting with --old_conf.
5461 Severity : enhancement
5463 Description: Handle symlinks in the path when checking if Lustre is mounted.
5464 Details : Resolve intermediate symlinks when checking if a client has
5465 mounted a filesystem to avoid duplicate client mounts.
5470 Description: lconf can hit an error exception but still return success.
5471 Details : The lconf command catches the Command error exception at the top
5472 level script context and will exit with the associated exit
5473 status, but doesn't ensure that this exit status is non-zero.
5478 Description: failure of ptlrpc thread startup can cause oops
5479 Details : Starting a ptlrpc service thread can fail if there are a large
5480 number of threads or the server memory is very fragmented.
5481 Handle this without oopsing.
5484 Frequency : always, only if liblustre and non-default acceptor port was used
5486 Description: liblustre cannot connect to servers with non-default acceptor port
5487 Details : tcpnal_set_default_params() was not called and was therefore
5488 ignoring the environment varaible TCPNAL_PORT, as well as other
5489 TCPNAL_ environment variables
5494 Description: two objects could be created on the same OST for a single file
5495 Details : If an OST is down, in some cases it was possible to create two
5496 objects on a single OST for a single file. No problems other
5497 than potential performance impact and spurious error messages.
5501 Bugzilla : 5681, 9562
5502 Description: Client may oops in ll_unhash_aliases
5503 Details : Client dcache may become inconsistent in race condition.
5504 In some cases "getcwd" can fail if the current directory is
5510 Description: Inode refcounting problems in NFS export code
5511 Details : link_raw functions used to call d_instantiate without obtaining
5512 extra inode reference first.
5516 Bugzilla : 9942, 9903
5517 Description: Referencing freed requests leading to crash, memleaks with NFS.
5518 Details : We used to require that call to ll_revalidate_it was always
5519 followed by ll_lookup_it. Also with revalidate_special() it is
5520 possible to call ll_revalidate_it() twice for the same dentry
5521 even if first occurence returned success. This fix changes semantic
5522 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5523 reference on a request referred from the intent.
5524 ll_intent_release() then releases such a request.
5527 Frequency : rare, normally benchmark loads only
5529 Description: unlinked inodes were kept in memory on the client
5530 Details : If a client is repeatedly creating and unlinking files it
5531 can accumulate a lot of stale inodes in the inode slab cache.
5532 If there is no other client load running this can cause the
5533 client node to run out of memory. Instead flush old inodes
5534 from client cache that have the same inode number as a new inode.
5537 Frequency : SLES9 2.6.5 kernel and long filenames only
5538 Bugzilla : 9969, 10379
5539 Description: utime reports stale NFS file handle
5540 Details : SLES9 uses out-of-dentry names in some cases, which confused
5541 the lustre dentry revalidation. Change it to always use the
5545 Frequency : rare, unless heavy write-truncate concurrency is continuous
5546 Bugzilla : 4180, 6984, 7171, 9963, 9331
5547 Description: OST becomes very slow and/or deadlocked during object unlink
5548 Details : filter_destroy() was holding onto the parent directory lock
5549 while truncating+unlinking objects. For very large objects this
5550 may block other threads for a long time and slow overall OST
5551 responsiveness. It may also be possible to get a lock ordering
5552 deadlock in this case, or run out of journal credits because of
5553 the combined truncate+unlink. Solution is to do object truncate
5554 first in one transaction without parent lock, and then do the
5555 final unlink in a new transaction with the parent lock. This
5556 reduces the lock hold time dramatically.
5559 Frequency : rare, 2.4 kernels only
5561 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5562 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5563 before unmounting their filesystem in order to avoid deadlock.
5564 The kernel_locked() function in 2.4 kernels only checks whether
5565 the kernel lock is held, not whether it is this process that is
5566 holding it as 2.6 kernels do.
5571 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5572 Details : The client ptlrpc code may be trying to reconnect to a down
5573 server before a previous connection attempt has timed out.
5574 Increase the reconnect interval to be longer than the connection
5575 timeout interval to avoid sending duplicate connections to
5579 Frequency : echo_client brw_test command
5581 Description: fix echo_client to work with OST preallocated code
5582 Details : OST preallocation code (5137) didn't take echo_client IO path
5583 into account: echo_client calls filter methods outside of any
5584 OST thread and, hence, there is no per-thread preallocated
5585 pages and buffers to use. Solution: hijack pga pages for IO. As
5586 a byproduct, this avoids unnecessary data copying.
5590 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5591 Description: Client can oops in mdc_commit_close() after open replay
5592 Details : It was possible for the MDS to return an open request with no
5593 transaction number in mds_finish_transno() if the client was
5594 evicted, but without actually returning an error. Clients
5595 would later try to replay that open and may trip an assertion
5596 Simplify the client close codepath, and always return an error
5597 from the MDS in case the open is not successful.
5600 Frequency : rare, 2.6 OSTs only
5602 Description: OST may deadlock under high load on fragmented files
5603 Details : If there was a heavy load and highly-fragmented OST filesystems
5604 it was possible to have all the OST threads deadlock waiting on
5605 allocation of biovecs, because the biovecs were not released
5606 until the entire RPC IO was completed. Instead, release biovecs
5607 as soon as they are complete to ensure forward IO progress.
5609 Severity : enhancement
5611 Description: Support for specifying external journal device at mount
5612 Details : If an OST or MDS device is formatted with an external journal
5613 device, this device major/minor is stored in the ext3 superblock
5614 and may not be valid for failover. Allow detecting and
5615 specifying the external journal at mount time.
5620 Description: Mounting an MDS with pending unlinked files may cause oops
5621 Details : target_finish_recovery() calls mds_postrecov() which returned
5622 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5623 considers this an error and immediately begins cleaning up the
5624 lov, just after starting the mds_lov process
5626 Severity : enhancement
5628 Description: Implement 'lfs df' to report actual free space on per-OST basis
5629 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5630 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5631 usage of objects; '-h' to report in human readable format.
5633 ------------------------------------------------------------------------------
5635 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5642 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5643 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5644 at startup during orphan unlinking. Instead, skip buddy update
5645 before setup, as it will be regenerated after recovery is complete.
5650 Description: If an OST is inactive, its locks might reference stale inodes.
5651 Details : lov_change_cbdata() must iterate over all namespaces, even if
5652 they are inactive to clear inode references from the lock.
5654 Severity : enhancement
5655 Frequency : occasional, if non-standard max_dirty_mb used
5657 Description: Client will block write RPCs if not enough grant
5658 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5659 then the client will block writes while waiting for another RPC
5660 to complete instead of consuming its dirty limit. With change
5661 we get improved performance when max_dirty_mb is small.
5663 Severity : enhancement
5664 Bugzilla : 3389, 6253
5665 Description: Add support for supplementary groups on the MDS.
5666 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5667 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5668 lookups for user supplementary groups into a cache.
5672 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5673 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5674 for liblustre O_EXCL is enforced by MDS.
5678 Description: readdir never returns NULL in liblustre.
5679 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5680 dentry in struct dirent.
5684 Frequency : liblustre only, depends on application IO pattern
5685 Description: liblustre clients evicted if not contacting servers
5686 Details : Don't put liblustre clients into the ping_evictor list, so
5687 they will not be evicted by the pinger ever.
5689 Severity : enhancement
5691 Description: Add ability to evict clients by NID from MDS.
5692 Details : By echoing "nid:$NID" string into
5693 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5694 $NID would be instantly evicted from this MDS and from all active
5695 OSTs connected to it.
5699 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5700 Details : lookup_it_finish() used to query file size from OSTs that was not
5705 Description: service threads change working directory to that of init
5706 Details : Starting lustre service threads may pin the working directory
5707 of the parent thread, making that filesystem busy. Threads
5708 now change to the working directory of init to avoid this.
5712 Frequency : during shutdown only
5713 Description: shutdown with a failed MDS or OST can cause unmount to hang
5714 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5719 Frequency : 2.6 only
5720 Description: chmod/chown may include an extra supplementary group
5721 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5722 supplementary group and if none is specified this is used.
5725 Bugzilla : 5479 (6816)
5727 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5728 Details : If lookup is in progress on a file that is unlinked we might try
5729 to revalidate the inode and fail in revalidate after lookup is
5730 complete and ll_file_open() enqueues the open again but
5731 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5734 Frequency : always, if lconf --abort_recovery used
5736 Description: lconf --abort_recovery fails with 'Operation not supported'
5737 Details : lconf was attempting to abort recovery on the MDT device and not
5740 ------------------------------------------------------------------------------
5742 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5747 Frequency : rare (only unsupported configurations with a node running as an
5749 Bugzilla : 6514, 5137
5750 Description: Mounting a Lustre file system on a node running as an OST could
5752 Details : OSTs now preallocates memory needed to write out data at
5753 startup, instead of when needed, to avoid having to
5754 allocate memory in possibly low memory situations.
5755 Specifically, if the file system is mounted on on OST,
5756 memory pressure could force it to try to write out data,
5757 which it needed to allocate memory to do. Due to the low
5758 memory, it would be unable to do so and the node would
5759 become unresponsive.
5761 Severity : enhancement
5763 Description: Addition of lconf --service command line option
5764 Details : lconf now accepts a '--service <arg>' option, which is
5765 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5767 Severity : enhancement
5769 Description: Failover mode is now the default for OSTs.
5770 Details : By default, OSTs will now run in failover mode. To return to
5771 the old behaviour, add '--failout' to the lmc line for OSTs.
5773 Severity : enhancement
5775 Description: Health checks are now provided for MDS and OSTs
5776 Details : Additional detailed health check information on MSD and OSTs
5777 is now provided through the procfs health_check value.
5780 Frequency : occasional, depends on IO load
5782 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5783 after numerous create/delete cycles
5784 Details : The ext3 inode allocation policy would not allocate new inodes
5785 very well on the OSTs because there are no new directories
5786 being created. Instead we look for groups with free space if
5787 the parent directories are nearly full.
5792 Description: Network or server problems during mount may cause partially
5793 mounted clients instead of returning an error.
5794 Details : The config llog parsing code may overwrite the error return
5795 code during mount error handling, returning success instead
5801 Description: MDS can fail to allocate large reply buffers
5802 Details : After long uptimes the MDS can fail to allocate large reply
5803 buffers (e.g. zconf client mount config records) due to memory
5804 fragmentation or consumption by the buffer cache. Preallocate
5805 some large reply buffers so that these replies can be sent even
5806 under memory pressure.
5810 Frequency : rare (liblustre)
5811 Description: fsx running with liblustre complained that using truncate() to
5812 extend the file doesn't work. This patch corrects that issue.
5813 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5814 ATTR_SIZE and lsm use in llu_setattr_raw.
5818 Frequency : rare, only 2.6 kernels
5819 Description: Unusual file access patterns on the MDS may result in inode
5820 data being lost in very rare circumstances.
5821 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5822 code in the 2.6 kernel caused the inode and/or EA data not to
5823 be read from disk, causing single-file corruption.
5827 Frequency : rare, only 2.6 filesystems using extents
5828 Description: Heavy concurrent write and delete load may cause data corruption.
5829 Details : It was possible under high-load situations to have an extent
5830 metadata block in the block device cache from a just-unlinked
5831 file overwrite a newly-allocated data block. We now unmap any
5832 metadata buffers that alias just-allocated data blocks.
5836 Frequency : filesystems with default stripe_count larger than 77
5837 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5838 Details : lconf specifies an inode size of 4096 bytes when the default
5839 stripe_count is larger than 77. This conflicts with the default
5840 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5841 this case to avoid pinning too much memory for large EAs.
5843 ------------------------------------------------------------------------------
5845 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5850 Frequency : rare (extremely heavy IO load with hundreds of clients)
5852 Description: Client is evicted, gets IO error writing to file
5853 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5854 caused clients to be evicted for AST timeouts. The fixes in
5855 bug 5192 mean we no longer need to have such short AST timeouts
5856 so ldlm_timeout has been increased.
5859 Frequency : occasional during --force or --failover shutdown under load
5860 Bugzilla : 5949, 4834
5861 Description: Server oops/LBUG if stopped with --force or --failover under load
5862 Details : a collection of import/export refcount and cleanup ordering
5863 issues fixed for safer force cleanup
5866 Frequency : only filesystems larger than 120 OSTs
5867 Bugzilla : 5990, 6223
5868 Description: lfs getstripe would oops on a very large filesystem
5869 Details : lov_getconfig used kfree on vmalloc'd memory
5872 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5873 Bugzilla : 6242, 6243
5874 Description: reading from files that had been truncated to a non-zero size
5875 but never opened returned no data
5876 Details : ll_file_read() reads zeros from no-object files to EOF
5881 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5882 Details : There's little harm in aborting MDS/OSS recovery and letting it
5883 try again, so I removed the LASSERT and return an error instead.
5885 Severity : enhancement
5887 Description: New debugging infrastructure for tracking down data corruption
5888 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5889 (b) cover more of the client-side code path, and (c) try to narrow
5890 down where problems occurred
5894 Bugzilla : 3819, 4364, 4397, 6313
5895 Description: Racing close and eviction MDS could cause assertion in mds_close
5896 Details : It was possible to get multiple mfd references during close and
5897 client eviction, leading to one thread referencing a freed mfd.
5899 Severity: : enhancement
5900 Bugzilla : 3262, 6359
5901 Description: Attempts to reconnect to servers are now more aggressive.
5902 Details : This builds on the enhanced upcall-less recovery that was added
5903 in 1.4.2. When trying to reconnect to servers, clients will
5904 now try each server in the failover group every 10 seconds. By
5905 default, clients would previously try one server every 25 seconds.
5910 Description: After recovery, certain operations trigger a failed
5911 assertion on a client.
5912 Details : Failing over an mds, using lconf -d --failover, while a
5913 client was doing a readdir() call would cause the client to
5914 LBUG after recovery completed and the readdir() was resent.
5916 Severity : enhancement
5918 Description: Default groups are now added by lconf
5919 Details : You can now run lconf --group <servicename> without having to
5920 manually add groups with lmc.
5923 Frequency : occasional
5925 Description: Nodes with an elan id of 0 trigger a failed assertion
5928 Frequency : always when accessing e.g. tty/console device nodes
5930 Description: tty and some other devices nodes cannot be used on lustre
5931 Details : file's private_data field is used by device data and lustre
5932 values in there got lost. New field was added to struct file to
5933 store fs-specific private data.
5936 Frequency : when exporting Lustre via NFS
5938 Description: NFSD failed occasionally when looking up a path component
5939 Details : NFSD is looking up ".." which was broken in ext3 directories
5940 that had grown large enough to become hashed.
5943 Frequency : Clusters with multiple interfaces not on the same subnet
5945 Description: Nodes will repeatedly try to reconnect to an interface which it
5946 cannot reach and report an error to the log.
5947 Details : Extra peer list entries will be created by lconf with some peers
5948 unreachable. lconf now validates the peer before adding it.
5951 Frequency : Only if a default stripe is set on the filesystem root.
5953 Description: Setting a default stripe on the filesystem root prevented the
5954 filesystem from being remounted.
5955 Details : The client was sending extra request flags in the root getattr
5956 request and did not allocate a reply buffer for the dir EA.
5959 Frequency : occasional, higher if lots of files are accessed by one client
5960 Bugzilla : 6159, 6097
5961 Description: Client trips assertion regarding lsm mismatch/magic
5962 Details : While revalidating inodes the VFS looks up inodes with ifind()
5963 and in rare cases can find an inode that is being freed.
5964 The ll_test_inode() code will free the lsm during ifind()
5965 when it finds an existing inode and then the VFS later attaches
5966 this free lsm to a new inode.
5970 Bugzilla : 6422, 7030
5971 Description: MDS deadlock between mkdir and client eviction
5972 Details : Creating a new file via mkdir or mknod (starting a transaction
5973 and getting the ns lock) can deadlock with client eviction
5974 (gets ns lock and trying to finish a synchronous transaction).
5977 Frequency : occasional
5978 Description: While starting a server, the fsfilt_ext3 module could not be
5980 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5981 kernels. Previously, lconf would still use the ext3 name
5982 when trying to load modules. Now, it will correctly use
5983 ext3 on 2.4 and ldiskfs on 2.6.
5985 Severity : enhancement
5986 Description: The default stripe count has been changed to 1
5987 Details : The interpretation of the default stripe count (0, to lfs
5988 or lmc) has been changed to mean striping across a single
5989 OST, rather than all available. For general usage we have
5990 found a stripe count of 1 or 2 works best.
5992 Severity : enhancement
5993 Description: Add support for compiling against Cray portals.
5994 Details : Conditional compiling for some areas that are different
5998 Frequency : occasional
5999 Bugzilla : 6409, 6834
6000 Description: Creating files with an explicit stripe count may lead to
6001 a failed assertion on the MDS
6002 Details : If some OSTs are full or unavailable, creating files may
6003 trigger a failed assertion on the MDS. Now, Lustre will
6004 try to use other servers or return an error to the
6008 Frequency : occasional
6010 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
6011 caused lock timeout during readahead (since 1.4.2).
6012 Details : Processes doing readahead might match a lock that hasn't been
6013 granted yet if there are overlapping and conflicting lock
6014 requests. The readahead process waits on ungranted lock
6015 (original lock is CBPENDING), while OST waits for that process
6016 to cancel CBPENDING read lock and eventually evicts client.
6018 Severity : enhancement
6020 Description: Initial enabling of flock support for clients
6021 Details : Implements fcntl advisory locking and file status functions.
6022 This feature is provided as an optional mount flag (default
6023 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
6024 locking are implemented yet, and those that are are not guaranteed
6025 to be completely correct in production environments.
6026 mount -t lustre -o [flock|noflock] ...
6029 Frequency : occasional
6031 Description: OSTs running 2.4 kernels but with extents enabled might trip an
6032 assertion in the ext3 JBD (journaling) layer.
6033 Details : The b_committed_data struct is protected by the big kernel lock
6034 in 2.4 kernels, serializing journal_commit_transaction() and
6035 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
6036 access to this struct. In 2.6 kernels there is finer grained
6037 locking to improve SMP performance of the JBD layer.
6041 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
6046 Description: merge and cleanup kernel patches.
6049 -----------------------------------------------------------------------------
6051 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
6053 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
6054 versions of Lustre. Please update all servers and clients to
6055 version 1.4.2 or later at the same time. You must also run
6056 "lconf --write-conf {config}.xml" on the MDS while it is stopped
6057 to update the configuration logs.
6059 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
6060 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
6061 - swab llog records of type '0' so we get proper header size/idx (5861)
6062 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
6063 - fix rename of one directory over another leaking an inode (5953)
6064 - avoid SetPageDirty on 2.6 (5981)
6065 - don't re-add just-being-destroyed locks to the waiting list (5653)
6066 - when creating new directories, inherit the parent's custom
6067 striping settings if present parent (3048)
6068 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
6069 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
6070 - don't hold client locks on temporary worklist from l_lru (5666)
6071 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
6072 - automatically evict dead clients (5921)
6073 - Update file size properly in create+truncate+fstat case (6196)
6074 - Do not unhash mountpoint dentries, do not allow removal of
6076 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
6077 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
6078 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
6079 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
6080 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
6081 - Correctly handle reads of files with no objects (6243)
6082 - lctl recover will also mark a device active if deactivate used (5933)
6084 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
6085 - allow --write-conf on an MDS with different nettype than client (5619)
6086 - don't write config llogs to MDS for mounts not from that MDS (5617)
6087 - lconf should create multiple TCP connections from a client (5201)
6088 - init scripts are now turned off by default; run chkconfig --on
6089 lustre and chkconfig --on lustrefs to use them
6090 - upcalls are no longer needed for clients to recover to failover
6092 - add --abort-recovery option to lconf to abort recovery on device
6094 - add support for an arbitrary number of OSTs (3026)
6095 - Quota support protocol changes.
6096 - forward compatibility changes to wire structs (6007)
6097 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
6098 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
6099 - improved "lustre status" script
6100 - initialize blocksize for non-regular files (6062)
6101 - added --disable-server and --disable-client configure options (5782)
6102 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
6103 - Vanilla 2.4.29 support
6104 - increase maximum number of obd devices to 520 (6242)
6105 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
6106 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
6107 SLES 9 2.6.5 kernels
6108 - limit stripes per file to 160 (the maximum EA size) (6093)
6110 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6113 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
6114 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
6115 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
6116 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
6117 - small ext3 extents cleanups and fixes (5733)
6118 - improved mballoc code, several small races and bugs fixed (5733, 5638)
6119 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
6120 - avoid needless client->OST connect, fix handle mismatch (5317)
6121 - fix DLM error path that led to out-of-sync client, long delays (5779)
6122 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
6123 - fix several locking issues related to i_size (5492,5624,5654,5672)
6124 - don't move pending lock onto export if it is already evicted (5683)
6125 - fix kernel oops when creating .foo in unlinked directory (5548)
6126 - fix deadlock in obdfilter statistics vs. object create (5811)
6127 - use time_{before,after} to avoid timer jiffies wrap (5882)
6128 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
6129 - Do not leak request if server was not able to process it (5154)
6130 - If mds_open unable to find parent dir, make that negative lookup(5154)
6131 - don't create new directories with extent-mapping (5909, 5936)
6133 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
6134 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
6135 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
6136 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
6137 - support for RHEL4 kernel on i686 (5773)
6138 - provide error messages when incompatible logs are encountered (5898)
6140 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
6141 * version 1.4.0.10 (1.4.1 release candidate 1)
6143 - don't keep a lock reference when lock is not granted (4238)
6144 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6145 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
6146 - drop import inflight refcount on signal_completed_replay error (5255)
6147 - unlock page after async write error during send (3677)
6148 - handle missing objects in filter_preprw_read properly (5265)
6149 - no transno return for symlink open, don't save no-trasno open (3440)
6150 - don't try to complete elan receive that already failed (4012)
6151 - free RPC server reply state on error (5406)
6152 - clean up thread from ptlrpc_start_thread() on error (5160)
6153 - readahead could read extra page into cache that wasn't ejected (5388)
6154 - prevent races in class_attach/setup/cleanup/detach (5260)
6155 - don't dereference de->d_inode after l_dput of de (5458)
6156 - use "int" for stripe value returned from lock_to_stripe (5544)
6157 - mballoc allocation and error-checking fixes in 2.6 (5504)
6158 - block device patches to fix I/O request sizes in 2.6 (5482)
6159 - look up hostnames for IB nals (5602)
6160 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
6161 - don't start multiple acceptors for the same port (5277)
6162 - fix incorrect LASSERT in mds_getattr_name (5635)
6163 - export a proc file for general "ping" checking (5628)
6164 - fix "lfs check" to not block when the MDS is down (5628)
6166 - service request history (4965)
6167 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6168 - create an "evict_client" /proc entry on OSTs, like the MDS has
6169 - fix mount usage message, return errors per mount(8) (5168)
6170 - change grep [] to grep "[]" in tests so they work in more UMLs
6171 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6172 - remove ancient LOV_MAGIC_V0 EA support (5047)
6173 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6174 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6175 - allow readahead allocations to fail when low on memory (5383)
6176 - mmap locking landed again, after considerable improvement (2828)
6177 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6179 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6182 - send OST transaction number in read/write reply to free req (4966)
6183 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6184 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6185 - the watchdog thread now runs as interruptible (5246)
6186 - flock/lockf fixes (but it's still disabled, pending 5135)
6187 - don't use EXT3 constants in llite code (5094)
6188 - memory shortage at startup could cause assertion (5176)
6190 - reorganization of lov code
6191 - single portals codebase
6193 - add extents/mballoc support (5025)
6194 - direct I/O reads in the obdfilter (4048)
6195 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6197 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6200 - send OST transaction number in read/write reply to free req (4966)
6201 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6202 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6203 - the watchdog thread now runs as interruptible (5246)
6204 - handle missing objects in filter_preprw_read properly (5265)
6205 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6206 - drop import inflight refcount on signal_completed_replay error (5255)
6207 - unlock page after async write error during send (3677)
6208 - return original error code on reconstructed replies (3761)
6209 - no transno return for symlink open, don't save no-trasno open (3440)
6211 - add pid to ldlm debugging output (4922)
6212 - bump the watchdog timeouts -- we can't handle 30sec yet
6213 - extra debugging for orphan dentry/inode bug (5259)
6215 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6218 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6219 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6220 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6221 - fix NULL dereference in /proc/sys/portals/routes (4827)
6222 - allow failed mdc_close() operations to be interrupted (4561)
6223 - stop precreate on OST before MDS would time out on it (4778)
6224 - don't send partial-page writes before EOF from client (4410)
6225 - discard client grant for sub-page writes on large-page clients (4520)
6226 - don't free dentries not owned by NFS code, check generation (4806)
6227 - fix lsm leak if mds_create_objects() fails (4801)
6228 - limit debug_daemon file size, always print CERROR messages (4789)
6229 - use transno after validating reply (3892)
6230 - process timed out requests if import state changes (3754)
6231 - update mtime on OST during writes, return in glimpse (4829)
6232 - add mkfsoptions to LDAP (4679)
6233 - use ->max_readahead method instead of zapping global ra (5039)
6234 - don't interrupt __l_wait_event() during strace
6236 - add software watchdogs to catch hung threads quickly (4941)
6237 - make lustrefs init script start after nfs is mounted
6238 - fix CWARN/ERROR duplication (4930)
6239 - return async write errors to application if possible (2248)
6240 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6241 - print NAL number in %x format (4645)
6242 - update barely-supported suse-2.4.21-171 series (4842)
6243 - support for sles 9 %post scripts
6244 - support for building 2.6 kernel-source packages
6245 - support for sles km_* packages
6247 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6250 - ignore -ENOENT errors in osc_destroy (3639)
6251 - notify osc create thread that OSC is being cleaned up (4600)
6252 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6253 - reconstruct ost_handle() like mds_handle() (4657)
6254 - create a new thread to do import eviction to avoid deadlock (3969)
6255 - let lconf resolve symlinked-to devices (4629)
6256 - don't unlink "objects" from directory with default EA (4554)
6257 - hold socknal file ref over connect in case target is down (4394)
6258 - allow more than 32000 subdirectories in a single directory (3244)
6259 - fix blocks count for O_DIRECT writes (3751)
6260 - OST returns ENOSPC from object create when no space left (4539)
6261 - don't send truncate RPC if file size isn't changing (4410)
6262 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6263 - bind to privileged port in socknal and tcpnal (3689)
6265 - rate limit CERROR/CWARN console message to avoid overload (4519)
6266 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6267 - basic mmap support (3918)
6268 - kernel patch series update from b1_4 (4711)
6270 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6273 - avoid crash during MDS cleanup with OST shut down (2775)
6274 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6275 - don't use bad inodes on the MDS (3744)
6276 - dynamic object preallocation to improve recovery speed (4236)
6277 - don't hold spinlock over lock dumping or change debug flags (4401)
6278 - don't zero obd_dev when it is force cleaned (3651)
6279 - print grants to console if they go negative (4431)
6280 - "lctl deactivate" will stop automatic recovery attempts (3406)
6281 - look for existing locks in ldlm_handle_enqueue() (3764)
6282 - don't resolve lock handle twice in recovery avoiding race (4401)
6283 - revalidate should check working dir is a directory (4134)
6285 - don't always mark "slow" obdfilter messages as errors (4418)
6287 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6290 - don't close LustreDB during write_conf until it is done (3860)
6291 - fix typo in lconf for_each_profile (3821)
6292 - allow dumping logs from multiple threads at one time (3820)
6293 - don't allow multiple threads in OSC recovery (3812)
6294 - fix debug_size parameters (3864)
6295 - fix mds_postrecov to initialize import for llog ctxt (3121)
6296 - replace config semaphore with spinlock (3306)
6297 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6298 - don't allow enqueue to complete on a destroyed export (3822)
6299 - down write_lock before checking llog header bitmap (3825)
6300 - recover from lock replay timeout (3764)
6301 - up llog sem before sending rpc (3652)
6302 - reduce ns lock hold times when setting kms (3267)
6303 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6304 - fix NULL deref and obd_dev leak on setup error (3312)
6305 - replace some LBUG about llog ops with error handling (3841)
6306 - don't match INVALID dentries from d_lookup and spin (3784)
6307 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6308 - fix invalid assertion in ptlrpc_set_wait (3880)
6310 - add libwrap support for the TCP acceptor (3996)
6311 - add /proc/sys/portals/routes for non-root route listing (3994)
6312 - allow setting MDS UUID in .xml (2580)
6313 - print the stack of a process that LBUGs (4228)
6315 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6318 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6319 - make sure to unset replay flag from failed open requests (3440)
6320 - if default stripe count is 0, use OST count for inode size (3636)
6321 - update parent mtime/ctime on client for create/unlink (2611)
6322 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6323 - free recovery state on server during a forced cleanup (3571)
6324 - unregister_reply for resent reqs (3063)
6325 - loop back devices mounting and status check on 2.6 (3563)
6326 - fix resource-creation race that can provoke i_size == 0 (3513)
6327 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6328 - more debugging for page-accounting assertion (3746)
6329 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6330 - avoid deadlock after precreation failure (3758)
6331 - fix race and lock order deadlock in orphan handling (3450, 3750)
6332 - add validity checks when grabbing inodes from l_ast_data (3599)
6334 - add /proc/.../recovery_status to obdfilter (3428)
6335 - lightweight CDEBUG infrastructure, debug daemon (3668)
6336 - change default OSC RPC parameters to be better on small clusters
6337 - turn off OST read cache for files smaller than 32MB
6338 - install man pages and include them in rpms (3100)
6339 - add new init script for (un)mounting lustre filesystems (2593)
6340 - run chkconfig in %post for init scripts (3701)
6341 - drop scimac NAL (unmaintained)
6343 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6346 - clean kiobufs before and after use (3485)
6347 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6348 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6349 - change init script to fail more clearly if not run as root (1528)
6350 - allow clients to reconnect during replay (1742)
6351 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6352 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6353 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6354 - fix cond_resched (3554)
6355 - only evict unfinished clients after recovery (3515)
6356 - allow bulk resend, prevent data loss (3570)
6357 - dynamic ptlrpc request buffer allocation (2102)
6358 - don't allow unlinking open directory if it isn't empty (2904)
6359 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6360 - remove extraneous obd dereference causing LASSERT failure (3334)
6361 - don't use get_cycles() when creating temp. files on the mds (3156)
6362 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6363 - handle EEXIST for set-stripe, set proper directory name (3336)
6365 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6366 - fix ksocknal_fmb_callback() error messages (2918)
6368 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6371 - don't copy lvb into (possibly NULL) reply on error (2983)
6372 - don't deref dentry after dput, don't free lvb on error (2922)
6373 - use the kms to determine writeback rpc length (2947)
6374 - increment oti_logcookies when osc is inactive (2948)
6375 - update client's i_blocks count via lvb messages (2543)
6376 - handle intent open/close of special files properly (1557)
6377 - mount MDS with errors=remount-ro, like obdfilter (2009)
6378 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6379 - don't use cancelling-locks' kms values (2947)
6380 - use highest lock extent for kms, not last one (2925)
6381 - don't dereference ERR_PTR() dentry in error handling path (3107)
6382 - fix thread race in portals_debug_dumplog() (3122)
6383 - create lprocfs device entries at setup instead of at attach (1519)
6384 - common AST error handler, don't evict client on completion race (3145)
6385 - zero nameidata in detach_mnt in 2.6 (3118)
6386 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6387 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6388 - initialize RPC timeout timer earlier for 2.6 (3219)
6389 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6390 - print nal/nid for unknown nid (3258)
6391 - additional checks for oscc recovery before doing precreate (3284)
6392 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6393 - don't crash in mdc_close for bad permissions on open (3285)
6394 - zero i_rdev for non-device files (3147)
6395 - clear page->private before handing to FS, better assertion (3119)
6396 - tune the read pipeline (3236)
6397 - fix incorrect decref of invalidated dentry (2350)
6398 - provide read-ahead stats and refine rpc in flight stats (3328)
6399 - don't hold journal transaction open across create RPC (3313)
6400 - update atime on MDS at close time (3265)
6401 - close LDAP connection when recovering to avoid server load (3315)
6402 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6403 - don't leak open file on MDS after open resend (3325)
6404 - serialize filter_precreate and filter_destroy_precreated (3329)
6405 - loop device shouldn't call sync_dev() for nul device (3092)
6406 - clear page cache after eviction (2766)
6407 - resynchronize MDS->OST in background (2824)
6408 - refuse to mount the same filesystem twice on same mountpoint (3394)
6409 - allow llmount to create routes for mounting behind routers (3320)
6410 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6411 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6412 - fix rare problem with rename on htree directories (3417)
6414 - allow default OST striping configuration per directory (1414)
6415 - fix compilation for qswnal for 2.6 kernels (3125)
6416 - increase maximum number of MDS request buffers for large systems
6417 - change liblustreapi to be useful for external progs like lfsck (3098)
6418 - increase local configuration timeout for slow disks (3353)
6419 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6421 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6424 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6425 - don't overwrite extent policy data in reply if lock was blocked (2901)
6426 - drop filter export grants atomically with removal from device (2663)
6427 - del obd_self_export from work_list in class_disconnect_exports (2908)
6428 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6429 - swab reply message in mdc_close, other PPC fixes (2464)
6430 - fix destroying of named logs (2325)
6431 - overwrite old logs when running lconf --write_conf (2264)
6432 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6433 - fix race in target_handle_connect (2898)
6434 - mds_reint_create() should take same inode create lock (2926)
6435 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6436 - don't close files for self_export to avoid uninitialized obd (2936)
6437 - allow MDS with the same name as client node (2939)
6438 - hold dentry reference for closed log files for unlink (2325)
6439 - reserve space for all logs during transactions (2059)
6440 - don't evict page beyond end of stripe extent (2925)
6441 - don't oops on a deleted current working directory (2399)
6442 - handle hard links to targets without a parent properly (2517)
6443 - don't dereference NULL lock when racing during eviction (2867)
6444 - don't grow lock extents when lots of conflicting locks (2919)
6446 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6449 - account for cache space usage on clients to avoid data loss (974)
6450 - lfsck support in lustre kernel code (2349)
6451 - reduce journal credits needed for BRW writes (2370)
6452 - orphan handling to avoid losing space on client/server crashes
6453 - ptlrpcd can be blocked, stopping ALL progress (2477)
6454 - use lock value blocks to assist in proper KMS, faster stat (1021)
6455 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6456 - recovery for initial connections (2355)
6457 - fixes for mds_cleanup_orphans (1934)
6458 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6459 - block all file creations until orphan recovery completes (1901)
6460 - client remove rq_connection from request struct (2423)
6461 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6462 - recovery timer race (2670)
6463 - mdc_close recovey bug (2532)
6464 - ptlrpc cleanup bug (2710)
6465 - mds timeout on local locks (2588)
6466 - namespace lock held during RPCs (2431)
6467 - handle interrupted sync write properly (2503)
6468 - don't try to handle a message that hasn't been replied to (2699)
6469 - client assert failure during cleanup after abort recovery (2701)
6470 - leak mdc device after failed mount (2712)
6471 - ptlrpc_check_set allows timedout requests to complete (2714)
6472 - wait for inflight reqs when ptlrpcd finishes (2710)
6473 - make sure unregistered services are removed from the srv_list
6474 - reset bulk XID's when resending them (caught by 1138 test)
6475 - unregister_bulk after timeout
6476 - fix lconf error (2694)
6477 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6478 - readahead locks pages, leaves pending causing memory pressure (2673)
6479 - increase OST request buffers to 4096 on large machines (2729)
6480 - fix up permission of existing directories in simple_mkdir (2661)
6481 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6482 - don't assign transno to errored transactions (2742)
6483 - don't delete objects on OST if given a bogus objid from MDS (2751)
6484 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6485 - if rq_no_resend, then timeout request after recovery (2432)
6486 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6487 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6488 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6489 - MDS recovery shouldn't delete objects that it has given out (2730)
6490 - if enqueue arrives after completion, don't clobber LVB (2819)
6491 - don't unlock pages twice when trigger_group_io returns error (2814)
6492 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6493 - don't write pages to disk if there was an error (1450)
6494 - don't ping imports that have recovery disabled (2676)
6495 - take buffered bytes into account when balancing socknal conn (2817)
6496 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6497 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6498 - remove little-endian swabbing of llog records (1987)
6499 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6500 - retry reposting request buffers if they fail (1191)
6501 - grow extent at grant time to avoid granting a revoked lock (2809)
6502 - lock revoke doesn't evict page if covered by a second lock (2765)
6503 - disable VM readahead to avoid reading outside lock extents (2805)
6505 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6506 - updated kernel patches for hp-2.4.20 kernel (2681)
6508 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6511 - fix truncated write corruption (2366)
6512 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6514 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6515 - ensure len(uuid) < 37 in lmc (1171)
6516 - fix ia64 OOPS in llog_test (2255)
6517 - zero end of page at obdfilter for partial page writes (2648)
6518 - don't leave stale dentries around after renames (bug 2428)
6519 - fix timeouts when evicting a client with a single lock held (2642)
6520 - set deadline for the initial HELLO message to drain (2634)
6521 - print out dotted-quad IP addresses in the socknal (2302)
6523 - additional debugging for MDS client eviction problem (2443)
6524 - fix mkfsoptions support for osts (2603, 2604)
6526 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6529 - add series for the vanilla 2.6.0 kernel
6530 - add series for the vanilla 2.4.24 kernel
6531 - add series for a cray x86/64 UL kernel drop
6532 - fix xattr patches for the vanilla 2.4.19 series
6534 - generate true UUIDs in lmc (1171)
6535 - have portals stack dumping break in UML (2466)
6536 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6537 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6538 - wake callback waiting threads on client eviction (2460)
6539 - Add --ptldebug and --subsystem to lmc (1719)
6540 - update assertion to allow safe interrupt allocation
6541 - set rq_no_resend for cancel requests (2432)
6542 - recalculate ptlrpcd timeout after resend (2494)
6543 - call vfs_rmdir when removing pending directories (2368)
6544 - fix renaming a file to itself (2429)
6545 - lmc creates a default one-stripe lov (2454)
6546 - expand procfs space to handle large clusters (2326)
6547 - increase UML stack to avoid overflow
6548 - update lconf's list of debug and subsystem masks
6549 - fix lfs find --obd (2510)
6550 - /proc tunable for disabling filter read caching (2591)
6551 - stop rpm packages from altering slapd.conf (2301)
6552 - disable nagle in the socknal under 0conf (2578)
6553 - choose mds inode size based on stripe count (2572)
6554 - fix kernel-source rpm problems (2516)
6556 - add --disable-doc to avoid pdf generation (2421)
6557 - update documentation, tests, type-os, comments
6558 - avoid format warnings on ia64
6559 - remove the TOE NAL
6560 - tiny code cleanups by removing unused fields
6562 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6565 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6566 - remove the most blatant lies from BUILDING file (2371)
6567 - change default debug level to reasonable production setting
6568 - reduce client side cache size to reduce cache flush time
6569 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6570 - make TCP zerocopy and pinger support enabled by default (2476)
6571 - sync writes completed after process exits caused crashes (2319)
6572 - maintain correct mount count on the MDS (2356)
6573 - backout 1557, because 2316 wasn't really fixed
6574 - better file I/O statistics gathering in /proc
6575 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6576 - another kernel patch to fix zero-copy TCP function export
6577 - don't take duplicate lock when processing re-sent getattr (2420)
6578 - lctl uses obd_self_export instead of creating new conn (2353)
6579 - MDS/OST recovery case which requires object creation asserted (2425)
6580 - move lfs from /usr/sbin to /usr/bin in packages
6581 - fix race between mds_client_add and mds_client_free (2417)
6582 - use kmalloc instead of slabs in portals (2430)
6583 - don't create duplicate records when a failover MDS is present (2442)
6584 - remove unnecessary mount age check (2332)
6585 - don't remove directory inodes from locks prematurely (2451)
6586 - don't break if MDS service name is the same as hostname (2103)
6587 - fix races in client write RPC generation when cache full (2482)
6589 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6592 - remove now-unused request->rq_obd (278)
6593 - if an allocation fails, print out how much memory we've used (1933)
6594 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6595 - add the "configurable stack size" patch to most series files (1256)
6596 - ability to write large log records, for 100+ OST configs (2306)
6597 - fix NULL deref when filter_prep fails (2314)
6598 - fix operator precedence error in filter_sync
6599 - dynamic allocation of socknal TX descriptors (2315)
6600 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6601 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6602 - ensure that CWARN messages in Portals always get to the syslog
6603 - __init/__exit are not for prototype decls (ldlm_init/exit)
6604 - x86-64 compile warning fixes
6605 - fix gateway LMC keyword conflict (2318)
6606 - fix MDS lock inversions in getattr/reint paths (1844)
6607 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6608 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6609 - DLM race condition prevented some lock evictions (2328)
6610 - ENOMEM detection and retry on socknal sends (2230)
6611 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6612 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6613 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6614 - better error messages when a client is rejected during recovery (1505)
6615 - avoid cancelling locks which were never granted, after failure (2330)
6616 - fix i_sem/journal inversion in mds_client_add (2333)
6617 - fix truncate/getattr lock cycle deadlock (2334)
6618 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6619 - fix two rare exit paths which could leak an l_lock() ref (2321)
6620 - fencepost error in MDS/OST orphan recovery (2226)
6621 - make log record alignment 8 bytes (1988)
6622 - lstripe now fails when requested offset > ost_count (2237)
6623 - ensure that all kernel series have a complete list.h (1607)
6624 - fix crashes in special-file operations (2316)
6625 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6627 - allow configurable automake binary, for testing new versions
6628 - small update to the lfs documentation
6630 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6632 * fix negative export reference count in fsfilt_sync (2312)
6634 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6635 * release candidate 0.9.1
6637 - orphans are moved into the PENDING directory for possible recovery
6638 - replayed opens now open by fid for orphan/rename safety (1042)
6639 - last close of an orphan inode generates a transno (683)
6640 - chdir() and mount() now pin the directory entry (1020)
6641 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6642 - discard very old requests without processing them (1502)
6643 - remove some common, well-understood CERRORs (1505)
6644 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6645 - clear "grant" flags in OST replies until OST grant code lands (1644)
6646 - fix read performance by not clobbering i_blksize on client (1598)
6647 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6648 - make LRU size a /proc tunable, clears locks when reduced (707)
6649 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6650 - server locks take references on exports now (1558)
6651 - build fixes for 2.4.20-rh trees (1663)
6652 - return an error from lov_create if all OSCs are inactive (1751)
6653 - fix import levels when a reconnect happens without a timeout (1597)
6654 - exit early from mds_open if we get a lookup error (1749)
6655 - partial page read at EOF wouldn't wait for disk before sending (1642)
6656 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6657 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6658 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6659 - write/truncate lock inversion (1639)
6660 - Don't auto-load obdclass, portals modules during cleanup (1495)
6661 - fix timestamps from jumping to "now" (1763)
6662 - extra journal assertions (1648)
6663 - add an extra multiunlink test (1771)
6664 - fix read_record/write_record API (1776)
6665 - fix leak of offset_extent, possible incorrect i_size later (1772)
6666 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6667 - Debugging for the kqswnal_get_idle_tx problems (1820)
6668 - Allow recovery to be attempted multiple times (1536)
6669 - Write out MDS last_rcvd file after it is first created (1600)
6670 - Fix tx_descriptor leak in failed transmit situations (1827)
6671 - ext3 journaling fixes for assertion failure after IO error (1871)
6672 - class_export_put() on freed export after completion AST error (1896)
6673 - Fix revalidate looping in VFS (1322)
6674 - Don't access a freed export during MDS_REINT timeout (1521)
6675 - Add open-unlink recovery support on the MDS (1673,1764)
6676 - Return an error if no MDS data was read from last_rcvd (1946)
6677 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6678 - Don't setup a disk device that doesn't match exported UUID (317)
6679 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6680 - avoid committing NULL handle in force close
6681 - local.sh is now a one-stripe LOV configuration
6682 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6683 - fix ext3 htree duplicate directory entry corruption (1516)
6684 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6685 - update to new LOV EA format (2097)
6686 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6688 2003-06-15 Phil Schwan <phil@clusterfs.com>
6691 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6692 - per-import/export recovery handling (958, 931, 959)
6693 - multiple last-rcvd slots, for serving multiple FSes (949)
6694 - connections are again shared between multiple imp/exports (963, 964)
6695 - "umount -f" would hang if any requests needed to be sent (393, 978)
6696 - avoid pinning large req buffer by copying for queued messages (989)
6697 - add "uuid" to "lctl device" command to help upcalls (991)
6698 - "open" RPCs with transnos would confuse recovery counters (1037)
6699 - do proper endian conversion of all wire messages (288, 340, 891)
6700 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6701 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6702 - fix ptlrpc_connection leak in target_handle_connect (1174)
6703 - fix import refcounting bug in OST and MDS cleanup (1134)
6704 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6705 - fix very unlikely obd_types race condition (501)
6706 - remove osc_open hack for echo_client (1187)
6707 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6708 - a failure in read_inode2 leads to deadlock (1139)
6709 - cancel ack-locks as soon as transaction is committed (1072)
6710 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6711 - make sure to commitrw after any preprw to avoid deadlock (1162)
6712 - failing to execute a file in a lustre FS would lock inode (1203)
6713 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6714 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6715 - obd_uuid2tgt would walk off the end of the list (1255)
6716 - on IA64 the osc would give portals incorrect bulk size (1258)
6717 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6718 - fix lock inversion caused by new llite matching code (1282)
6719 - limit the number of dirty pages on a client to 10MB (1286)
6720 - timed out locks were not being corrected cancelled (1289)
6721 - fix O_DIRECT above 4GB on IA-32 (1292)
6722 * major user-visible changes
6723 - fail out/fail over policy now controlled by the upcall (993)
6725 - add OBD_PING to check server availability and failure (954)
6726 - lustre messages are now sent in sending host order (288, 340, 891)
6727 - add eadatalen to MDS getattr reply (340)
6728 - OST read replies may contain second buffer, with per-page status (593)
6730 2003-03-11 Phil Schwan <phil@clusterfs.com>
6733 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6734 - failed open()s could cause deadlock; fixed (867, 869)
6735 - stop cancelling OST locks when files are closed (481)
6736 - overlapping XID spaces caused network corruption (851, 853)
6737 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6738 - setattr_raw added, to do single-RPC, server-side setattrs
6739 - lmc/lconf syntax change for OST UUIDs
6740 - fix crashy race condition between ptlrpc_free_req and osc_close
6741 - don't use request in mdc_enqueue if we hit a timeout (889)
6742 - don't set the inode i_size for regular files from the MDS (896)
6743 - handle out of order completion AST (842)
6744 - don't LBUG if a lock request times out after receiving AST (913)
6745 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6746 - if a bad lock AST arrives, send an error instead of dropping entirely
6747 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6748 - fix leak in bulk IO when only partially completed (899, 900, 926)
6749 - fix O_DIRECT for ia64 (55)
6750 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6751 - C-z after timeout could hang a process forever; fixed (977)
6753 - client-side I/O cache (678, 924, 929, 941, 970)
6755 - READPAGE and SETATTRs which don't take server-side locks get
6758 2003-02-11 Phil Schwan <phil@clusterfs.com>
6761 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6762 - Fix fsfilt_extN_readpage() to read a full page of directory
6763 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6764 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6765 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6766 - Most intent operations were removed, in favour of a new RPC mode
6767 that does a single RPC to the server and bypasses most of the VFS
6768 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6769 - Aggressively cancel local locks on DLM servers
6770 - mds_reint_unlink sends EA to the client if it's the last nlink.
6771 client uses that EA to unlink OST objects.
6772 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6773 - recursive symlinks were fixed (439)
6774 - fixed NULL deref in DEBUG_REQ
6775 - filter_update_lastobjid no longer calls sync, which annoyed extN
6776 - fixed multi-client small-writes to a single file problem (445)
6777 - fixed mtime updates during file writes (607)
6778 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6779 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6780 - fixed Linux OST size reporting problem (444, 656)
6781 - OST now updates object mtime with writes or setattr (607, 619)
6782 - client verifies file size before zeroing page past EOF (445)
6783 - OST now writes last allocated objid to disk with allocation (108)
6784 - LOV on echo now works (409)
6786 - mds_reint_unlink sends a new buffer, with the EA included. this
6787 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6788 set if a regular file was being unlinked, and it was the last link
6789 - use PtlGet from the target for bulk writes (315)
6790 - OST now updates object mtime with writes or setattr (607, 619)
6791 - LDLM now has a grant-time callback to revalidate locked items, if
6793 - Many MDS operations were reorganized to combat race conditions
6795 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6796 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6798 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6801 - Fully reactivate OST imports after reconnection (512, others)
6802 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6803 - More graceful error handling for truncating on dead OST (515)
6804 - Don't error out unless we're actually accessing dead stripes (474)
6805 - Fix garbage sizes when stripes are missing (410)
6806 - LRU counters were broken, causing constant lock purge (433, 432)
6807 - garbage on read from stripes with failed OSTs (441)
6808 - mark OSCs as active before reconnecting during recovery (438)
6809 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6810 - lfind did not preserve OST order in output (443)
6811 - symlinks cause hung clients, incorrect data (439)
6812 - stop dereferencing request after dropping refcount (457)
6813 - don't LASSERT(spin_is_locked) on non-SMP (455)
6814 - fixes for many rename() bugs
6815 - fstat didn't correctly synchronize attributes (399)
6816 - server must handle lock cancellation during blocking AST prep (487)
6817 - bulk descriptors were free()d too soon (511)
6818 - fix paths in lconf, which would load incorrect modules (451, 507)
6819 - fix confusing lconf 'host not found' error message (386)
6820 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6821 - fix race condition in mdc_blocking_ast() for inode access (526)
6822 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6823 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6824 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6825 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6826 - fix lstripe in several ways (kernel side) (527)
6827 - fix request leak in ldlm_cli_enqueue (262)
6828 - incorrect OSC was marked inactive after OST failure
6829 - call mds_fs_cleanup before unmounting filesystem (524)
6830 - fix races between taking ns_lock and ldlm_lock_change_resource
6831 - fix races updating LOV export open file list
6832 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6833 - fix recovery NULL deref in ldlm_cli_cancel_unused
6834 - fix some DLM races by using new hash table for lock handles (419)
6835 - permit the client to specify desired inodes, at replay
6836 - duplicate requests when we queue them for replay reintegration
6837 - fix last_rcvd offset calculation
6838 - sync after each recovered transaction, so we always make progress
6839 - never, not always, ERESTART requests without transnos
6840 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6842 - skip replay if the MDS says that the client is already connected
6843 - don't check for a recovery-enabled export to match lctl's UUID
6844 - don't INC_USE_COUNT for phantom exports
6845 - don't crash when cleaning up phantom exports (567)
6846 - don't double-finish or set replay data for errored mdc_open requests
6847 - abort requests when they time out, so we don't get old replies
6848 - send/receive replies for AST messages again
6849 - if the client says that it doesn't have the lock, cancel it on the
6851 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6852 mark it as destroyed, it will all work out when decref is called
6853 - fix module use counts (22, 581)
6855 - ASTs now expect a reply (server cancels lock on error reply)
6857 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6860 - fix many simultaneous client startup (392)
6861 - fix dentry->d_it clobbering
6862 - credentials weren't being shipped for readdir/getattr operations
6863 - remove invalid assertions triggered during some concurrent MD
6865 - proper Lustre versions added (336, 389)
6866 - fix memory leak for create error case (398)
6867 - fix LOV locking bug that would get cli/srv out of sync
6868 - fix echo client over LOV (409)
6869 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6870 - fix double-O_EXCL intent crash (424)
6871 - avoid sending multiple lock CANCELs (352)
6873 - MDS can do multi-client recovery (modulo bugs in new code)
6875 - many updates, edits, cleanups
6877 2002-11-18 Phil Schwan <phil@clusterfs.com>
6880 - fix null d_it dereference (346)
6881 - fix full OST/dbench hang (333)
6882 - fix permission problem with file removal (286)
6883 - fix removal of OSCs from LOV when they fail
6884 - fix NULL deref during bulk timeout (214)
6885 - fix problems related to multiple filesystems on one MDS (241)
6886 - fixed serious subtle metadata locking bugs
6887 - free locks on clients when inodes are removed due to memory
6889 - fix inode pointer in lock data (285)
6890 - partial support for multiple MDS on a single host (241)
6891 - data locks weren't cancelled at clear_inode time (290, 311)
6892 - intent locks could lead to unbounded lock growth (205)
6893 - added a maximum lock count, an LRU list, and a flusher
6894 - fix multiple rename (365)
6895 - properly abstracted the echo client
6896 - OSC locked 1 byte too many; fixed
6897 - rewrote brw callback code:
6898 - fixed recovery bugs related to LOVs (306)
6899 - fixed too-many-pages-in-one-write crash (191)
6900 - fixed (again) crash in sync_io_timeout (214)
6901 - probably fixed callback-related race (385)
6903 - Add capability to MDS protocol
6904 - LDLM cancellations and callbacks on different portals
6906 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6909 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6910 - defer open object destruction to close time (601981, 138)
6911 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6912 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6913 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6914 - don't call obd_setattr in ll_file_release for destroyed objects
6915 * protocol change to lustre_msg: move |version| and add |flags|
6916 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6917 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6918 * added replay of create, unlink, link and rename operations during
6919 MDS failover; recovery should be much more robust now
6920 * remove failed OSCs from LOVs (only lov_create uses this so far)
6921 * the lustre-HOWTO was brought (more) up to date (582544)
6923 2002-10-23 Phil Schwan <phil@clusterfs.com>
6926 - in-use dentries weren't being reused properly (617851)
6927 - prevent multiple LDLM setup (599178)
6928 - fix LOV size calculations for truncate (617853)
6929 - fix client handling of MDS intent errors (POSIX)
6930 - fix permission bug in lovstripe.c test (624321)
6931 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6932 - truncate past end of file could corrupt data
6933 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6934 - a race in recovery could return ETIMEDOUT to apps (623947)
6935 - building outside the source directory was fixed
6936 * the lustre-HOWTO was brought (more) up to date (582544)
6937 * major progress was made on recovery functionality
6939 2002-10-10 Phil Schwan <phil@clusterfs.com>
6942 - recovery deadlock fix
6943 - rm -rf causes LBUG fix (617817)
6944 - file open by multiple tasks fix (618962)
6945 - directory permissions bugs (602707 and 620007)
6946 - journal_stop fixed with locking (611313)
6947 - O_APPEND failures resolved (618273, perhaps 614459)
6948 - lconf PATH fix (619770)
6949 - IA64 build fix (621450)
6950 - RPC buffer sizes scale with amount of memory
6952 2002-10-01 Phil Schwan <phil@clusterfs.com>
6955 - locks would be cancelled without throwing away data pages,
6956 resulting in inconsistent data (605627)
6957 - inode attributes were not always being refreshed (605627, 612449)
6958 - lconf now continues to cleanup after lctl reports an error
6959 - MDS now enforces user permissions (602707)
6960 - lprocfs cleanup fixed, but not yet enabled (614157)
6961 - fixed infinite server hang, should a client not respond to an AST
6962 - avoid going into recovery if user calls readlink() with a buffer
6963 that's too small (613941)
6964 - AST RPCs no longer require replies (614867) -- this may be changed
6965 - don't crash server if client sends an IOV that's too big (611336)
6966 - fixed lock conversion deadlock (611892)
6967 - fixed the following of symlinks (614622)
6968 * recovery: the server can remove locks from a client that dies, other
6969 clients can make progress
6970 * more extN patch fixes
6971 * compile-time configurable ptlrpc buffer allocations
6973 - collaborative read cache document
6974 - Lustre Lite Performance CDR document-in-progress
6976 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6979 - fix typo in patch-2.4.18
6981 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6984 - clear ptlrpc request each time in handle_incoming_request()
6985 - unlink of files now destroys the object on the OST
6987 2002-09-19 Peter Braam <braam@clusterfs.com>
6989 * add hard link support
6990 * change obdfile creation method
6991 * kernel patch changed
6993 2002-09-19 Peter Braam <braam@clusterfs.com>
6996 - stack overflow bug in extN fixed
6998 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
7000 * documentation updates
7001 - add man pages for config tools
7002 - update tests/README to describe testing with new config tools
7003 - finish metadata API descriptions
7004 * bug fixes and cleanups
7005 - statfs workaround for 16TB limit
7006 - LOV stripe allocation improved, can stripe on subset of OSTs
7007 - LOV file size/IO offset was wrong for files > 4GB in size
7008 - object EA data was being dropped, caused files to be unreadable
7009 - memory overflow with non-LOV OST caused memory corruption
7010 - fixed regression tests to work with new config tools, obdfilter
7011 - fixed bug when directory size became larger than 1 block
7012 - fixed bug (for single client case) when PWD was deleted
7013 - invalidate local directory pages when doing intent-based ops
7014 - avoid LDLM oops when lock callback contained bad data
7016 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
7018 * documentation updates
7019 * bug fixes and cleanups
7020 - configuration tools
7023 - 64-bit compile warnings
7024 - 64-bit internal statfs data
7026 * test_brw on persistent OST devices
7028 * lprocfs (disabled)
7030 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
7032 * documentation updates
7033 * bug fixes and cleanups
7034 * configuration tools
7036 2002-08-30 Peter J. Braam <braam@clusterfs.com>
7039 * many small fixes to 0_5_4
7040 * io/network handling
7041 * thinkos in MDS operations
7043 2002-08-24 Peter J. Braam <braam@clusterfs.com>
7046 * crucial basic fixes to 0.5.3
7047 * IOR, Iozone work over Elan
7050 2002-08-07 Phil Schwan <phil@clusterfs.com>
7051 * version 0_5_3, our first alpha
7052 * we use the new Portals iovs
7053 * documentation updates
7054 * bug fixes and cleanups
7055 * small changes in the DLM wire protocol
7057 2002-07-25 Peter J. Braam <braam@clusterfs.com>
7058 * version 0_5_1 with some initial stability,
7059 * locking on MD and file I/O.
7060 * documentation updates
7061 * several bug fixes since 0.5.0
7062 * small changes in wire protocol
7064 2002-07-18 Phil Schwan <phil@clusterfs.com>
7066 * delivered as Lustre Light Alpha
7067 * fixed a crash after handling invalid MDS requests
7068 * fixed directory pages for architectures with non-4k pages sizes
7070 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
7071 * release version v0_4_4
7072 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
7074 * Fixes a number of interruption problems with OST operations.
7075 * Update documentation for portals header changes
7076 * Move all wire protocol structs/defines to lustre_idl.h
7077 * Fixes symlink length bug.
7078 * Add tcpdump to repository.
7080 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
7081 * release version v0_4_3
7082 * Fixes statfs for inodes on extN.
7083 * Fixes bug in runtests which would delete /etc/hosts.
7084 * Use 64-bit object IDs wherever possible (not into VFS though)
7085 Remove ost_get_info, which is unused by lustre, and out of date.
7087 2002-07-03 Peter Braam <braam@clusterfs.com>
7088 * release version v0_4_2 Fixes a lookup error (type not passed)
7089 * move forward to head of Portals
7090 * move forward to latest Lustre kernel
7092 2002-06-25 Peter Braam <braam@clusterfs.com>
7093 * release version v0_4_1. Hopefully stable on single node use.