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 3.10.0-1160.45.1.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 3.10.0-1160.45.1.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 4.4.120-92.70 (SLES12 SP2)
41 4.4.0-142 (Ubuntu 16.04)
42 vanilla linux 4.15.0 (ZFS only)
43 * Recommended e2fsprogs version: 1.45.6.wc1 or newer
44 * Recommended ZFS / SPL version: 0.7.13
45 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
46 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
47 stack overflow. For more information, please refer to bugzilla 17630.
48 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
49 the re-exporting nodes to be mounted with "32bitapi" mount option
50 * Note for Ubuntu 14.04.5:
51 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
52 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
56 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
57 for currently supported client and server kernel versions.
58 * Server primary kernels built and tested during release cycle:
59 3.10.0-957.27.2.el7 (RHEL7.6)
60 3.10.0-1062.18.1.el7 (RHEL7.7)
61 3.10.0-1127.19.1.el7 (RHEL7.8)
62 3.10.0-1160.45.1.el7 (RHEL7.9)
63 * Server known to build on the following kernels (others may also work):
64 2.6.32-696.18.7.el6 (RHEL6.9)
65 3.10.0-862.14.4.el7 (RHEL7.5)
66 4.4.120-92.70 (SLES12 SP2)
67 4.4.180-94.97 (SLES12 SP3)
68 4.4.0-142 (Ubuntu 16.04)
69 4.15.0-45 (Ubuntu 18.04)
70 vanilla linux 4.6.7 (ZFS only)
71 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
72 * Client primary unpatched kernels built and tested during release cycle:
73 3.10.0-957.27.2.el7 (RHEL7.6)
74 3.10.0-1062.18.1.el7 (RHEL7.7)
75 3.10.0-1127.19.1.el7 (RHEL7.8)
76 3.10.0-1160.45.1.el7 (RHEL7.9)
77 4.18.0-80.11.2.el8 (RHEL8.0)
78 4.4.180-94.97 (SLES12 SP3)
79 4.12.14-95.54 (SLES12 SP4)
80 4.12.14-122.91 (SLES12 SP5)
81 4.15.0-45 (Ubuntu 18.04)
82 * Client known to build on unpatched kernels (others may also work):
83 2.6.32-696.18.7.el6 (RHEL6.9)
84 2.6.32-754.6.3.el6 (RHEL6.10)
85 3.10.0-862.14.4.el7 (RHEL7.5)
86 4.14.0-49.13.1.el7a (RHEL7.5)
87 4.14.0-115.2.2.el7a (RHEL7.6)
88 4.18.0-147.3.1.el8 (RHEL8.1)
89 4.18.0-193.28.1.el8 (RHEL8.2)
90 4.18.0-240.22.1.el8 (RHEL8.3)
91 4.18.0-305.19.1.el8 (RHEL8.4)
92 4.18.0-348.2.1.el8 (RHEL8.5)
93 4.4.120-92.70 (SLES12 SP2)
94 4.4.0-142 (Ubuntu 16.04)
95 vanilla linux 4.15.0 (ZFS only)
96 * Recommended e2fsprogs version: 1.45.6.wc1 or newer
97 * Recommended ZFS / SPL version: 0.7.13
98 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
99 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
100 stack overflow. For more information, please refer to bugzilla 17630.
101 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
102 the re-exporting nodes to be mounted with "32bitapi" mount option
103 * Note for Ubuntu 14.04.5:
104 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
105 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
109 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
110 for currently supported client and server kernel versions.
111 * Server primary kernels built and tested during release cycle:
112 3.10.0-957.27.2.el7 (RHEL7.6)
113 3.10.0-1062.18.1.el7 (RHEL7.7)
114 3.10.0-1127.19.1.el7 (RHEL7.8)
115 3.10.0-1160.15.2.el7 (RHEL7.9)
116 * Server known to build on the following kernels (others may also work):
117 2.6.32-696.18.7.el6 (RHEL6.9)
118 3.10.0-862.14.4.el7 (RHEL7.5)
119 4.4.120-92.70 (SLES12 SP2)
120 4.4.180-94.97 (SLES12 SP3)
121 4.4.0-142 (Ubuntu 16.04)
122 4.15.0-45 (Ubuntu 18.04)
123 vanilla linux 4.6.7 (ZFS only)
124 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
125 * Client primary unpatched kernels built and tested during release cycle:
126 3.10.0-957.27.2.el7 (RHEL7.6)
127 3.10.0-1062.18.1.el7 (RHEL7.7)
128 3.10.0-1127.19.1.el7 (RHEL7.8)
129 3.10.0-1160.15.2.el7 (RHEL7.9)
130 4.18.0-80.11.2.el8 (RHEL8.0)
131 4.4.180-94.97 (SLES12 SP3)
132 4.12.14-95.54 (SLES12 SP4)
133 4.12.14-122.57 (SLES12 SP5)
134 4.15.0-45 (Ubuntu 18.04)
135 * Client known to build on unpatched kernels (others may also work):
136 2.6.32-696.18.7.el6 (RHEL6.9)
137 2.6.32-754.6.3.el6 (RHEL6.10)
138 3.10.0-862.14.4.el7 (RHEL7.5)
139 4.14.0-49.13.1.el7a (RHEL7.5)
140 4.14.0-115.2.2.el7a (RHEL7.6)
141 4.18.0-147.3.1.el8 (RHEL8.1)
142 4.18.0-193.28.1.el8 (RHEL8.2)
143 4.18.0-240.1.1.el8 (RHEL8.3)
144 4.4.120-92.70 (SLES12 SP2)
145 4.4.0-142 (Ubuntu 16.04)
146 vanilla linux 4.15.0 (ZFS only)
147 * Recommended e2fsprogs version: 1.45.6.wc1 or newer
148 * Recommended ZFS / SPL version: 0.7.13
149 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
150 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
151 stack overflow. For more information, please refer to bugzilla 17630.
152 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
153 the re-exporting nodes to be mounted with "32bitapi" mount option
154 * Note for Ubuntu 14.04.5:
155 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
156 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
160 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
161 for currently supported client and server kernel versions.
162 * Server primary kernels built and tested during release cycle:
163 3.10.0-957.27.2.el7 (RHEL7.6)
164 3.10.0-1062.18.1.el7 (RHEL7.7)
165 3.10.0-1127.8.2.el7 (RHEL7.8)
166 * Server known to build on the following kernels (others may also work):
167 2.6.32-696.18.7.el6 (RHEL6.9)
168 3.10.0-862.14.4.el7 (RHEL7.5)
169 4.4.120-92.70 (SLES12 SP2)
170 4.4.180-94.97 (SLES12 SP3)
171 4.4.0-142 (Ubuntu 16.04)
172 4.15.0-45 (Ubuntu 18.04)
173 vanilla linux 4.6.7 (ZFS only)
174 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
175 * Client primary unpatched kernels built and tested during release cycle:
176 3.10.0-957.27.2.el7 (RHEL7.6)
177 3.10.0-1062.18.1.el7 (RHEL7.7)
178 3.10.0-1127.8.2.el7 (RHEL7.8)
179 4.18.0-80.11.2.el8 (RHEL8.0)
180 4.4.180-94.97 (SLES12 SP3)
181 4.12.14-95.45 (SLES12 SP4)
182 4.12.14-122.20 (SLES12 SP5)
183 4.15.0-45 (Ubuntu 18.04)
184 * Client known to build on unpatched kernels (others may also work):
185 2.6.32-696.18.7.el6 (RHEL6.9)
186 2.6.32-754.6.3.el6 (RHEL6.10)
187 3.10.0-862.14.4.el7 (RHEL7.5)
188 4.14.0-49.13.1.el7a (RHEL7.5)
189 4.14.0-115.2.2.el7a (RHEL7.6)
190 4.18.0-147.8.1.el8 (RHEL8.1)
191 4.18.0-193.1.2.el8 (RHEL8.2)
192 4.4.120-92.70 (SLES12 SP2)
193 4.4.0-142 (Ubuntu 16.04)
194 vanilla linux 4.15.0 (ZFS only)
195 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
196 * Recommended ZFS / SPL version: 0.7.13
197 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
198 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
199 stack overflow. For more information, please refer to bugzilla 17630.
200 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
201 the re-exporting nodes to be mounted with "32bitapi" mount option
202 * Note for Ubuntu 14.04.5:
203 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
204 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
208 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
209 for currently supported client and server kernel versions.
210 * Server primary kernels built and tested during release cycle:
211 3.10.0-957.27.2.el7 (RHEL7.6)
212 3.10.0-1062.9.1.el7 (RHEL7.7)
213 * Server known to build on the following kernels (others may also work):
214 2.6.32-696.18.7.el6 (RHEL6.9)
215 3.10.0-862.14.4.el7 (RHEL7.5)
216 4.4.120-92.70 (SLES12 SP2)
217 4.4.180-94.97 (SLES12 SP3)
218 4.4.0-142 (Ubuntu 16.04)
219 4.15.0-45 (Ubuntu 18.04)
220 vanilla linux 4.6.7 (ZFS only)
221 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
222 * Client primary unpatched kernels built and tested during release cycle:
223 3.10.0-957.27.2.el7 (RHEL7.6)
224 3.10.0-1062.9.1.el7 (RHEL7.7)
225 4.18.0-80.11.2.el8 (RHEL8.0)
226 4.4.180-94.97 (SLES12 SP3)
227 4.12.14-95.29 (SLES12 SP4)
228 4.15.0-45 (Ubuntu 18.04)
229 * Client known to build on unpatched kernels (others may also work):
230 2.6.32-696.18.7.el6 (RHEL6.9)
231 2.6.32-754.6.3.el6 (RHEL6.10)
232 3.10.0-862.14.4.el7 (RHEL7.5)
233 4.14.0-49.13.1.el7a (RHEL7.5)
234 4.14.0-115.2.2.el7a (RHEL7.6)
235 4.18.0-147.3.1.el8 (RHEL8.1)
236 4.4.120-92.70 (SLES12 SP2)
237 4.4.0-142 (Ubuntu 16.04)
238 vanilla linux 4.15.0 (ZFS only)
239 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
240 * Recommended ZFS / SPL version: 0.7.13
241 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
242 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
243 stack overflow. For more information, please refer to bugzilla 17630.
244 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
245 the re-exporting nodes to be mounted with "32bitapi" mount option
246 * Note for Ubuntu 14.04.5:
247 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
248 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
252 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
253 for currently supported client and server kernel versions.
254 * Server primary kernels built and tested during release cycle:
255 3.10.0-957.27.2.el7 (RHEL7.6)
256 3.10.0-1062.1.1.el7 (RHEL7.7)
257 * Server known to build on the following kernels (others may also work):
258 2.6.32-696.18.7.el6 (RHEL6.9)
259 3.10.0-862.14.4.el7 (RHEL7.5)
260 4.4.120-92.70 (SLES12 SP2)
261 4.4.180-94.97 (SLES12 SP3)
262 4.4.0-142 (Ubuntu 16.04)
263 4.15.0-45 (Ubuntu 18.04)
264 vanilla linux 4.6.7 (ZFS only)
265 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
266 * Client primary unpatched kernels built and tested during release cycle:
267 3.10.0-957.27.2.el7 (RHEL7.6)
268 3.10.0-1062.1.1.el7 (RHEL7.7)
269 4.18.0-80.7.1.el8 (RHEL8.0)
270 4.4.180-94.97 (SLES12 SP3)
271 4.12.14-95.29 (SLES12 SP4)
272 4.15.0-45 (Ubuntu 18.04)
273 * Client known to build on unpatched kernels (others may also work):
274 2.6.32-696.18.7.el6 (RHEL6.9)
275 2.6.32-754.6.3.el6 (RHEL6.10)
276 3.10.0-862.14.4.el7 (RHEL7.5)
277 4.14.0-49.13.1.el7a (RHEL7.5)
278 4.14.0-115.2.2.el7a (RHEL7.6)
279 4.4.120-92.70 (SLES12 SP2)
280 4.4.0-142 (Ubuntu 16.04)
281 vanilla linux 4.15.0 (ZFS only)
282 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
283 * Recommended ZFS / SPL version: 0.7.13
284 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
285 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
286 stack overflow. For more information, please refer to bugzilla 17630.
287 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
288 the re-exporting nodes to be mounted with "32bitapi" mount option
289 * Note for Ubuntu 14.04.5:
290 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
291 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
295 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
296 for currently supported client and server kernel versions.
297 * Server known to build on the following kernels (others may also work):
298 2.6.32-696.18.7.el6 (RHEL6.9)
299 3.10.0-862.14.4.el7 (RHEL7.5)
300 3.10.0-957.10.1.el7 (RHEL7.6)
301 4.4.120-92.70 (SLES12 SP2)
302 4.4.162-94.69 (SLES12 SP3)
303 4.4.0-142 (Ubuntu 16.04)
304 4.15.0-45 (Ubuntu 18.04)
305 vanilla linux 4.6.7 (ZFS only)
306 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
307 * Client known to build on unpatched kernels (others may also work):
308 2.6.32-696.18.7.el6 (RHEL6.9)
309 2.6.32-754.6.3.el6 (RHEL6.10)
310 3.10.0-862.14.4.el7 (RHEL7.5)
311 3.10.0-957.10.1.el7 (RHEL7.6)
312 4.14.0-49.13.1.el7a (RHEL7.5)
313 4.14.0-115.2.2.el7a (RHEL7.6)
314 4.4.120-92.70 (SLES12 SP2)
315 4.4.162-94.69 (SLES12 SP3)
316 4.12.14-95.13 (SLES12 SP4)
317 4.4.0-142 (Ubuntu 16.04)
318 4.15.0-45 (Ubuntu 18.04)
319 vanilla linux 4.15.0 (ZFS only)
320 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
321 * Recommended ZFS / SPL version: 0.7.13
322 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
323 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
324 stack overflow. For more information, please refer to bugzilla 17630.
325 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
326 the re-exporting nodes to be mounted with "32bitapi" mount option
327 * Note for Ubuntu 14.04.5:
328 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
329 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
334 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
335 for currently supported client and server kernel versions.
336 * Server known to build on the following kernels (others may also work):
337 2.6.32-696.18.7.el6 (RHEL6.9)
338 3.10.0-862.14.4.el7 (RHEL7.5)
339 3.10.0-957.10.1.el7 (RHEL7.6)
340 4.4.120-92.70 (SLES12 SP2)
341 4.4.162-94.69 (SLES12 SP3)
342 4.4.0-142 (Ubuntu 16.04)
343 4.15.0-45 (Ubuntu 18.04)
344 vanilla linux 4.6.7 (ZFS only)
345 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
346 * Client known to build on unpatched kernels (others may also work):
347 2.6.32-696.18.7.el6 (RHEL6.9)
348 2.6.32-754.6.3.el6 (RHEL6.10)
349 3.10.0-862.14.4.el7 (RHEL7.5)
350 3.10.0-957.10.1.el7 (RHEL7.6)
351 4.14.0-49.13.1.el7a (RHEL7.5)
352 4.14.0-115.2.2.el7a (RHEL7.6)
353 4.4.120-92.70 (SLES12 SP2)
354 4.4.162-94.69 (SLES12 SP3)
355 4.12.14-95.13 (SLES12 SP4)
356 4.4.0-142 (Ubuntu 16.04)
357 4.15.0-45 (Ubuntu 18.04)
358 vanilla linux 4.15.0 (ZFS only)
359 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
360 * Recommended ZFS / SPL version: 0.7.13
361 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
362 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
363 stack overflow. For more information, please refer to bugzilla 17630.
364 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
365 the re-exporting nodes to be mounted with "32bitapi" mount option
366 * Note for Ubuntu 14.04.5:
367 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
368 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
372 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
373 for currently supported client and server kernel versions.
374 * Server known to build on the following kernels (others may also work):
375 2.6.32-696.18.7.el6 (RHEL6.9)
376 3.10.0-862.14.4.el7 (RHEL7.5)
377 3.10.0-957.el7 (RHEL7.6)
378 4.4.120-92.70 (SLES12 SP2)
379 4.4.162-94.69 (SLES12 SP3)
380 4.4.0-131 (Ubuntu 16.04)
381 4.15.0-32 (Ubuntu 18.04)
382 vanilla linux 4.6.7 (ZFS only)
383 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
384 * Client known to build on unpatched kernels (others may also work):
385 2.6.32-696.18.7.el6 (RHEL6.9)
386 2.6.32-754.6.3.el6 (RHEL6.10)
387 3.10.0-862.14.4.el7 (RHEL7.5)
388 3.10.0-957.el7 (RHEL7.6)
389 4.14.0-49.13.1.el7a (RHEL7.5)
390 4.14.0-115.2.2.el7a (RHEL7.6)
391 4.4.120-92.70 (SLES12 SP2)
392 4.4.162-94.69 (SLES12 SP3)
393 4.4.0-131 (Ubuntu 16.04)
394 4.15.0-32 (Ubuntu 18.04)
395 vanilla linux 4.15.0 (ZFS only)
396 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
397 * Recommended ZFS / SPL version: 0.7.9
398 * Tested with ZFS / SPL version: 0.7.9
399 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
400 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
401 stack overflow. For more information, please refer to bugzilla 17630.
402 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
403 the re-exporting nodes to be mounted with "32bitapi" mount option
404 * Note for Ubuntu 14.04.5:
405 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
406 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
408 --------------------------------------------------------------------------------
411 2018-04-03 Intel Corporation
413 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
414 for currently supported client and server kernel versions.
415 * Server known to build on patched kernels:
416 2.6.32-431.29.2.el6 (RHEL6.5)
417 2.6.32-504.30.3.el6 (RHEL6.6)
418 2.6.32-573.26.1.el6 (RHEL6.7)
419 2.6.32-642.15.1.el6 (RHEL6.8)
420 2.6.32-696.18.7.el6 (RHEL6.9)
421 3.10.0-693.21.1.el7 (RHEL7.4)
422 3.0.101-0.47.71 (SLES11 SP3)
423 3.0.101-107 (SLES11 SP4)
424 3.12.74-60.64.40 (SLES12 SP1)
425 4.4.114-92.64 (SLES12 SP2)
426 4.4.114-94.11 (SLES12 SP3)
427 3.13.0-101 (Ubuntu 14.04, ZFS only)
428 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
429 4.4.0-116 (Ubuntu 16.04)
430 vanilla linux 4.6.7 (ZFS only)
431 * Client known to build on unpatched kernels:
432 2.6.32-431.29.2.el6 (RHEL6.5)
433 2.6.32-504.30.3.el6 (RHEL6.6)
434 2.6.32-573.26.1.el6 (RHEL6.7)
435 2.6.32-642.15.1.el6 (RHEL6.8)
436 2.6.32-696.18.7.el6 (RHEL6.9)
437 3.10.0-693.21.1.el7 (RHEL7.4)
438 3.0.101-0.47.71 (SLES11 SP3)
439 3.0.101-107 (SLES11 SP4)
440 3.12.74-60.64.40 (SLES12 SP1)
441 4.4.114-92.64 (SLES12 SP2)
442 4.4.114-94.11 (SLES12 SP3)
443 3.13.0-101 (Ubuntu 14.04)
444 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
445 4.4.0-116 (Ubuntu 16.04)
447 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
448 * Recommended ZFS / SPL version: 0.7.6
449 * Tested with ZFS / SPL version: 0.7.6
450 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
451 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
452 stack overflow. For more information, please refer to bugzilla 17630.
453 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
454 the re-exporting nodes to be mounted with "32bitapi" mount option
455 * Note for Ubuntu 14.04.5:
456 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
457 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
459 --------------------------------------------------------------------------------
461 2017-07-05 Intel Corporation
463 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
464 for currently supported client and server kernel versions.
465 * Server known to build on patched kernels:
466 2.6.32-431.29.2.el6 (RHEL6.5)
467 2.6.32-504.30.3.el6 (RHEL6.6)
468 2.6.32-573.26.1.el6 (RHEL6.7)
469 2.6.32-642.15.1.el6 (RHEL6.8)
470 2.6.32-696.3.1.el6 (RHEL6.9)
471 3.10.0-514.21.1.el7 (RHEL7.3)
472 3.0.101-0.47.71 (SLES11 SP3)
473 3.0.101-100 (SLES11 SP4)
474 3.12.74-60.64.40 (SLES12 SP1)
475 4.4.59-92.17 (SLES12 SP2)
476 vanilla linux 4.6.7 (ZFS only)
477 * Client known to build on unpatched kernels:
478 2.6.32-431.29.2.el6 (RHEL6.5)
479 2.6.32-504.30.3.el6 (RHEL6.6)
480 2.6.32-573.26.1.el6 (RHEL6.7)
481 2.6.32-642.15.1.el6 (RHEL6.8)
482 2.6.32-696.3.1.el6 (RHEL6.9)
483 3.10.0-514.21.1.el7 (RHEL7.3)
484 3.0.101-0.47.71 (SLES11 SP3)
485 3.0.101-100 (SLES11 SP4)
486 3.12.74-60.64.40 (SLES12 SP1)
487 4.4.59-92.17 (SLES12 SP2)
489 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
490 * Recommended ZFS / SPL version: 0.6.5.9
491 * Tested with ZFS / SPL version: 0.6.5.9
492 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
493 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
494 stack overflow. For more information, please refer to bugzilla 17630.
495 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
496 the re-exporting nodes to be mounted with "32bitapi" mount option
498 --------------------------------------------------------------------------------
500 12-07-2016 Intel Corporation
502 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
503 for currently supported client and server kernel versions.
504 * Server known to build on patched kernels:
505 2.6.32-431.29.2.el6 (RHEL6.5)
506 2.6.32-504.30.3.el6 (RHEL6.6)
507 2.6.32-573.26.1.el6 (RHEL6.7)
508 2.6.32-642.11.1.el6 (RHEL6.8)
509 3.10.0-514.el7 (RHEL7.3)
510 3.0.101-0.47.71 (SLES11 SP3)
511 3.0.101-80 (SLES11 SP4)
512 3.12.62-60.64.8 (SLES12 SP1)
513 vanilla linux 4.5.7 (ZFS only)
514 * Client known to build on unpatched kernels:
515 2.6.32-431.29.2.el6 (RHEL6.5)
516 2.6.32-504.30.3.el6 (RHEL6.6)
517 2.6.32-573.26.1.el6 (RHEL6.7)
518 2.6.32-642.11.1.el6 (RHEL6.8)
519 3.10.0-514.el7 (RHEL7.3)
520 3.0.101-0.47.71 (SLES11 SP3)
521 3.0.101-80 (SLES11 SP4)
522 3.12.62-60.64.8 (SLES12 SP1)
524 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
525 * Recommended ZFS / SPL version: 0.6.5.8
526 * Tested with ZFS / SPL version: 0.6.5.8
527 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
528 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
529 stack overflow. For more information, please refer to bugzilla 17630.
530 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
531 the re-exporting nodes to be mounted with "32bitapi" mount option
533 --------------------------------------------------------------------------------
535 02-29-2016 Intel Corporation
537 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
538 for currently supported client and server kernel versions.
539 * Server known to build on patched kernels:
540 2.6.32-431.29.2.el6 (RHEL6.5)
541 2.6.32-504.30.3.el6 (RHEL6.6)
542 2.6.32-573.12.1.el6 (RHEL6.7)
543 3.10.0-327.3.1.el7 (RHEL7.2)
544 3.0.101-0.47.71 (SLES11 SP3)
545 3.0.101-68 (SLES11 SP4)
546 vanilla linux 4.2.1 (ZFS only)
547 * Client known to build on unpatched kernels:
548 2.6.32-431.29.2.el6 (RHEL6.5)
549 2.6.32-504.30.3.el6 (RHEL6.6)
550 2.6.32-573.12.1.el6 (RHEL6.7)
551 3.10.0-327.3.1.el7 (RHEL7.2)
552 3.0.101-0.47.71 (SLES11 SP3)
553 3.0.101-68 (SLES11 SP4)
556 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
557 * Recommended ZFS / SPL version: 0.6.4.2
558 * Tested with ZFS / SPL version: 0.6.4.2
559 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
560 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
561 stack overflow. For more information, please refer to bugzilla 17630.
562 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
563 the re-exporting nodes to be mounted with "32bitapi" mount option
565 --------------------------------------------------------------------------------
567 03-10-2015 Intel Corporation
569 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
570 for currently supported client and server kernel versions.
571 * Server known to build on patched kernels:
572 2.6.32-431.29.2.el6 (RHEL6.5)
573 2.6.32-504.8.1.el6 (RHEL6.6)
574 3.0.101-0.46 (SLES11 SP3)
575 * Client known to build on unpatched kernels:
576 2.6.32-431.29.2.el6 (RHEL6.5)
577 2.6.32-504.8.1.el6 (RHEL6.6)
578 3.10.0-123.20.1.el7 (RHEL7)
579 3.0.101-0.46 (SLES11 SP3)
580 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
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 Severity : enhancement
589 Description: control OST-index in IDIF via ROCOMPAT flag.
590 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
591 last_rcvd file. For new formatted OST device, it will be auto set;
592 for the case of upgrading from old OST device, you can enable it
593 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
594 enabled, for new created OST-object, its IDIF-in-LMA will contain
595 the OST-index; for the existing OST-object, the OSD will convert
596 old format IDIF as new format IDIF with OST-index stored in the
597 LMA EA when accessing such OST-object or via OI scrub. Once such
598 flag is enabled, it cannot be reverted back, so the system cannot
599 be downgraded to the orignal incompatible version.
601 --------------------------------------------------------------------------------
603 07-30-2014 Intel Corporation
605 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
606 for currently supported client and server kernel versions.
607 * Server known to build on patched kernels:
608 2.6.32-431.20.3.el6 (RHEL6.5)
609 3.0.101-0.31 (SLES11 SP3)
610 * Client known to build on unpatched kernels:
611 2.6.32-431.20.3.el6 (RHEL6.5)
612 3.0.101-0.31 (SLES11 SP3)
615 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
616 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
617 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
618 stack overflow. For more information, please refer to bugzilla 17630.
619 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
620 the re-exporting nodes to be mounted with "32bitapi" mount option
622 --------------------------------------------------------------------------------
624 10-31-2013 Intel Corporation
626 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
627 * Server builds for kernels:
628 2.6.32-358.18.1.el6 (RHEL6.4)
629 * Client builds for unpatched kernels:
630 2.6.32-358.18.1.el6 (RHEL6.4)
631 3.0.92-0.8 (SLES11 SP3)
633 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
634 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
635 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
636 stack overflow. For more information, please refer to bugzilla 17630.
637 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
638 the re-exporting nodes to be mounted with "32bitapi" mount option
640 --------------------------------------------------------------------------------
642 05-30-2013 Intel Corporation
644 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
645 * Server builds for kernels:
646 2.6.32-358.6.2.el6 (RHEL6.4)
647 * Client builds for unpatched kernels:
648 2.6.32-358.6.2.el6 (RHEL6.4)
649 3.0.74-0.6.6 (SLES11 SP2)
651 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
652 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
653 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
654 stack overflow. For more information, please refer to bugzilla 17630.
655 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
656 the re-exporting nodes to be mounted with "32bitapi" mount option
658 --------------------------------------------------------------------------------
660 9-30-2012 Whamcloud, Inc.
662 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
663 * Server support for kernels:
664 2.6.32-279.5.1.el6 (RHEL6)
665 * Client support for unpatched kernels:
666 2.6.32-279.5.1.el6 (RHEL6)
668 2.6.18-238.19.1.el5 (RHEL5)
669 2.6.16.60-0.69.1 (SLES10)
670 * Recommended e2fsprogs version: 1.41.90.wc4
671 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
672 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
673 stack overflow. For more information, please refer to bugzilla 17630.
674 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
675 the re-exporting nodes to be mounted with "32bitapi" mount option
677 Severity : enhancement
679 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
681 Severity : enhancement
683 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
685 Severity : enhancement
687 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
689 Severity : enhancement
691 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
693 Severity : enhancement
695 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
697 --------------------------------------------------------------------------------
699 04-30-2012 Whamcloud, Inc.
701 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
702 * Server support for kernels:
703 2.6.32-220.4.2.el6 (RHEL6)
704 * Client support for unpatched kernels:
705 2.6.32-220.4.2.el6 (RHEL6)
707 2.6.18-238.19.1.el5 (RHEL5)
708 2.6.16.60-0.69.1 (SLES10)
709 * Recommended e2fsprogs version: 1.41.90.wc4
710 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
711 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
712 stack overflow. For more information, please refer to bugzilla 17630.
713 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
714 the re-exporting nodes to be mounted with "32bitapi" mount option
716 Severity : enhancement
718 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
720 Severity : enhancement
722 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
724 Severity : enhancement
726 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
728 Severity : enhancement
730 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
732 --------------------------------------------------------------------------------
734 09-30-2011 Whamcloud, Inc.
736 * Server support for kernels:
737 2.6.18-238.19.1.el5 (RHEL5)
738 2.6.32-131.6.1.el6 (RHEL6)
739 * Client support for unpatched kernels:
740 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
741 2.6.18-238.19.1.el5 (RHEL5)
742 2.6.32-131.6.1.el6 (RHEL6)
743 2.6.29.4-167.fc11 (FC11)
744 2.6.16.60-0.69.1 (SLES10)
746 * Recommended e2fsprogs version: 1.41.90.wc3
747 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
748 Lustre filesystem with 4K stack may cause a stack overflow. For more
749 information, please refer to bugzilla 17630.
750 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
751 mounted with "32bitapi" mount option
753 Severity : enhancement
754 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
756 Severity : enhancement
757 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
759 Severity : enhancement
760 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
763 Frequency : only with SLES10
765 Description: Use OFED "KMP" provided by Novell
766 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
767 Lustre is now built against this package. That means you need to
768 install the ofed-kmp package from Novell for the patchless client.
769 Note that the ofed-kmp that Novell ships may not exactly match the
770 kernel version but should still be compatible.
772 Severity : enhancement
774 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
775 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
777 --------------------------------------------------------------------------------
779 2010-07-15 Oracle, Inc.
781 * Server support for kernels:
782 2.6.18-164.11.1.el5 (RHEL 5)
783 2.6.18-164.11.1.0.1.el5 (OEL 5)
784 * Client support for unpatched kernels:
785 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
786 2.6.18-164.11.1.el5 (RHEL 5),
787 2.6.18-164.11.1.0.1.el5 (OEL 5)
788 2.6.16.60-0.42.8 (SLES 10),
790 2.6.29.4-167.fc11 (FC11)
791 * Recommended e2fsprogs version: 1.41.10-sun2
793 Severity : enhancement
795 Description: Distribute lfs_migrate script with lustre
796 Details : The lfs_migrate mentioned in the manual is now included in the
797 lustre tree in lustre/scripts and will distributed in lustre RPMs.
799 Severity : enhancement
801 Description: Failover nid list added to import proc file.
803 Severity : enhancement
805 Description: "som_preview" mount option required to use SOM
806 Details : The clients will need to use som_preview mount option
810 Description: Lustre does not do 1MB IOs to HW RAID
811 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
813 Severity : enhancement
815 Description: lfs quota output is not very convenient for awk/sed-parsing
816 Details : Some positions in lfs quota output table could be empty or
817 non-empty which made it hard to parse it with scripts, now
818 a dash is put instead of space where there is not supposed
823 Description: Update to OFED 1.5.1
824 Details : OFED 1.5.1 will be used in environments that don't use
825 vendor supplied OFED.
829 Description: Remove lnet. prefix from lctl params display
830 Details : Also renamed lnet.memused to lnet_memused to
835 Description: Use conf_param -d to remove permanant settings.
839 Description: Add list_param -R
840 Details : lctl list_param now supports recursive mode.
843 Bugzilla : 21829, 22312
844 Description: Fix llobdstat.
845 Details : 1. Now updating the write_bytes counter
846 2. Port script fixes from 1.8.
850 Description: decrease the usage of memory on clients.
851 Details : 1. On clients, recycle dentries and inodes unused.
852 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
857 Description: llobdstat fix and enhancement
858 Details : add a counter to set a limit to how many samples will be returned
859 fix a wildcard in the path to limit to obdfilter stats only
863 Description: lfs check only allowed for root
864 Details : Code cleanups and added a regresion test of non-root use of
869 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
870 Details : Fix hr->hr_index overload.
874 Description: Fix FIEMAP functionality.
878 Description: FC11 patchless client support added
882 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
883 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
887 Description: New acc-sm script and various test-framework enhancments.
891 Description: Previously landed patch was reverted.
895 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
899 Description: Add ldiskfs maxdirsize mount option.
903 Description: Fix mkfs.lustre support for 16TB LUNs.
907 Description: Print mmp_check_interval
908 Details : print mmp_check_interval and make it possible to abort
909 mount operation in case it takes too long.
913 Description: Add cascading_rw to lustre mpi tests.
917 Description: Store SOM data in EA.
918 Details : Also several SOM related cleanups.
922 Description: Add refcount to md_open_data
926 Description: Use refcount on lov_request_set to prevent memory corruption.
930 Description: Statahead: use parent dir's i_mutex for synchronization.
934 Description: loadgen improvements.
938 Description: Enable peer health detection in ksocklnd and ko2iblnd.
942 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
946 Description: Fix race on obd_devcie:md_stats
950 Description: Fix SLES 11 client builds.
951 Details : Resolved uninitalized pointer gcc warnings.
955 Description: Allow non-root access for "lfs check".
956 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
960 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
962 Severity : enhancement
964 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
965 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
969 Description: Sanity test_27p fails.
970 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
971 Sanity tests 27 were cleaned up and improved.
975 Description: Unify req format on client/servers
976 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
977 buffers and swabbers approach..
982 Description: drop join file stuff.
984 Severity : enhancement
986 Description: Add 2.6.27 and SLES11 (patchless client) support
990 Description: Fix in ptlrpc_expire_one_request() to print the signed time
993 Severity : enhancement
995 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
998 Severity : enhancement
1000 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
1002 Severity : enhancement
1004 Description: correctly handle big reply message.
1005 Details : send LNet event if reply is bigger than buffer and adjust this buffer
1010 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
1012 Severity : enhancement
1014 Description: Add support for OEL5.
1018 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
1020 Severity : enhancement
1022 Description: "weak-modules" support
1023 Details : Implement "weak-modules" support which enables kernel modules
1024 to be used with any kernel that implements the same kABI. In
1025 order to achieve this modules are now installed in
1026 /lib/modules/$(uname -r)/updates/kernel on all distributions.
1030 Description: client could not reconnect to OST because of an active request.
1031 Details : The client now retries to reconnect to the same server, if a
1032 connect request failed with EBUSY or -EAGAIN. Test case 35b
1033 conf-sanity.sh has been created.
1037 Description: Don't allow make backward step on assignin osc next id.
1038 Details : race between allocation next id and ll_sync thread can be cause
1039 of set wrong osc next id and can be kill valid ost objects.
1043 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
1047 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
1051 Description: Error handling in osc_statfs_interpret() has been improved.
1052 Details : Check in osc_statfs_interpret() for EBADR.
1056 Description: Do not send statfs() requests to OSTs disabled by administrator.
1057 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
1059 Severity : enhancement
1061 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
1063 Severity : enhancement
1065 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
1067 Severity : enhancement
1069 Description: Enhancement for lfs(1) command to use numeric uid/gid.
1073 Description: Improved cleanup procedure for conf-sanity test_46a.
1077 Description: Do not start bulk transfer, if client tried to reconnect.
1078 Details : Do not start bulk transfer, if client tried to reconnect,
1079 but failed due to active rpcs. Check for exp_abort_active_req
1080 has been implemented.
1084 Description: Prevent inconsistences between linux and lustre mount structures.
1085 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
1086 Make the sleep interruptible.
1088 Severity : enhancement
1090 Description: provide server to client comms path
1091 Details : server to client communications path via new LDLM_SET_INFO rpc
1095 Description: truncate starts GFP_FS allocation under transaction causing deadlock
1096 Details : ldiskfs_truncate calls grab_cache_page which may start page
1097 allocation under an open transaction. This may lead to
1098 calling prune_icache with consequent lustre reentrance.
1102 Description: send correctly lsm on open replay
1103 Details : MDS is trust to LSM size on replay open, but client can set wrong size
1108 Description: enable client interface failover
1109 Details : When a child reconnects from another NID, properly update export
1110 nid hash position and ldlm reverse import.
1112 Severity : enhancement
1114 Description: Communicate OST degraded/readonly state via statfs to MDS
1115 Details : Flags in the statfs returned from OSTs indicate whether the
1116 OST is in a degraded RAID state, or if the filesystem has
1117 turned read-only after a filesystem error is detected.
1122 Description: don't panic if EPROTO was hit when reading symlink
1123 Details : correctly handling request reference in error cases.
1125 Severity : enhancement
1127 Description: Add LustreNetLink, a kernel-userspace communcation path.
1129 Severity : enhancement
1131 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1136 Description: access to llog context before init.
1137 Details : move handling CATALOGS file at osc layer and forbid access to llog
1138 context before init.
1141 Frequency : rare, only when client is writing to OST on local node
1143 Description: Avoid deadlock for local client writes
1144 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1145 memory freeing context. This allows OST threads to set the
1146 PF_MEMALLOC flag on task structures in order to allocate memory
1147 from reserved pools and complete IO.
1148 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1149 so that the OST threads generate memory pressure and allow
1150 inactive pages to be reclaimed.
1152 Severity : enhancement
1154 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1159 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1160 Details : move ldlm namespace creation in setup phase to avoid grab
1161 _lprocfs_lock with cli_sem held.
1165 Description: Temporarily disable grant shrink.
1166 Details : Disable the feature for debugging.
1169 Frequency : only when formatting test filesystems
1171 Description: Unable to run several mkfs.lustre on loop devices at the same time
1172 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1173 formatting. The solution is to proper handle the error.
1175 Severity : enhancement
1177 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1179 Severity : enhancement
1181 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1186 Description: resolve race between obd_disconnect and class_disconnect_exports
1187 Details : if obd_disconnect will be called to already disconnected export he
1188 forget release one reference and osc module can't unloaded.
1190 Severity : enhancement
1192 Description: Allow tuning service thread via /proc
1193 Details : For each service a new
1194 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1195 created that can be used to set min/max thread counts, and get the
1196 current number of running threads.
1201 Descriptoin: don't return error if have partially created objects for file.
1202 Details : lov_update_create_set uses set->set_success as index for created
1203 objects, so if some requests will be failed, they will have hole
1204 at end of array and qos_shrink_lsm for allocate correct lsm.
1206 Severity : enhancement
1208 Description: Update OFED support to 1.4.1
1213 Description: Reduce small size read RPC
1214 Details : Set read-ahead limit for every file and only do read-ahead when
1215 available read-ahead pages are bigger than 1M to avoid small size
1218 Severity : enhancement
1220 Description: move AT tunable parameters for more consistent usage
1221 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1223 Severity : enhancement
1225 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1226 Details : allow skip disconnected ost for send statfs request and hide error
1232 Description: fix racy locking of mballoc block bitmaps causing BUG
1233 Details : The locking of the mballoc buddy bitmap and the in-memory
1234 block bitmap was using two different spin locks in some
1235 cases. This made it possible to incorrectly access the
1236 mballoc bitmap while another process was modifying it,
1237 causing a sanity assertion to fail. While no on-disk corruption
1238 was reported, there was some risk of this happening.
1241 Frequency : rare, on llog test 6
1243 Descriptoin: don't allow connect to already connected import
1244 Details : allowing connect to already connected import is hide connecting problem.
1249 Descriptoin: allow kill process which wait statahead result
1250 Details : for some reasons 'ls' can stick in waiting result from statahead,
1251 in this case need way for kill this process.
1253 Severity : enhancement
1255 Description: Add state history info file, enhance import info file
1256 Details : Track import connection state changes in a new osc/mdc proc file;
1257 add overview-type data to the osc/mdc import proc file.
1259 Severity : enhancement
1261 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1263 Severity : enhancement
1265 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1270 Descriptoin: OST grant too much space to client even there are not enough space.
1271 Details : Client will shrink its grant cache to OST if there are no write
1272 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1273 this grant cache if there are already not enough avaible space
1274 (left_space < total_clients * 32M).
1277 Frequency : start MDS on uncleanly shutdowned MDS device
1279 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1280 Details : stay in waiting mds<>ost recovery finished produce random bugs
1281 due race between two ll_sync thread for one lov target. send
1282 ACTIVATE event only if connect realy finished and import have
1286 Frequency : rare, connect and disconnect target at same time
1288 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1289 Details : don't call obd_disconnect under lov_lock. this long time
1290 operation and can block ptlrpcd which answer to connect request.
1295 Descriptoin: don't lose wakeup for imp_recovery_waitq
1296 Details : recover_import_no_retry or invalidate_import and import_close can
1297 both sleep on imp_recovery_waitq, but we was send only one wakeup
1301 Frequency : always with long access acl
1303 Descriptoin: mds can't pack reply with long acl.
1304 Details : mds don't control size of acl but they limited by reint/getattr
1307 Severity : enhancement
1309 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1311 Severity : enhancement
1313 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1316 Frequency : start MDS on uncleanly shutdowned MDS device
1318 Descriptoin: aborting recovery hang on MDS
1319 Details : don't throttle destroy RPCs for the MDT.
1322 Frequency : on remount
1324 Description: external journal device not working after the remount
1325 Details : clear dev_rdonly flag for external journal devices in
1331 Description: shutdown vs evict race
1332 Details : client_disconnect_export vs connect request race.
1333 if client will evicted at this time - we start invalidate
1334 thread without referece to import and import can be freed
1338 Frequency : rare, need acl's on inode.
1340 Description: client can't handle ost additional correctly
1341 Details : if ost was added after client connected to mds client can have
1342 hit lnet_try_match_md ... to big messages to wide striped files.
1343 in this case need teach client to handle config events about add
1344 lov target and update client max ea size at that event.
1346 Severity : enhancement
1348 Description: Changelogs
1349 Details : Changelogs are a lightweight mechanism to track filesystem
1350 metadata and namespace changes. The changelog is recorded
1351 permanently on the MDTs, and is periodically "consumed" / purged
1352 when records are no longer needed.
1354 Severity : enhancement
1356 Description: compact fld format with extents
1357 Details : Store range of seq rather than every seq in FLD. Seq
1358 controller update FLD rather than clients. In Case of CMD, mdt0
1359 has FLD, all other metadata server act as non persistent proxy
1360 for FLD queries and cache fld entries in fld cache.
1365 Description: don't skip ost target if they assigned to file
1366 Details : Drop slow OSCs if we can, but not for requested start idx.
1367 This means "if OSC is slow and it is not the requested
1368 start OST, then it can be skipped, otherwise skip it only
1369 if it is inactive/recovering/out-of-space.
1373 Description: more cleanup in mds_lov
1374 Details : not send LOV EA under replay, we can't know about they size at this
1375 time. Don't allow client connect to mds before any ost connected,
1376 for avoid problems with LOV EA size and returning EIO to client.
1378 Severity : enhancement
1380 Description: Interoperability at server side (Disk interoperability)
1382 Severity : enhancement
1384 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1386 Severity : enhancement
1388 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1390 Severity : enhancement
1392 Description: New client IO stack (CLIO).
1394 Severity : enhancement
1396 Description: Commit on sharing. Eliminate inter-client dependencies between
1397 uncommitted transactions by doing transaction commits.
1398 Thereby clients may recovery independently.
1401 Frequency : Create a symlink file with a very long name
1403 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1404 Details : If there is no extra space in the request for early cancels,
1405 ldlm_req_handles_avail() returns 0 instead of a negative value.
1407 Severity : enhancement
1409 Description: Add /proc entry for import status
1410 Details : The mdc, osc, and mgc import directories now have
1411 an import directory that contains useful import data for debugging
1412 connection problems.
1414 Severity : enhancement
1416 Description: Re-disable certain /proc logging
1417 Details : Enable and disable client's offset_stats, extents_stats and
1418 extents_stats_per_process stats logging on the fly.
1421 Frequency : Only on FC kernels 2.6.22+
1423 Description: oops in statahead
1424 Details : Do not drop reference count for the dentry from VFS when lookup,
1425 VFS will do that by itself.
1427 Severity : enhancement
1429 Description: Generic /proc file permissions
1430 Details : Set /Proc file permissions in a more generic way to enable non-
1431 root users operate on some /proc files.
1435 Description: Hitting mdc_commit_close() ASSERTION
1436 Details : Properly handle request reference release in
1437 ll_release_openhandle().
1441 Description: quota recovery deadlock during mds failover
1442 Details : This patch includes att18982, att18236, att18237 in bz14840.
1444 1. fix osts hang when mds does failover with quotaon
1445 2. prevent watchdog storm when osts threads wait for the
1450 Frequency : only patchless client
1451 Description: add workaround for race between add/remove dentry from hash
1453 Severity : enhancement
1455 Description: Allow OST glimpses to return PW locks
1459 Description: LBUG when llog conf file is full
1460 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1464 Description: Prevent import from entering FULL state when server in recovery
1468 Description: service mount cannot take device name with ":"
1469 Details : Only when device name contains ":/" will mount treat it as
1475 Description: replace ptlrpcd with the statahead thread to interpret the async
1476 statahead RPC callback
1480 Frequency : on recovery
1481 Description: I/O failures after umount during fail back
1482 Details : if client reconnected to restarted server we need join to recovery
1483 instead of find server handler is changed and process self
1484 eviction with cancel all locks.
1486 Severity : enhancement
1488 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1490 Severity : enhancement
1492 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1494 Severity : enhancement
1496 Description: Upcall on Lustre log has been dumped
1497 Details : Allow for a user mode script to be called once a Lustre log has
1498 been dumped. It passes the filename of the dumped log to the
1499 script, the location of the script can be specified via
1500 /proc/sys/lnet/debug_log_upcall.
1505 Description: avoid idr_remove called for id which is not allocated.
1506 Details : Move assigment s_dev for clustered nfs to end of initialization,
1507 to avoid problem with error handling.
1512 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1513 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1514 into CONN_USED_HASH and this prodice warning when put connection
1515 again in unused hash.
1518 Severity : enhancement
1520 Description: File striping can now be set to use an arbitrary pool of OSTs.
1522 Severity : enhancement
1524 Description: Export bytes_read/bytes_write count on OSC/OST.
1528 Description: Early reply size mismatch, MGC loses connection
1529 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1530 the connect flags are properly negotiated.
1535 Description: quotas are not honored with O_DIRECT
1536 Details : all writes with the flag O_DIRECT will use grants which leads to
1537 this problem. Now using OBD_BRW_SYNC to guard this.
1541 Description: add quota statistics
1542 Details : 1. sort out quota proc entries and proc code.
1543 2. add quota statistics
1545 Severity : enhancement
1547 Description: enable quota support for HEAD.
1551 Description: Properly propagate oinfo flags from lov to osc for statfs
1552 Details : restore missing copy oi_flags to lov requests.
1554 Severity : enhancement
1556 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1557 : l_getgroups(8), lst(8), routerstat(8)
1558 Details : included man pages for llobdstat(8), llstat(8),
1559 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1561 Severity : enhancement
1563 Description: configure's --enable-quota should check the
1564 : kernel .config for CONFIG_QUOTA
1565 Details : configure is terminated if --enable-quota is passed but
1566 : no quota support is in kernel
1570 Description: Remove portals compatibility
1571 Details : Remove portals compatibility, not interoperable with releases
1576 Description: Resolve device initialization race
1577 Details : Prevent proc handler from accessing devices added to the
1578 obd_devs array but yet be intialized.
1580 Severity : enhancement
1582 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1584 Severity : enhancement
1586 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1591 Description: Using wrong pointer in osc_brw_prep_request
1592 Details : Access to array[-1] can produce panic if kernel compiled with
1593 CONFIG_PAGE_ALLOC enabled
1597 Description: Client runs out of low memory
1598 Details : Consider only lowmem when counting initial number of llap pages
1602 Description: *optional* service tags registration
1603 Details : if the "service tags" package is installed on a Lustre node
1604 When the filesystem is mounted, a local-node service tag will
1605 be created. See http://inventory.sun.com/ for more information
1606 about the Service Tags asset management system.
1610 Description: Kernel BUG tries to release flock
1611 Details : Lustre does not destroy flock lock before last reference goes
1612 away. So always drop flock locks when client is evicted and
1613 perform unlock regardless of successfulness of speaking to MDS.
1617 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1621 Description: sanity test 65a fails if stripecount of -1 is set
1622 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1627 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1628 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1631 Severity : enhancement
1633 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1637 Description: do not process already freed flock
1638 Details : flock can possibly be freed by another thread before it reaches
1639 to ldlm_flock_completion_ast.
1643 Description: LBUG during stress test
1644 Details : Need properly lock accesses the flock deadlock detection list.
1648 Description: oops in page fault handler
1649 Details : kernel page fault handler can return two special 'pages' in error
1650 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1654 Description: timeout with invalidate import.
1655 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1656 should be handled by ptlrpcd. This produce long age waiting and
1657 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1659 Severity : enhancement
1661 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1665 Description: enable MGS and MDT services start separately
1666 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1667 a co-located MGS without starting the MGS, which is a complement
1668 to 'nosvc' mount option.
1672 Frequency : after recovery
1673 Description: precreate to many object's after del orphan.
1674 Details : del orphan st in oscc last_id == next_id and this triger growing
1675 count of precreated objects. Set flag LOW to skip increase count
1676 of precreated objects.
1680 Frequency : rare, on clear nid stats
1681 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1682 Details : when clean nid stats sometimes try destroy live entry,
1683 and this produce panic in free.
1687 Description: Stack overflow during MDS log replay
1688 ease stack pressure by using a thread dealing llog_process.
1692 Description: wait until IO finished before start new when do lock cancel.
1693 Details : VM protocol want old IO finished before start new, in this case
1694 need wait until PG_writeback is cleared until check dirty flag and
1695 call writepages in lock cancel callback.
1697 Severity : enhancement
1699 Description: using special macro for print time and cleanup in includes.
1703 Description: mds_mfd_close() ASSERTION(rc == 0)
1704 Details : In mds_mfd_close(), we need protect inode's writecount change
1705 within its orphan write semaphore to prevent possible races.
1709 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1713 Frequency : rare, on shutdown ost
1714 Description: don't hit live lock with umount ost.
1715 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1716 use shrink_dcache_sb instead.
1720 Description: don't panic with use echo client
1721 Details : echo client pass NULL as client nid pointer and this produce null
1722 pointer dereference.
1726 Description: fix build on ppc32
1727 Details : compile code with -m64 flag produce wrong object file for ppc32.
1731 Description: add message levels for liblustreapi
1735 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1736 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1737 will result in an -EAGAIN and in the exit call path we call
1738 alloc_rr() which will with increasing aggressiveness attempt to
1739 aquire precreated objects on the minimum number of required OSCs.
1743 Description: Use old size assignment to avoid deadlock
1744 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1745 the scheduling while holding a spinlock. We do not need locking
1746 for size in ll_update_inode() because size is only updated from
1747 the MDS for directories or files without objects, so there is no
1748 other place to do the update, and concurrent access to such inodes
1749 are protected by the inode lock.
1753 Description: resolve "_IOWR redefined" build error on SLES10
1757 Description: dump the memory debugging after all modules are unloaded to
1758 suppress false negative in conf_sanity test 39
1760 Severity : enhancement
1762 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1767 Description: fix SLES kernel versioning
1768 Details : the kernel version for our SLES 10 kernel did not include a "-"
1769 before the "smp" at the end. while this was not a problem in
1770 general, it did mean that software trying to use the kernel
1771 version to try to detect a vendor specific kernel would fail.
1772 this was most evident by the OFED build scripts.
1776 Description: Don't update lov_desc members until making sure they are valid
1777 Details : When updating lov_desc members via proc fs, need fix their
1778 validities before doing the real update.
1782 Description: don't put request into delay list while invalidate in flight.
1783 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1784 import in flight. this produce timeout for invalidate and sometimes
1785 can cause stale data.
1789 Frequency : on ppc only
1790 Description: not convert ost objects for directory because it's not exist.
1791 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1795 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1797 Severity : enhancement
1799 Description: Update to support 2.6.22.14 vanilla kernel.
1803 Frequency : rare, on recovery
1804 Description: read procfs can produce deadlock in some situation
1805 Details : Holding lprocfs lock which send rpc can produce block for destroy
1806 obd objects and this also block reconnect with -EALREADY. This
1807 isn't fix all lprocfs bugs - but make it rare.
1809 Severity : enhancement
1811 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1814 Frequency : frequent on X2 node
1816 Description: mdc_set_open_replay_data LBUG
1817 Details : Set replay data for requests that are eligible for replay.
1821 Description: lustre_mgs: operation 101 on unconnected MGS
1822 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1823 MGC, and late on MGC cannot successfully connect to MGS and a lot
1824 of the error messages complaining that MGS is not connected.
1828 Frequency : on network error
1829 Description: panic with double free request if network error
1830 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1831 but it's true only for synchronus enqueue, for async enqueue
1832 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1835 Severity : enhancement
1837 Description: client-side metadata stat-ahead during readdir(directory readahead)
1838 Details : perform client-side metadata stat-ahead when the client detects
1839 readdir and sequential stat of dir entries therein
1842 Frequency : on start mds
1844 Description: Implement get_info(last_id) in obdfilter.
1847 Frequency : occasional
1849 Description: Correctly check stale fid, not start epoch if ost not support SOM
1850 Details : open with flag O_CREATE need set old fid in op_fid3 because
1851 op_fid2 was overwritten with new generated fid, but mds can answer
1852 with one of these two fids and both is not stale. Setattr
1853 incorrectly started an epoch and assume will be called
1854 done_writing, but without SOM done_writing ever being called.
1857 Frequency : rare, depends on device drivers and load
1859 Description: MDS or OSS nodes crash due to stack overflow
1860 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1861 In some cases, in conjunction with device drivers that use a lot
1862 of stack the MDS (or possibly OSS) service threads could overflow
1863 the stack. One change which was identified to consume additional
1864 stack has been reworked to avoid the extra stack usage.
1867 Frequency : occasional
1869 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1870 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1871 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1872 oscc_has_objects and see OSC has no objests and call
1873 oscc_internal_create to resend create request.
1875 Severity : enhancement
1877 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1879 Severity : enhancement
1881 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1884 Frequency : very rare
1886 Description: Fix replay if there is an un-replied request and open
1887 Details : In some cases, older replay request will revert the
1888 mcd->mcd_last_xid on MDS which is used to record the client's
1889 latest sent request.
1891 Severity : enhancement
1893 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1895 Severity : enhancement
1897 Description: Add rhel5 support to HEAD.
1899 Serverity : enhancement
1901 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1906 Description: Don't allow skipping OSTs if index has been specified.
1907 Details : Don't allow skipping OSTs if index has been specified, make
1908 locking in internal create lots better.
1912 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1913 Details : access to bitfield in structure is always rounded to long
1914 and this produce problem with not atomic change any bit.
1918 Description: Lustre make rpms failed.
1919 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1920 builds Lustre rpms from tarball.
1923 Frequency : rare on shutdown ost
1925 Description: If llog cancel was not send before clean_exports phase, this can
1926 produce deadlock in llog code.
1927 Details : If llog thread has last reference to obd and call class_import_put
1928 this produce deadlock because llog_cleanup_commit_master wait when
1929 last llog_commit_thread exited, but this never success because was
1930 called from llog_commit_thread.
1934 Description: allow userland application know is lost one of stripes.
1935 Details : fill lvb_blocks with error code on ost and return it to
1936 application if error flag found.
1940 Description: NULL lov_tgts causing MDS oops
1941 Details : more safe checks for NULL lov_tgts for avoid oops.
1943 Severity : enhancement
1945 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1949 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1950 Details : If an OST is down the MDS will hang indefinitely in
1951 obd_statfs_rqset() waiting for the statfs data. While for
1952 MDS QOS usage of statfs, it should not stuck in waiting.
1954 Severity : enhancement
1956 Description: remote_acl support
1957 Details : Support ACL-based permission check for remote user.
1958 Support setfacl/getfacl for remote user with the utils
1959 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1960 the system "{s,g}etfacl" utils.
1962 Severity : enhancement
1964 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1966 Severity : enhancement
1968 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1972 Description: make mgs_setparam() handle fsname containing dash
1973 Details : fsname containing a dash does not work with lctl conf_param
1975 Severity : enhancement
1977 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1979 Severity : enhancement
1981 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1985 Frequency : rare, at shutdown
1986 Description: access already free / zero obd_namespace.
1987 Details : if client_disconnect_export was called without force flag set,
1988 and exist connect request in flight, this can produce access to
1989 NULL pointer (or already free pointer) when connect_interpret
1990 store ocd flags in obd_namespace.
1994 Frequency : only at startup
1995 Description: not alloc memory with spinlock held.
1996 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1997 if any spinlock held.
1999 Severity : enhancement
2001 Description: make lustre randomly fail allocating memory
2002 Details : Make lustre randomly failed allocating memory for testing purpose.
2004 Severity : enhancement
2006 Description: lost problems with lov objid file
2007 Details : Fixes some scability and access to not inited memory problems
2008 in work with lov objdid file.
2013 Description: lfs find does not continue on file error
2014 Details : Continue other files processing when a file/dir is absent.
2018 Description: Inconsistent usage of lustre_pack_reply()
2019 Details : Standardize the usage of lustre_pack_reply() such that it
2020 always generate a CERROR on failure.
2023 Frequency : occasional
2025 Description: MDS hang or stay in waiting lock
2026 Details : If client receive lock with CBPENDING flag ldlm need send lock
2027 cancel as separate rpc, to avoid situation when cancel request
2028 can't processed due all i/o threads stay in wait lock.
2032 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
2036 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
2040 Description: Update to SLES9 kernel-2.6.5-7.287.3.
2044 Description: Update to RHEL4 latest kernel.
2046 Severity : enhancement
2048 Description: Build SLES10 patchless client fails
2049 Details : The configure was broken by run ./configure with
2050 --with-linux-obj=.... argument for patchless client. When the
2051 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
2052 can't search header adequately. Use absolute path such as
2053 -I($LINUX)/include instead.
2057 Description: interrupt oig_wait produce painc on resend.
2058 Details : brw_redo_request can be used for resend requests from ptlrpcd and
2059 private set, and this produce situation when rq_ptlrpcd_data not
2060 copyed to new allocated request and triggered LBUG on assert
2061 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
2062 ptlrpcd set if request is changed and can be safety changed to use
2067 Description: LASSERT_{REQ,REP}SWAB macros are buggy
2068 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
2069 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
2070 these macros and replace them with their difinitions instead.
2074 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
2075 Details : Update which_patch & target file for SLES10 latest kernel.
2078 Description: Remove client patches from SLES 10 kernel.
2079 Details : This causes SLES 10 clients to behave as patchless clients
2080 even on a Lustre-patched (server) kernel.
2082 Severity : enhancement
2084 Description: self-adjustable client's lru lists
2085 Details : use adaptive algorithm for managing client cached locks lru
2086 lists according to current server load, other client's work
2087 pattern, memory activities, etc. Both, server and client
2088 side namespaces provide number of proc tunables for controlling
2091 Severity : enhancement
2093 Description: light-weight GSS support
2094 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
2095 and reduce performance overhead.
2097 Severity : enhancement
2099 Description: Linux keyring support
2100 Details : Support using service of Linux keyring for Lustre GSS internal
2101 context refresh/cache mechanism.
2105 Description: Fix errors in lfs documentation
2106 Details : Fixes man pages
2110 Description: don't use GFP_* in generic Lustre code.
2111 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2115 Description: obdclass is limited by single OBD_ALLOC(idarray)
2116 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2120 Description: warnings with build patchless client with vanila 2.6.19 and up
2121 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2122 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2126 Frequency : Only for SLES
2127 Description: sanity_quota fail test_1
2128 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2129 file, which makes TSTID[2] inunique.
2133 Frequency : Only for SLES9
2134 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2135 Details : Update target/ChangeLog/which_patch .
2139 Frequency : Only for test_10 in sanity-quota.sh
2140 Description: a bug in quota test code
2141 Details : checking a test flag in a wrong place
2145 Frequency : Only for test_13 in sanity-quota.sh
2146 Description: a bug in quota test script
2147 Details : shouldn't check overfull allocated quota
2151 Description: jbd statistics
2152 Details : Port older jbd statistics patch for sles10
2154 Severity : enhancement
2156 Description: Add printing inode info into message about error in writepage.
2161 Description: reply_lock_interpret crash due to race with it and lock cancel.
2162 Details : Do not replay locks that are being cancelled. Do not reference
2163 locks by their address during replay, just by their handle.
2167 Frequency : When flocks are used.
2168 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2169 Details : Flock locks might destroy just granted lock if it could be merged
2170 with another existing flock, this is done in completion handler,
2171 so teach ldlm_cli_enquque_fini that this is a valid case for
2177 Description: Oops in read and write path when failing to allocate lock.
2178 Details : Check if lock allocation failed and return error back.
2182 Description: lstripe command fails for valid OST index
2183 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2184 instead of lov->desc.ld_active_tgt_count.
2188 Description: sanity.sh failed test 103
2189 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2194 Description: df doesn't work properly if diskfs blocksize != 4K
2195 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2199 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2200 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2202 Severity : enhancement
2204 Description: Adaptive timeouts
2205 Details : RPC timeouts adapt to changing server load and network
2206 conditions to reduce resend attempts and improve recovery time.
2210 Description: llapi_file_create() does not allow some changes
2211 Details : add llapi_file_open() that allows specifying the mode and
2212 open flags, and also returns an open file handle.
2216 Description: merge and cleanup kernel patches.
2217 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2221 Description: Add journal checksum support.(Kernel part)
2222 Details : The journal checksum feature adds two new flags i.e
2223 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2224 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2225 indicates that the commit block contains the checksum for
2226 the blocks described by the descriptor blocks. Now commit
2227 record can be sent to disk without waiting for descriptor
2228 blocks to be written to disk. This behavior is controlled
2229 using JBD2_FEATURE_ASYNC_COMMIT flag.
2233 Description: OSS needs mutliple precreate threads
2234 Details : Add ability to start more than one create thread per OSS.
2238 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2239 Details : execute lfs setstripe on client
2243 Description: mds_obd_create error creating tmp object
2244 Details : When the user sets quota on root, llog will be affected and can't
2245 create files and write files.
2248 Frequency : Always on ia64 patchless client, and possibly others.
2250 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2251 Details : This allows the patchless client to be loaded on architectures
2252 without this export.
2256 Description: RedHat Update kernel for RHEL5
2257 Details : Add the kernel config file for RHEL5.
2261 Description: RedHat Update kernel for RHEL5
2262 Details : Modify the kernel config file more closer RHEL5.
2266 Description: Build failure against Centos5 (RHEL5)
2267 Details : Define PAGE_SIZE when it isn't present.
2271 Description: test_7 on sanity-quota.sh can't be used on separate machines
2272 Details : add facet to handle it
2276 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2277 Details : with NFS, the anon dentry's parent was set to itself in
2278 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2279 corresponding dentry other than use rec->ur_name.
2281 Severity : enhancement
2283 Description: lfs setstripe enhancement
2284 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2288 Description: enable data checksumming by default
2289 Details : enable checksum by default, allow --disable-checksum
2290 configure option and "-o nochecksum" mount option. Checksums
2291 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2295 Description: lustre support for RHEL5
2296 Details : Add support for RHEL5.
2300 Description: Client eviction due to failover config
2301 Details : after a connection loss, the lustre client should attempt to
2302 reconnect to the last active server first before trying the
2303 other potential connections.
2307 Description: when mds and osts use different quota unit(32bit and 64bit),
2308 quota will be released repeatly.
2309 Details : void sending multiple quota reqs to mds, which will keep the
2310 status between the reqs.
2314 Description: osts not allocated evenly to files
2315 Details : change the condition to increase offset_idx
2318 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2320 Description: Data corruption for OSTs that are formatted larger than 2TB
2322 Details : When generating the bio request for lustre file writes the
2323 sector number would overflow a temporary variable before being
2324 used for the IO. The data reads correctly from Lustre (which
2325 will overflow in a similar manner) but other file data or
2326 filesystem metadata may be corrupted in some cases.
2330 Description: Tune the kernel for good SCSI performance.
2331 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2332 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2337 Description: rewrite ext2-derived code in obdclass/uuid.c
2338 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2339 from scratch preserving functionality.
2343 Description: "lfs find -obd UUID" prints directories
2344 Details : "lfs find -obd UUID" will return all directory names instead
2345 of just file names. It is incorrect because the directories
2346 do not reside on the OSTs.
2350 Description: use i_size_read and i_size_write in 2.6 port
2351 Details : replace inode->i_size access with i_size_read/write()
2354 Frequency : when removing large files
2356 Description: scheduling issue during removal of large Lustre files
2357 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2358 It causes scheduling issues when removing large files (17TB in the
2362 Frequency : only with liblustre clients on XT3
2364 Description: evictions taking too long
2365 Details : allow llrd to evict clients directly on OSTs
2368 Frequency : only on ppc
2370 Description: /proc/fs/lustre/devices broken on ppc
2371 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2372 We should make sure the type of 'index' is loff_t and then cast
2373 explicitly as needed below. Do not assign an explicitly cast
2377 Frequency : only for rhel5
2379 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2380 Details : Modify the target file & which_kernel.
2382 Severity : enhancement
2384 Description: omit set fsid for export NFS
2385 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2386 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2390 Frequency : Always, for kernels after 2.6.16
2391 Description: Fix warning idr_remove called for id=.. which is not allocated.
2392 Details : Last kernels save old s_dev before kill super and not allow
2393 to restore from callback - restore it before call kill_anon_super.
2397 Description: buffer overruns could theoretically occur
2398 Details : llapi_semantic_traverse() modifies the "path" argument by
2399 appending values to the end of the origin string, and a
2400 overrun may occur. Adding buffer overrun check in liblustreapi.
2404 Description: Fix error on 'ls .' at the top of the Lustre mount.
2405 Details : Don't revalidate dentry if it is a root dentry.
2409 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2410 Details : Modify vm-tunables-rhel4.patch.
2414 Description: Kernel config for 2.6.18-vanilla.
2415 Details : Modify targets/2.6-vanilla.target.in.
2416 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2417 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2418 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2419 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2423 Description: improve handling recoverable errors
2424 Details : if request processig with error which can be recoverable on server
2425 request should be resend, otherwise page released from cache and
2430 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2431 Details : Modify target file & which_patch.
2432 A flaw was found in the IA32 system call emulation provided
2433 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2434 value could be stored in the %RAX register, which could trigger an
2435 out-of-bounds system call table access. An untrusted local user
2436 could exploit this flaw to run code in the kernel
2437 (ie a root privilege escalation). (CVE-2007-4573).
2441 Description: O_DIRECT bypasses client statistics.
2442 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2443 read_ahead_stats were updated. copied the stats section from
2444 osc_send_oap_rpc() into async_internal().
2448 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2452 Description: change order of libsysio includes
2453 Details : '#include sysio.h' should always come before '#include xtio.h'
2457 Description: handle "serious error: objid * already exists" more gracefully
2458 Details : If LAST_ID value on disk is smaller than the objects existing in
2459 the O/0/d* directories, it indicates disk corruption and causes an
2460 LBUG(). If the object is 0-length, then we should use the existing
2461 object. This will help to avoid a full fsck in most cases.
2463 Severity : enhancement
2465 Description: adapt the lustre_config script to support the upgrade case
2466 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2467 server targets to 1.6 in parallel.
2471 Description: To avoid grant space > avaible space when the disk is almost
2472 full. Without this patch you might see the error "grant XXXX >
2473 available" or some LBUG about grant, when the disk is almost
2475 Details : In filter_check_grant, for non_grant cache write, we should
2476 check the left space by if (*left > ungranted + bytes), instead
2477 of (*left > ungranted), because only we are sure the left space
2478 is enough for another "bytes", then the ungrant space should be
2479 increase. In client, we should update cl_avail_grant only there
2480 is OBD_MD_FLGRANT in the reply.
2484 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2485 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2486 described in CVE-2007-4573. This problem could allow untrusted
2487 local users to gain root access.
2490 Frequency : when using O_DIRECT and quotas
2492 Description: Incorrect file ownership on O_DIRECT output files
2493 Details : block usage reported by 'lfs quota' does not take into account
2494 files that have been written with O_DIRECT.
2499 Description: touch file failed when fs is not full
2500 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2501 otherwise we can get ENOSP while fs is not full.
2505 Description: parallel lock callbacks
2506 Details : Instead of sending blocking and completion callbacks as separated
2507 requests, adding them to a set and sending in parallel.
2510 Frequency : only for Cray XT3
2511 Bugzilla : 12829/13455
2512 Description: Changing primary group doesn't change the group lustre assigns to
2514 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2515 overridden with the primary group provided by the group upcall,
2516 whereas the supplied fsgid can be trusted if it is in the list of
2517 supplementary groups returned by the group upcall.
2519 Severity : enhancement
2521 Description: Allow masking D_WARNING, D_ERROR messages from console
2522 Details : Console messages can now be disabled via lnet.printk.
2526 Description: User code with malformed file open parameter crashes client node
2527 Details : Before packing join_file req, all the related reference should be
2528 checked carefully in case some malformed flags cause fake
2529 join_file req on client.
2533 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2534 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2535 mode is equal to the granted mode. However, it should also check
2536 that LDLM_FL_CP_REQD is not set.
2540 Description: Heavy nfs access might result in deadlocks
2541 Details : After ELC code landed, it is now improper to enqueue any mds
2542 locks under och_sem, because enqueue might want to decide to
2543 cancel open locks for same inode we are holding och_sem for.
2547 Description: Client eviction while running blogbench
2548 Details : A lot of unlink operations with concurrent I/O can lead to a
2549 deadlock causing evictions. To address the problem, the number of
2550 oustanding OST_DESTROY requests is now throttled to
2551 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2552 callbacks are processed in priority.
2556 Description: enable ACLs on MDS by default
2557 Details : ACLs must be enabled on MDS by default.
2560 Frequency : PPC/PPC64 only
2562 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2563 Details : fix duplicated definitions between asm-ppc64/types.h and
2564 lustre_types.h on PPC.
2567 Frequency : PPC/PPC64 only
2569 Description: asm-ppc/segment.h does not exist
2570 Details : fix compile issue on PPC.
2574 Description: better handle error messages in extents code
2577 Frequency : RHEL4 only
2579 Description: mkfs is very slow on IA64/RHEL4
2580 Details : A performance regression has been discovered in the MPT Fusion
2581 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2582 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2583 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2585 Severity : enhancement
2587 Description: SNMP support enhancement
2588 Details : Adding total number of sampled request for an MDS node in snmp
2591 Severity : enhancement
2593 Description: Optimize ldlm waiting list processing for PR extent locks
2594 Details : When processing waiting list for read extent lock and meeting read
2595 lock that is same or wider to it that is not contended, skip
2596 processing rest of the list and immediatelly return current
2597 status of conflictness, since we are guaranteed there are no
2598 conflicting locks in the rest of the list.
2602 Description: Time out and refuse to reconnect
2603 Details : When the failover node is the primary node, it is possible
2604 to have two identical connections in imp_conn_list. We must
2605 compare not conn's pointers but NIDs, otherwise we can defeat
2606 connection throttling.
2610 Description: port llog fixes from b1_6 into HEAD
2611 Details : Port llog reference couting and some llog cleanups from b1_6
2612 (bug 10800) into HEAD, for protect from panic and access to already
2613 free llog structures.
2617 Description: Detect stride IO mode in read-ahead
2618 Details : When a client does stride read, read-ahead should detect that and
2619 read-ahead pages according to the detected stride pattern.
2623 Description: data checksumming impacts single node performance
2624 Details : add support for several checksum algorithm. Currently, only CRC32
2625 and Adler-32 are supported. The checksum type can be changed on
2626 the fly via /proc/fs/lustre/osc/*/checksum_type.
2630 Description: use adler32 for page checksums
2631 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2636 Description: build for x2 fails
2637 Details : fix compile issue on Cray systems.
2641 Description: Properly match for duplicate locks
2642 Details : Due to different lock order from skiplists code, we need to
2643 traverse entire list for now
2646 Frequency : only on PPC/SLES10
2648 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2649 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2650 be defined. Add a hack in mkfs_lustre.c to work around this compile
2655 Description: LASSERT on MDS when client holding flock lock dies
2656 Details : ldlm pool logic depends on number of granted locks equal to
2657 number of released locks which is not true for flock locks, so
2658 just exclude such locks from consideration.
2662 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2663 Details : Use fsfilt_sync() for both the whole filesystem sync and
2664 individual file sync to eliminate dangerous inode locking
2665 with I_LOCK that can lead to a deadlock.
2669 Description: performance in 1.6.3
2670 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2671 sure that 1MB requests are not fragmented by the block layer.
2673 Severity : enhancement
2675 Description: organize the server-side client stats on per-nid basis
2676 Details : Change the structure of stats under obdfilter and mds to
2684 The "uuid"s file would list the uuids of _active_ exports.
2685 And the clear entry is to clear all stats and stale nids.
2687 Severity : enhancement
2689 Description: eliminate client locks in face of contention
2690 Details : file contention detection and lockless i/o implementation
2691 for contended files.
2695 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2698 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2700 Description: MDS may lose file striping (and hence file data) in some cases
2701 Details : If there are additional extended attributes stored on the MDS,
2702 in particular ACLs, SELinux, or user attributes (if user_xattr
2703 is specified for the client mount options) then there is a risk
2704 of attribute loss. Additionally, the Lustre file striping
2705 needs to be larger than default (e.g. striped over all OSTs),
2706 and an additional attribute must be stored initially in the
2707 inode and then increase in size enough to be moved to the
2708 external attribute block (e.g. ACL growing in size) for file
2713 Description: skiplist implementation simplification
2714 Details : skiplists are used to group compatible locks on granted list
2715 that was implemented as tracking first and last lock of each
2716 lock group the patch changes that to using doubly linked lists
2720 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2721 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2722 because LASSERT on both the data supplied by a client, and the
2723 data on disk is dangerous and incorrect.
2725 Severity : enhancement
2727 Description: Slow truncate/writes to huge files at high offsets.
2728 Details : Directly associate cached pages to lock that protect those pages,
2729 this allows us to quickly find what pages to write and remove
2730 once lock callback is received.
2734 Description: more ldlm soft lockups
2735 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2736 starve other threads from the resource lock for a long time in
2737 case of long waiting queue, so change the debug level from
2738 D_OTHER to the less frequently used D_INFO.
2740 Severity : enhancement
2742 Description: add -gid, -group, -uid, -user options to lfs find
2746 Description: Hung threads in invalidate_inode_pages2_range
2747 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2748 once one is completed. As a result, some RPCs are stuck in the
2749 queue and are never sent.
2753 Description: filter threads hungs on waiting journal commit
2754 Details : Cleanup filter group llog code, then only filter group llog will
2755 be only created in the MDS/OST syncing process.
2759 Description: Procfs and llog threads access destoryed import sometimes.
2760 Details : Sync the import destoryed process with procfs and llog threads by
2761 the import refcount and semaphore.
2763 Severity : enhancement
2765 Description: openlock cache of b1_6 port to HEAD
2770 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2771 Details : If insertion of an extent fails, then discard the inode
2772 preallocation and free data blocks else it can lead to duplicate
2777 Description: don't always update ctime in ext3_xattr_set_handle()
2778 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2779 In some cases the ctime should not be updated, for example for
2780 2.0->1.8 compatibility it is necessary to delete an xattr and it
2781 should not update the ctime.
2785 Bugzilla : 15713/16362
2786 Description: Assertion in iopen_connect_dentry in 1.6.3
2787 Details : looking up an inode via iopen with the wrong generation number can
2788 populate the dcache with a disconneced dentry while the inode
2789 number is in the process of being reallocated. This causes an
2790 assertion failure in iopen since the inode's dentry list contains
2791 both a connected and disconnected dentry.
2795 Description: assertion failure in ldlm_handle2lock()
2796 Details : fix a race between class_handle_unhash() and class_handle2object()
2797 introduced in lustre 1.6.5 by bug 13622.
2802 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2803 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2808 Description: oops in sd_iostats_seq_show()
2809 Details : unloading/reloading the scsi low level driver triggers a kernel
2810 bug when trying to access the sd iostat file.
2815 Description: Kernel panics during QLogic driver reload
2816 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2817 patch, causing memory corruption.
2822 Description: journal_dev option does not work in b1_6
2823 Details : pass mount option during pre-mount.
2825 Severity : enhancement
2827 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2828 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2829 extent information of a file. It can be used to map logical blocks
2830 in a file to physical blocks in the block device.
2834 Description: LDLM soft lockups - improvement
2835 Details : It is be possible to send the lock handle along with each read
2836 or write request because the client is already doing a lock match
2837 itself so there isn't any reason the OST should have to re-do that
2841 Frequency : only on Cray X2
2843 Description: X2 build failures
2844 Details : fix build failures on Cray X2.
2848 Description: xid & resent requests
2849 Details : Initialize RPC XID from clock at startup (randomly if clock is
2852 Severity : enhancement
2854 Description: Add lustre_start utility to start or stop multiple Lustre servers
2859 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2860 Details : In case of memory pressure, list_del() can be called twice on
2861 req->rq_history_list, causing a kernel oops.
2865 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2866 Details : fix stack overflow in the distributed lock manager by defering
2867 export eviction after a failed AST to the elt thread instead of
2868 handling it in the dlm interpret routine.
2872 Description: Convert some comments to new format.
2873 Details : Update documenting comments to match doxygen conventions.
2877 Description: Grammar fixes.
2878 Details : A couple of trivial spelling fixes.
2882 Description: OSD_COUNTERS-mandatory
2883 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2884 to simplify the code.
2888 Description: simplify cmm_device freeing logic.
2889 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2893 Description: Add lockdep support to dt_object_operations locking interface.
2894 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2895 parameter indicating lock ordering. Update mdd code to use new
2900 Description: Introduce failloc constants for lockless IO tests.
2901 Details : Add two new failloc constants to test lockless IO. Only one of
2902 them in implemented---another is checked in yet to be landed
2907 Description: Add lockdep support for inode mutex.
2908 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2909 in the situations where Lustre has to lock more than one inode
2914 Description: Add optional invariants checking support.
2915 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2916 configure switch. This macro is to be used for consistency and
2917 sanity checks that are too expensive to be left in `production'
2922 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2923 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2928 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2929 Details : Use spin_lock_nested() in (the only) situation where more than
2930 one ldlm_lock is locked simultaneously. Also, fix possible
2931 dead-lock in ldlm_lock_change_resource() by enforcing particular
2936 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2937 Details : Instead of specifying each ldlm_lock call-back through separate
2938 parameter, wrap them into struct ldlm_callback_suite.
2942 Description: Kill join_lru obd method and its callers.
2943 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2944 in memory---a requirement implemented through ->o_join_lru() obd
2945 method in HEAD. Remove this method and its users.
2949 Description: Add asynchronous ldlm ENQUEUE completion handler.
2950 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2951 case that stock ldlm_completion_ast() cannot handle as it waits
2952 until lock is granted. Introduce new ldlm_completion_ast_async()
2953 for this. Also comment ldlm_completion_ast().
2957 Description: ldlm_error <-> errno conversion.
2958 Details : Add functions to map (rather arbitrary) between LDLM error codes
2959 and standard errno values. CLIO needs this to prevent LDLM specific
2960 constants from escaping ldlm and osc.
2964 Description: Kill unused ldlm_handle2lock_ns() function.
2965 Details : Kill unused ldlm_handle2lock_ns() function.
2969 Description: Add lu_ref support to ldlm_lock
2970 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2971 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2972 and ldlm_resource. LDLM interface has to be changed, because code that
2973 releases a reference on a lock, has to "know" what reference this is.
2974 In the most frequent case
2976 lock = ldlm_handle2lock(handle);
2978 LDLM_LOCK_PUT(lock);
2980 no changes are required. When any other reference (received _not_ from
2981 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2982 instead of LDLM_LOCK_PUT().
2984 Arguably, changes are pervasive, and interface requires some discipline
2985 for proper use. On the other hand, it was very instrumental in finding
2986 a few leaked lock references.
2990 Description: Add ldlm_lock_addref_try().
2991 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2992 attempts to addref a lock that might be being canceled
2997 Description: Add ldlm_weigh_callback().
2998 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2999 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
3000 hard-coded `number of pages' logic.
3004 Description: Add lockdep annotations to llog code.
3005 Details : Use appropriately tagged _nested() locking calls in the places
3006 where llog takes more than one ->lgh_lock lock.
3010 Description: Add loi_kms_set().
3011 Details : Wrap kms updates into a helper function.
3015 Description: Constify instances of struct lsm_operations.
3016 Details : Constify instances of struct lsm_operations.
3020 Description: lu_conf support.
3021 Details : On a server, a file system object is uniquely identified
3022 by a fid, which is sufficient to locate and load all object
3023 state (inode). On a client, on the other hand, more data are
3024 necessary instantiate an object. Change lu_object_find() and
3025 friends to take additional `lu_conf' argument describing object.
3026 Typically this includes layout information.
3030 Description: lu_context fixes.
3031 Details : Introduce new lu_context functions that are needed on the client
3032 side, where some system threads (ptlrpcd) are shared by multiple
3033 modules, and so cannot be stopped during module shutdown.
3037 Description: Add start and stop methods to lu_device_type_operations.
3038 Details : Introduce two new methods in lu_device_type_operations, that are
3039 invoked when first instance of a given type is created and last one
3040 is destroyed respectively. This is need by CLIO.
3044 Description: Add lu_ref support to struct lu_device.
3045 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
3046 track leaked references.
3050 Description: Introduce lu_kmem_descr.
3051 Details : lu_kmem_descr and its companion interface allow to create
3052 and destroy a number of kmem caches at once.
3056 Description: Fix lu_object finalization race.
3057 Details : Fix a race between lu_object_find() finding an object and its
3058 concurrent finalization. This race is (most likely) not possible
3059 on the server, but might happen on the client.
3063 Description: Introduce lu_ref interface.
3064 Details : lu_ref is a debugging module allowing to track references to
3065 a given object. It is quite cpu expensive, and has to be
3066 explicitly enabled with --enable-lu_ref. See usage description
3071 Description: Factor lu_site procfs stats into a separate function.
3072 Details : Separate lu_site stats printing code into a separate function
3073 that can be reused on a client.
3077 Description: Constify instances of struct {lu,dt,md}_device_operations.
3078 Details : Constify instances of struct {lu,dt,md}_device_operations.
3082 Description: Introduce struct md_site and move meta-data specific parts of
3083 struct lu_site here.
3084 Details : Move md-specific fields out of struct lu_site into special struct
3085 md_site, so that lu_site can be used on a client.
3089 Description: Kill mdd_lov_destroy().
3090 Details : Remove unused mdd code.
3094 Description: Add st_block checking to multistat.c.
3095 Details : Add st_block checking to multistat.c.
3099 Description: Add lu_ref support to struct obd_device.
3100 Details : Add lu_ref tracking to obd_device.
3104 Description: Kill obd_set_fail_loc().
3105 Details : Remove unused code.
3109 Description: Add special type for ptlrpc_request interpret functions.
3110 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
3111 there. Actual usage will be in CLIO.
3115 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
3116 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
3117 doesn't work with lockdep.
3121 Description: Add rwv.c test program.
3122 Details : New testing program exercising readv(2) and writev(2) (Qian).
3126 Description: Add sendfile.c test program.
3127 Details : New testing program exercising sendfile(2) (Jay).
3131 Description: Ratelimit a message that can be very frequent.
3132 Details : Ratelimit a memory allocation failure message that can
3137 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3138 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3142 Description: Liblustre build fixes.
3143 Details : Liblustre build fixes.
3147 Description: libcfs: add cfs_{need,cond}_resched() interface.
3148 Details : libcfs: add cfs_{need,cond}_resched() definition and
3149 implementations for Linux, NT, and liblustre.
3151 Severity : enhancement
3153 Description: More exported tunables for mballoc
3154 Details : Add support for tunable preallocation window and new tunables for
3155 large/small requests
3159 Description: Detect corruption of block bitmap and checking for preallocations
3160 Details : Checks validity of on-disk block bitmap. Also it does better
3161 checking of number of applied preallocations. When corruption is
3162 found, it turns filesystem readonly to prevent further corruptions.
3166 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3167 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3168 window to 0. If it does hit cache miss, and read-pattern is still
3169 stride-io mode, does not reset the stride window, but also does not
3170 increase the stride window length in this case.
3174 Frequency : only for big-endian servers
3175 Description: Check if system is big-endian while mounting fs with extents feature
3176 Details : Mounting a filesystem with extents feature will fail on big-endian
3177 systems since ext3-based ldiskfs is not supported on big-endian
3178 systems. This can be over-riden with "bigendian_extents" mount option.
3180 Severity : enhancement
3182 Description: The root squash functionality
3183 Details : A security feature, which is to prevent users from being able
3184 to mount lustre on their desktop, run as root, and delete
3185 all of the files in the filesystem. The goal is accomplished by
3186 remapping user id (UID) and group id (GID) of the root user to
3187 a UID and GID specified by the system administartor via Lustre
3188 configuration management server (MGS). The functionality also
3189 allows to specify sets of clients for which the remapping does
3194 Description: Excessive recovery window
3195 Details : With AT enabled, the recovery window can be excessively long (6000+
3196 seconds). To address this problem, we no longer use
3197 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3198 timeout no longer depends on the service time, it is set to
3199 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3200 time via pb_service_time.
3204 Description: Watchdog triggered on MDS failover
3205 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3206 the OSTs know that the MDS "UUID" can be reused for the same export
3207 from a different NID, so we do not need to wait for the export to be
3211 Frequency : rare, only if using MMP with Linux RAID
3213 Description: MMP doesn't work with Linux RAID
3214 Details : While using HA for Lustre servers with Linux RAID, it is possible
3215 that MMP will not detect multiple mounts. To make this work we
3216 need to unplug the device queue in RAID when the MMP block is being
3217 written. Also while reading the MMP block, we should read it from
3218 disk and not the cached one.
3220 Severity : enhancement
3222 Description: open file using fid
3223 Details : A file can be opened using just its fid, like
3224 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3227 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3230 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3231 Details : add proper locking for creation of jbd_4k slab cache
3235 Description: MMP check in ext3_remount() fails without displaying any error
3236 Details : When multiple mount protection fails during remount, proper error
3239 Severity : enhancement
3241 Description: Allow stripe size to be up to 4G-64k
3242 Details : Fix math logic to allow large stripe sizes.
3246 Description: add check for >8TB ldiskfs filesystems
3247 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3248 Don't allow >8TB ldiskfs filesystems to be mounted without
3249 force_over_8tb mount option
3254 Descriptoin: OSS complains "too many blocks" and crashes
3255 Details : In filter_commitrw_write():
3256 1. lquota_chkquota();
3257 2. obdo_from_inode();
3258 3. lquota_pending_commit();
3259 "2." has changed uid and gid in this case, then caused
3260 lquota_chkquota() and lquota_pending_commit() to use
3261 different uid and gid. That is the root of the bug.
3265 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3266 Details : Detect and print out the 'stripe count = -1' case properly.
3270 Description: Out or order replies might be lost on replay
3271 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3272 request with tid smaller than one currently being added, add it
3273 to the start, not end of the list.
3275 --------------------------------------------------------------------------------
3277 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3279 * Support for kernels:
3280 2.6.5-7.283 (SLES 9),
3281 2.6.9-55.EL (RHEL 4),
3282 2.6.16.46-0.14 (SLES 10),
3283 2.6.18.8 vanilla (kernel.org)
3284 * Client support for unpatched kernels:
3285 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3286 2.6.16 - 2.6.22 vanilla (kernel.org)
3287 * Due to recently discovered recovery problems, we do not recommend
3288 using patchless RHEL 4 clients with this or any earlier release.
3289 * Recommended e2fsprogs version: 1.39.cfs8
3290 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3291 * Starting with this release, the ldiskfs backing filesystem required
3292 by Lustre is now in its own package, lustre-ldiskfs. This package
3293 should be installed. It is versioned separately from Lustre and
3294 may be released separately in future.
3298 Description: block reactivating mgc import until all deactivates complete
3299 Details : Fix race when failing back MDT/MGS to itself (testing)
3302 Frequency : at statup only
3304 Description: mds_lov_synchronize race leads to various problems
3305 Details : simultaneous MDT->OST connections at startup can cause the
3306 sync to abort, leaving the OSC in a bad state.
3308 Severity : enhancement
3310 Description: add optional extra BUILD_VERSION info
3311 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3312 to override the lustre version.
3315 Frequency : 2.6.18 servers only
3317 Description: ll_kern_mount() doesn't release the module reference
3318 Details : The ldiskfs module reference count never drops down to 0
3319 because ll_kern_mount() doesn't release the module reference.
3324 Description: server LBUG when using old ost_num_threads parameter
3325 Details : Accept the old ost_num_threads parameter but warn that it
3326 is deprecated, and fix an off-by-one error that caused an LBUG.
3331 Description: Transient SCSI error results in persistent IO issue
3332 Details : iobuf->dr_error is not reinitialized to 0 between two
3336 Frequency : sometimes when underlying device returns I/O errors
3338 Description: OSTs not going read-only during write failures
3339 Details : OSTs are not remounted read-only when the journal commit threads
3340 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3341 instead of the ext3 wrappers.
3345 Description: poor connect scaling with increasing client count
3346 Details : Don't run filter_grant_sanity_check for more than 100 exports
3347 to improve scaling for large numbers of clients.
3350 Frequency : SLES10 only
3352 Description: sanity-quota.sh quotacheck failed: rc = -22
3353 Details : Quotas cannot be enabled on SLES10.
3356 Frequency : liblustre clients only
3358 Description: getdirentries does not give error when run on compute nodes
3359 Details : getdirentries does not fail when the size specified as an argument
3360 is too small to contain at least one entry
3362 Severity : enhancement
3364 Description: Add LNET router traceability for debug purposes
3365 Details : If a checksum failure occurs with a router as part of the
3366 IO path, the NID of the last router that forwarded the bulk data
3367 is printed so it can be identified.
3372 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3373 Details : Due to a race condition, liblustre clients were occasionally
3374 evicted incorrectly.
3376 Severity : enhancement
3378 Description: lfs setstripe use optional parameters instead of postional
3381 Severity : enhancement
3383 Description: Nanosecond timestamp support for ldiskfs
3384 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3385 resolution timestamps. There is not yet support for this at
3386 the Lustre filesystem level.
3389 Frequency : during server recovery
3391 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3392 Details : request with rq_no_resend flag not awake l_wait_event if they get
3396 Frequency : nfs export on patchless client
3398 Description: connectathon hang when test nfs export over patchless client
3399 Details : Disconnected dentry cannot be found with lookup, so we do not need
3400 to unhash it or make it invalid
3403 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3404 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3405 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3410 Description: open req refcounting wrong on reconnect
3411 Details : If reconnect happened between getting open reply from server and
3412 call to mdc_set_replay_data in ll_file_open, we will schedule
3413 replay for unreferenced request that we are about to free.
3414 Subsequent close will crash in variety of ways.
3415 Check that request is still eligible for replay in
3416 mdc_set_replay_data().
3421 Description: disable writes to filesystem when reading health_check file
3422 Details : the default for reading the health_check proc file has changed
3423 to NOT do a journal transaction and write to disk, because this
3424 can cause reads of the /proc file to hang and block HA state
3425 checking on a healthy but otherwise heavily loaded system. It
3426 is possible to return to the previous behaviour during configure
3427 with --enable-health-write.
3429 Severity : enhancement
3431 Description: 64-bit inode version
3432 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3433 to the inode. This will be required for version-based recovery.
3438 Description: MDS fails to start if a duplicate client export is detected
3439 Details : in some rare cases it was possible for a client to connect to
3440 an MDS multiple times. Upon recovery the MDS would detect this
3441 and fail during startup. Handle this more gracefully.
3443 Severity : enhancement
3445 Description: Add -o localflock option to simulate old noflock
3447 Details : This will achieve local-only flock/fcntl locks
3453 Description: log_commit_thread vs filter_destroy race leads to crash
3454 Details : Take import reference before releasing llog record semaphore
3459 Description: Wrong request locking in request set processing
3460 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3461 imp_delayed_list, in this place should be used imp_lock.
3464 Frequency : when reconnection
3466 Description: Grant Leak when osc reconnect to OST
3467 Details : When osc reconnect ost, OST(filter) should check whether it
3468 should grant more space to client by comparing fed_grant and
3469 cl_avail_grant, and return the granted space to client instead
3470 of "new granted" space, because client will call osc_init_grant
3471 to update the client grant space info.
3474 Frequency : when client reconnect to OST
3476 Description: Grant Leak when osc do resend and replay bulk write
3477 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3478 bulk write request, because the grant info while be sync between
3479 OSC and OST when reconnect, and we should ignore the grant info
3480 these of resend/replay write req.
3485 Description: Grant space more than available left space sometimes.
3486 Details : When then OST is about to be full, if two bulk writing from
3487 different clients came to OST. Accord the avaliable space of the
3488 OST, the first req should be permitted, and the second one
3489 should be denied by ENOSPC. But if the seconde arrived before
3490 the first one is committed. The OST might wrongly permit second
3491 writing, which will cause grant space > available space.
3494 Frequency : when client is evicted
3496 Description: Grant might be wrongly erased when osc is evicted by OST
3497 Details : when the import is evicted by server, it will fork another
3498 thread ptlrpc_invalidate_import_thread to invalidate the
3499 import, where the grant will be set to 0. While the original
3500 thread will update the grant it got when connecting. So if
3501 the former happened latter, the grant will be wrongly errased
3502 because of this race.
3507 Description: Checking Stale with correct fid
3508 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3509 to check whether it is stale, because sometimes, we want the
3510 enqueue happened anyway, and op_data.fid2 will not be initialized.
3512 Severity : enhancement
3514 Description: update patchless client
3515 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3518 Frequency : only with 2.4 kernel
3520 Description: random memory corruption
3521 Details : size of struct ll_inode_info is to big for union inode.u and this
3522 can be cause of random memory corruption.
3527 Description: Memory leak in recovery
3528 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3529 It should also check obd_fail before fsfilt_start, otherwise if
3530 fsfilt_start return -EROFS,(failover mds during mds recovery).
3531 then the req will return with repmsg->transno = 0 and rc = EROFS.
3532 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3533 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3534 be freed no matter whether fsfilt_commit success or not.
3537 Frequency : only with huge count clients
3539 Description: Prevents from taking the superblock lock in llap_from_page for
3541 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3542 for llap_from_page call in ll_removepage prevents from taking the
3543 superblock lock for a soon died page.
3548 Description: Not check open intent error before release open handle
3549 Details : in some rare cases, the open intent error is not checked before
3550 release open handle, which may cause
3551 ASSERTION(open_req->rq_transno != 0), because it tries to release
3552 the failed open handle.
3557 Description: Set cat log bitmap only after create log success.
3558 Details : in some rare cases, the cat log bitmap is set too early. and it
3559 should be set only after create log success.
3563 Description: Accessing a block bevice can re-enable I/O when Lustre is
3564 tearing down a device.
3565 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3570 Description: service threads may hog cpus when there are a lot of requests
3571 Details : Insert cond_resched to give other threads a chance to use some CPU
3576 Description: the cat log was not initialized in recovery
3577 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3578 unlink log on mds will not be initialized until mds post
3579 recovery. And also in mds post recovery, the unlink log will
3580 initialization will be done asynchronausly, so there will be race
3581 between add unlink log and unlink log initialization.
3585 Description: brw_stats were being printed incorrectly
3586 Details : brw_stats were being printed as log2 but all of them were not
3587 recorded as log2. Also remove some code duplication arising from
3588 filter_tally_{read,write}.
3592 Frequency : rare, only in recovery.
3593 Description: ASSERTION(req->rq_type != LI_POISON) failed
3594 Details : imp_lock should be held while iterating over imp_sending_list for
3595 prevent destroy request after get timeout in ptlrpc_queue_wait.
3599 Description: replay-single.sh test 52 fails
3600 Details : A lock's skiplist need to be cleanup when it being unlinked
3601 from its resource list.
3605 Description: Short directio read returns full requested size rather than
3607 Details : Direct I/O operations should return actual amount of bytes
3608 transferred rather than requested size.
3610 Severity : enhancement
3612 Description: metadata RPC reduction (e.g. for rm performance)
3613 Details : decrease the amount of synchronous RPC between clients and servers
3614 by canceling conflicing lock before the operation on the client
3615 and packing thier handles into the main operation RPC to server.
3617 Severity : enhancement
3619 Description: Async OSC create to avoid the blocking unnecessarily.
3620 Details : If an OST has no remaining object, system will block on the
3621 creation when it needs to create a new object on this OST. Now,
3622 ways use pre-created objects when available, instead of blocking on
3623 an empty osc while others are not empty. If we must block, we
3624 block for the shortest possible period of time.
3628 Description: improve handling recoverable errors
3629 Details : if request processig with error which can be recoverable on server
3630 request should be resend, otherwise page released from cache and
3633 Severity : enhancement
3635 Description: refine locking for avoid write wrong info into lov_objid file
3636 Details : fix possible races with add new target and write/update data in
3639 --------------------------------------------------------------------------------
3641 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3646 Frequency : on some architectures
3648 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3649 Details : Uninitialized flags sometimes cause configuration commands to
3653 Frequency : patchless clients only
3655 Description: missing __iget() symbol export
3656 Details : The __iget() symbol export is missing. To avoid the need for
3657 this on patchless clients the deathrow inode reaper is turned
3658 off, and we depend on the VM to clean up old inodes. This
3659 dependency was during via the fix for bug 12181.
3661 --------------------------------------------------------------------------------
3663 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3665 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3666 INTEROPERATE with older versions automatically. In many cases a
3667 special upgrade step is needed. Please read the
3668 user documentation before upgrading any part of a live system.
3669 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3670 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3671 * WARNING: Lustre configuration and startup changes are required with
3672 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3676 Severity : enhancement
3678 Description: MountConf
3679 Details : Lustre configuration is now managed via mkfs and mount
3680 commands instead of lmc and lconf. New obd types (MGS, MGC)
3681 are added for dynamic configuration management. See
3682 https://mail.clusterfs.com/wikis/lustre/MountConf for
3685 Severity : enhancement
3687 Description: dynamic OST addition
3688 Details : OSTs can now be added to a live filesystem
3690 Severity : enhancement
3692 Description: startup order invariance
3693 Details : MDTs and OSTs can be started in any order. Clients only
3694 require the MDT to complete startup.
3696 Severity : enhancement
3698 Description: parallel, asynchronous orphan cleanup
3699 Details : orphan cleanup is now performed in separate threads for each
3700 OST, allowing parallel non-blocking operation.
3702 Severity : enhancement
3704 Description: optimized stripe assignment
3705 Details : stripe assignments are now made based on ost space available,
3706 ost previous usage, and OSS previous usage, in order to try
3707 to optimize storage space and networking resources.
3709 Severity : enhancement
3711 Description: Permanently set tunables
3712 Details : All writable /proc/fs/lustre tunables can now be permanently
3713 set on a per-server basis, at mkfs time or on a live system.
3715 Severity : enhancement
3717 Description: Lustre message v2
3718 Details : Add lustre message format v2.
3720 Severity : enhancement
3722 Description: client OST exclusion list
3723 Details : Clients can be started with a list of OSTs that should be
3724 declared "inactive" for known non-responsive OSTs.
3728 Description: SPEC SFS validation failure on NFS v2 over lustre.
3729 Details : Changes the blocksize for regular files to be 2x RPC size,
3730 and not depend on stripe size.
3732 Severity : enhancement
3734 Description: Multiple MD RPCs in flight.
3735 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3736 To avoid overly-overloading MDS, introduce a limit on number of
3737 MDS RPCs in flight for a single client and add /proc controls
3738 to adjust this limit.
3740 Severity : enhancement
3742 Description: client read/write statistics
3743 Details : Add client read/write call usage stats for performance
3744 analysis of user processes.
3745 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3746 file access. extents_stats shows chunk size distribution.
3747 extents_stats_per_process show chunk size distribution per
3750 Severity : enhancement
3752 Description: mds statistics
3753 Details : Add detailed mds operations statistics in
3754 /proc/fs/lustre/mds/*/stats.
3758 Description: Failure of copying files with lustre special EAs.
3759 Details : Client side always return success for setxattr call for lustre
3760 special xattr (currently only "trusted.lov").
3765 Description: Refcount LNET uuids
3766 Details : The global LNET uuid list grew linearly with every startup;
3767 refcount repeated list entries instead of always adding to
3770 Severity : enhancement
3772 Description: Dynamic service threads
3773 Details : Within a small range, start extra service threads
3774 automatically when the request queue builds up.
3777 Frequency : mixed-endian client/server environments
3779 Description: mixed-endian crashes
3780 Details : The new msg_v2 system had some failures in mixed-endian
3783 Severity : enhancement
3785 Description: Easy OST removal
3786 Details : OSTs can be permanently deactivated with e.g. 'lctl
3787 conf_param lustre-OST0001.osc.active=0'
3789 Severity : enhancement
3791 Description: MGS proc entries
3792 Details : Added basic proc entries for the MGS showing what filesystems
3795 Severity : enhancement
3797 Description: provide MGS failover
3798 Details : Added config lock reacquisition after MGS server failover.
3800 Severity : enhancement
3802 Description: add Linux 2.4 support
3803 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3807 Description: a large application tries to do I/O to the same resource and dies
3808 in the middle of it.
3809 Details : Check the req->rq_arrival time after the call to
3810 ost_brw_lock_get(), but before we do anything about
3811 processing it & sending the BULK transfer request. This
3812 should help move old stale pending locks off the queue as
3813 quickly as obd_timeout.
3816 Frequency : when an incorrect nid is specified during startup
3818 Description: ptlrpc connect to non-existant node causes kernel crash
3819 Details : LNET can't be re-entered from an event callback, which
3820 happened when we expire a message after the export has been
3821 cleaned up. Instead, hand the zombie cleanup off to another
3824 Severity : enhancement
3826 Description: plain/inodebits lock performance improvement
3827 Details : Grouping plain/inodebits in granted list by their request modes
3828 and bits policy, thus improving the performance of search through
3832 Frequency : only if OST filesystem is corrupted
3834 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3835 Details : for a short time RPCs with bulk IO are in the replay list,
3836 but replay of bulk IOs is unimplemented. If the OST filesystem
3837 is corrupted due to disk cache incoherency and then replay is
3838 started it is possible to trip an assertion. Avoid putting
3839 committed RPCs into the replay list at all to avoid this issue.
3842 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3844 Description: System hang on startup
3845 Details : This bug allowed the liblustre (e.g. catamount) client to
3846 return to the app before handling all startup RPCs. This
3847 could leave the node unresponsive to lustre network traffic
3848 and manifested as a server ptllnd timeout.
3850 Severity : enhancement
3852 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3853 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3854 boolean to control whether to print peer debug info when a
3855 client's RPC times out.
3858 Frequency : only for kernels with patches from Lustre below 1.4.3
3860 Description: Remove old rdonly API
3861 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3864 Frequency : only for devices with external journals
3866 Description: Set external device read-only also
3867 Details : During a commanded failover stop, we set the disk device
3868 read-only while the server shuts down. We now also set any
3869 external journal device read-only at the same time.
3872 Frequency : when upgrading from 1.4 while trying to change parameters
3874 Description: The wrong (new) MDC name was used when setting parameters for
3875 upgraded MDT's. Also allows changing of OSC (and MDC)
3876 parameters if --writeconf is specified at tunefs upgrade time.
3879 Frequency : when setting specific ost indicies
3881 Description: QOS code breaks on skipped indicies
3882 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3883 created with --index need not be sequential.
3885 Severity : enhancement
3887 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3888 Details : The uninit_groups feature works in conjunction with the kernel
3889 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3890 up the pass1 processing of e2fsck. This is a read-only feature
3891 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3892 mount filesystems that have had this feature enabled.
3894 Severity : enhancement
3896 Description: Improve multi-block allocation algorithm to avoid fragmentation
3897 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3898 allocation locality and avoid filesystem fragmentation.
3900 ------------------------------------------------------------------------------
3902 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3904 * Support for kernels:
3905 2.4.21-47.0.1.EL (RHEL 3)
3906 2.6.5-7.283 (SLES 9)
3907 2.6.9-42.0.10.EL (RHEL 4)
3908 2.6.12.6 vanilla (kernel.org)
3909 2.6.16.27-0.9 (SLES 10)
3910 * Recommended e2fsprogs version: 1.39.cfs5
3912 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3916 Frequency : occasional, depends on client load and configuration
3917 Bugzilla : 12181, 12203
3918 Description: data loss for recently-modified files
3920 Details : In some cases it is possible that recently written or created
3921 files may not be written to disk in a timely manner (this should
3922 normally be within 30s unless client IO load is very high).
3923 The problem appears as zero-length files or files that are a
3924 multiple of 1MB in size after a client crash or client eviction
3925 that are missing data at the end of the file.
3927 This problem is more likely to be hit on clients where files are
3928 repeatedly created and unlinked in the same directory, clients
3929 have a large amount of RAM, have many CPUs, the filesystem has
3930 many OSTs, the clients are rebooted frequently, and/or the files
3931 are not accessed by other nodes after being written.
3933 The presence of the problem can be detected by looking at
3934 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3935 smaller than the second (nr_unused) then dirty files will not
3936 be flushed automatically to disk. "sync; sleep 10" should be
3937 run several times on the node before unmounting it to update
3938 Lustre (this is also safe to run on nodes without this problem).
3940 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3941 that can cause this same problem, so customers using that kernel
3942 also need to update the kernel in addition to Lustre. In order
3943 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3945 It is normal that files written just before a client crash (less
3946 than 30s) may not yet have been flushed to disk, even for local
3950 Frequency : frequent on thin XT3 nodes
3952 Description: UUID collision on thin XT3 Linux nodes
3953 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3954 randomly, since we relied on an insufficiently-seeded PRNG.
3959 Description: OSS hangs after "All ost request buffers busy"
3960 Details : A deadlock between quota and journal operations caused OSS
3961 hangs after printing "All ost request buffers busy."
3964 Frequency : always on liblustre builds
3966 Description: Cleanup compiler warnings on liblustre
3969 Frequency : always on liblustre builds on XT3
3971 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3975 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3976 > 32000 subdirectories
3977 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3978 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3979 subdirectory count crosses 32000. This will aid e2fsck to
3980 correctly handle more than 32000 subdirectories.
3983 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3985 Description: System hang on startup
3986 Details : This bug allowed the liblustre (e.g. catamount) client to
3987 return to the app before handling all startup RPCs. This
3988 could leave the node unresponsive to lustre network traffic
3989 and manifested as a server ptllnd timeout.
3991 Severity : enhancement
3993 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3994 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3995 boolean to control whether to print peer debug info when a
3996 client's RPC times out.
4001 Description: make O_SYNC working on 2.6 kernels
4002 Details : 2.6 kernels use different method for mark pages for write,
4003 so need add a code to lustre for O_SYNC work.
4008 Description: Failure to close file and release space on NFS
4009 Details : Put inode details into lock acquired in ll_intent_file_open.
4010 Use mdc_intent_lock in ll_intent_open to properly
4011 detect all kind of errors unhandled by mdc_enqueue.
4016 Description: proc file read during shutdown sometimes raced obd removal,
4018 Details : Add lock to prevent obd access after proc file removal.
4021 Frequency : Only for files larger than 4GB on 32-bit clients.
4023 Description: improperly doing page alignment of locks
4024 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
4025 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
4030 Description: RPCs being resent when they shouldn't be
4031 Details : Some RPCs that should not be resent are being resent. This
4032 can cause inconsistencies in the RPC state machine. Do not
4033 resend such requests.
4036 Frequency : rare, only with NFS export
4038 Description: Crash on NFS re-export node
4039 Details : under very unusual load conditions an assertion is hit in
4040 ll_intent_file_open()
4043 Frequency : only if OST filesystem is corrupted
4045 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4046 Details : for a short time RPCs with bulk IO are in the replay list,
4047 but replay of bulk IOs is unimplemented. If the OST filesystem
4048 is corrupted due to disk cache incoherency and then replay is
4049 started it is possible to trip an assertion. Avoid putting
4050 committed RPCs into the replay list at all to avoid this issue.
4055 Description: large O_DIRECT requests fail under memory pressure/fragmentation
4056 Details : Large single O_DIRECT read and write calls can fail to allocate
4057 a sufficiently large buffer to process the request. In case of
4058 allocation failure the allocation is retried with a smaller
4059 buffer and broken into smaller requests.
4061 Severity : enhancement
4063 Description: Add -o localflock option to simulate old noflock behaviour.
4064 Details : This will achieve local-only flock/fcntl locks coherentness.
4069 Description: versioning check is incomplete
4070 Details : Checking the version difference of client vs. server, report
4071 error if the gap is too big.
4076 Description: add support PG_writeback bit
4077 Details : add support for PG_writeback bit for Lustre, for more carefull
4078 work with page cache in 2.6 kernel. This also fix some deadlocks
4079 and remove hack for work O_SYNC with 2.6 kernel.
4081 Severity : enhancement
4083 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4084 Details : The uninit_groups feature works in conjunction with the kernel
4085 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4086 up the pass1 processing of e2fsck. This is a read-only feature
4087 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4088 mount filesystems that have had this feature enabled.
4090 Severity : enhancement
4092 Description: Improve multi-block allocation algorithm to avoid fragmentation
4093 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4094 allocation locality and avoid filesystem fragmentation.
4096 ------------------------------------------------------------------------------
4098 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
4100 * Support for kernels:
4101 2.6.9-42.0.3.EL (RHEL 4)
4102 2.6.5-7.276 (SLES 9)
4103 2.4.21-47.0.1.EL (RHEL 3)
4104 2.6.12.6 vanilla (kernel.org)
4105 2.6.16.21-0.8 (SLES10)
4106 * Recommended e2fsprogs version: 1.39.cfs2-0
4108 * The backwards-compatible /proc/sys/portals symlink has been removed
4109 in this release. Before upgrading, please ensure that you change
4110 any configuration scripts or /etc/sysctl.conf files that access
4111 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4112 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4113 in advance of the upgrade on any system running Lustre 1.4.6 or
4114 newer, since /proc/sys/lnet was added in that version.
4115 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4119 Frequency : only when quota is used
4121 Description: avoid scanning export list for quota master
4122 Details : Change the algorithms to avoid scanning export list in order
4123 to improve the efficiency.
4126 Frequency : MDS failover only, very rarely
4128 Description: "went back in time" messages on mds failover
4129 Details : The greatest transno may be lost when the current operation
4130 finishes with an error (transno==0) and the client's last_rcvd
4131 record is over-written. Save the greatest transno in the
4132 mds_last_transno for this case.
4135 Frequency : always for specific kernels and striping counts
4137 Description: client may get "Matching packet too big" without ACL support
4138 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4139 when trying to access files in certain configurations. The
4140 clients should in fact be denied when mounting because they do
4141 not understand ACLs.
4144 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4146 Description: many clients connecting with IO in progress causes connect timeouts
4147 Details : Avoid synchronous journal commits to avoid delays caused by many
4148 clients connecting/disconnecting when bulk IO is in progress.
4149 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4150 OST_IO_PORTAL to avoid delays behind potentially many pending
4154 Frequency : occasionally with multiple writers to a single file
4156 Description: shared writes to file may result in wrong size reported by stat()
4157 Details : Allow growing of kms when extent lock is cancelled
4160 Frequency : always with random mmap IO to multi-striped file
4162 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4163 Details : Only if the hole is at the end of OST object so that kms is too
4164 small. Fix is to increase kms accordingly in ll_nopage.
4167 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4169 Description: writes to a missing object would leak memory on the OST
4170 Details : If there is an inconsistency between the MDS and OST filesystems,
4171 such that the MDS references an object that doesn't exist, writes
4172 to that object will leak memory due to incorrect cleanup in the
4173 error handling path, eventually running out of memory on the OST.
4178 Description: Creating too long symlink causes lustre errors
4179 Details : Check symlink and name lengths before sending requests to MDS.
4182 Frequency : only if flock is enabled (not on by default)
4184 Description: posix locks not released on fd closure on 2.6.9+
4185 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4186 kernels, this caused such locks not to be released on fd close and
4187 then assertions on fs unmount about still used locks.
4190 Frequency : MDS failover only, very rarely
4192 Description: clients may get ASSERTION(granted_lock != NULL)
4193 Details : When request was taking a long time, and a client was resending
4194 a getattr by name lock request. The were multiple lock requests
4195 with the same client lock handle and
4196 mds_getattr_name->fixup_handle_for_resent_request found one of the
4197 lock handles but later failed with ASSERTION(granted_lock != NULL).
4202 Description: handle->h_buffer_credits > 0, assertion failure
4203 Details : h_buffer_credits is zero after truncate, causing assertion
4204 failure. This patch extends the transaction or creates a new
4208 Frequency : NFS re-export or patchless client
4209 Bugzilla : 11179, 10796
4210 Description: Crash on NFS re-export node (__d_move)
4211 Details : We do not want to hash the dentry if we don't have a lock.
4212 But if this dentry is later used in d_move, we'd hit uninitialised
4213 list head d_hash, so we just do this to init d_hash field but
4214 leave dentry unhashed.
4217 Frequency : NFS re-export or patchless client
4219 Description: NFS exports has problem with symbolic link
4220 Details : lustre client didn't properly install dentry when re-exported
4221 to NFS or running patchless client.
4224 Frequency : NFS re-export or patchless client
4226 Description: Various nfs/patchless fixes.
4227 Details : fixes reuse disconected alias for lookup process - this fixes
4228 warning "find_exported_dentry: npd != pd",
4229 fix permission error with open files at nfs.
4230 fix apply umask when do revalidate.
4233 Frequency : occasional
4235 Description: Crash on NFS re-export node
4236 Details : calling clear_page() on the wrong pointer triggered oops in
4237 generic_mapping_read().
4240 Frequency : rarely, using O_DIRECT IO
4242 Description: unaligned directio crashes client with LASSERT
4243 Details : check for unaligned buffers before trying any requests.
4246 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4248 Description: stale data returned from RAID cache
4249 Details : If only a small amount of IO is done to the RAID device before
4250 reading it again it is possible to get stale data from the RAID
4251 cache instead of reading it from disk.
4254 Frequency : always for sles10 kernel
4256 Description: sles10 support
4257 Details : ll_follow_link: compile fixes and using of nd_set_link
4258 under newer kernels.
4261 Frequency : depends on arch, kernel and compiler version, always on sles10
4264 Description: recursive or deep enough symlinks cause stack overflow
4265 Details : getting rid of large stack-allocated variable in
4269 Frequency : depends on hardware
4271 Description: lustre write performance loss in the SLES10 kernel
4272 Details : the performance loss is caused by using of write barriers in the
4273 ext3 code. The SLES10 kernel turns barrier support on by
4274 default. The fix is to undo that change for ldiskfs.
4276 ------------------------------------------------------------------------------
4278 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4280 * Support for kernels:
4281 2.6.9-42.0.3EL (RHEL 4)
4282 2.6.5-7.276 (SLES 9)
4283 2.4.21-47.0.1.EL (RHEL 3)
4284 2.6.12.6 vanilla (kernel.org)
4288 Frequency : quota enabled and large files being deleted
4290 Description: releasing more than 4GB of quota at once hangs OST
4291 Details : If a user deletes more than 4GB of files on a single OST it
4292 will cause the OST to spin in an infinite loop. Release
4293 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4298 Description: statfs data retrieved from /proc may be stale or zero
4299 Details : When reading per-device statfs data from /proc, in the
4300 {kbytes,files}_{total,free,avail} files, it may appear
4301 as zero or be out of date.
4304 Frequency : systems with MD RAID1 external journal devices
4306 Description: lconf's call to blkid is confused by RAID1 journal devices
4307 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4308 of returning all block devices that match the journal UUID.
4311 Frequency : always, for aggregate stripe size over 4GB
4313 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4314 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4315 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4318 Frequency : always if "lfs find" used on a local file/directory
4320 Description: "lfs find" segfaults if used on a local file/directory
4321 Details : The case where a directory component was not specified wasn't
4322 handled correctly. Handle this properly.
4325 Frequency : always on ppc64
4327 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4328 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4329 perform endianness conversion.
4332 Frequency : rarely (truncate to non-zero file size after write under load)
4333 Bugzilla : 10730, 10687
4334 Description: Files padded with zeros to next 4K multiple
4335 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4336 it is possible that files that are truncated to a non-zero size
4337 immediately after being written are filled with zero bytes beyond
4338 the truncated size. No file data is lost.
4340 Severity : enhancement
4342 Description: Allow recovery/failover for liblustre clients.
4343 Details : liblustre clients were unaware of failover configurations until
4346 Severity : enhancement
4348 Description: user file locks should fail when not mounting with flock option
4349 Details : Set up an error-returning stub in ll_file_operations.lock field
4350 to prevent incorrect behaviour when client is mounted without
4351 flock option. Also, set up properly f_op->flock field for
4355 Frequency : always on ia64
4357 Description: "lfs df" loops on printing out MDS statfs information
4358 Details : The obd_ioctl_data was not initialized and in some systems
4359 this caused a failure during the ioctl that did not return
4360 an error. Initialize the struct and return an error on failure.
4363 Frequency : SLES 9 only
4365 Description: Error of copying files with lustre special EAs as root
4366 Details : Client side always return success for setxattr call for lustre
4367 special xattr (currently only "trusted.lov").
4370 Frequency : rarely on clusters with both ia64+i386 clients
4372 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4373 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4374 clients are doing concurrent writes to the same file it is
4375 possible that the ia64 clients may LASSERT because the OST
4376 extent locks are not PAGE_SIZE aligned. Ensure that grown
4377 locks are always aligned on the request boundary.
4380 Frequency : specific use, occasional
4382 Description: Overwriting in use executable truncates on-disk binary image
4383 Details : If one node attempts to overwrite an executable in use by
4384 another node, we now correctly return ETXTBSY instead of
4385 truncating the file.
4387 Severity : enhancement
4389 Description: Async OSC create to avoid the blocking unnecessarily.
4390 Details : If an OST has no remaining object, system will block on the creating
4391 when need to create a new object on this OST. Now, ways use
4392 pre-created objects when available, instead of blocking on an
4393 empty osc while others are not empty. If we must block, we block
4394 for the shortest possible period of time.
4399 Description: chmod on Lustre root is propagated to other clients
4400 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4401 invalid by the follow_mount time.
4404 Frequency : liblustre clients only
4406 Description: Race in 'instant cancel' lock handling could lead to such locks
4407 never to be granted in case of SMP MDS
4408 Details : Do not destroy not yet granted but cbpending locks in
4412 Frequency : replay/resend of open
4414 Description: non null lock assetion failure in mds_intent_policy
4415 Details : Trying to replay/resend lockless open requests resulted in
4416 mds_open() returning 0 with no lock. Now it sets a flag if
4417 a lock is going to be returned.
4419 Severity : enhancement
4421 Description: Checksum enhancements
4422 Details : New checksum enhancements allow for resending RPCs that failed
4425 Severity : enhancement
4427 Description: Tunables on number of dirty pages in cacche
4428 Details : Allow to set limit on number of dirty pages cached.
4433 Description: client crash on unmount - lock still has references
4434 Details : In some error handling cases it was possible to leak a lock
4435 reference on a client while accessing a file. This was not
4436 harmful to the client during operation, but would cause the
4437 client to crash when the filesystem is unmounted.
4440 Frequency : specific case, rare
4442 Description: ETXTBSY on mds though file not in use by client
4443 Details : ETXTBSY is no longer incorrectly returned when attempting to
4444 chmod or chown a directory that the user previously tried to
4445 execute or a currently-executing binary.
4448 Frequency : extremely rare except on liblustre-based clients
4450 Description: Lustre space not freed when files are deleted
4451 Details : Clean up open-unlinked files after client eviction. Previously
4452 the unlink was skipped and the files remained as orphans.
4457 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4458 Details : In some cases it was possible to send a blocking callback to
4459 a client doing a glimpse, even though that client didn't get
4460 a lock granted. When the glimpse lock is cancelled on the OST
4461 the freed lock is left on the waiting list and corrupted the list.
4464 Frequency : all core dumps
4466 Description: Broke core dumps to lustre
4467 Details : Negative dentry may be unhashed if parent does not have UPDATE
4468 lock, but some callers, e.g. do_coredump, expect dentry to be
4469 hashed after successful create, hash it in ll_create_it.
4471 ------------------------------------------------------------------------------
4473 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4475 * Support for kernels:
4476 2.6.9-42.0.2.EL (RHEL 4)
4477 2.6.5-7.276 (SLES 9)
4478 2.4.21-40.EL (RHEL 3)
4479 2.6.12.6 vanilla (kernel.org)
4483 Frequency : always on RHEL 3
4485 Description: Number of open files grows over time
4486 Details : The number of open files grows over time, whether or not
4487 Lustre is started. This was due to a filp leak introduced
4488 by one of our kernel patches.
4490 ------------------------------------------------------------------------------
4492 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4494 * Support for kernels:
4495 2.6.9-42.EL (RHEL 4)
4496 2.6.5-7.276 (SLES 9)
4497 2.4.21-40.EL (RHEL 3)
4498 2.6.12.6 vanilla (kernel.org)
4503 Bugzilla : 5719, 9635, 9792, 9684
4504 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4505 Details : If a server is under heavy load and cannot reply to new
4506 connection requests before the client resends the (re)connect,
4507 the connection handling code can behave badly if two service
4508 threads are concurrently handing separate (re)connections from
4509 the same client. Add better locking to the connection handling
4510 code, and ensure that only a single connection will be processed
4511 for a given client UUID, even if the lock is dropped.
4513 Severity : enhancement
4515 Description: add TCP zero-copy support to kernel
4516 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4517 sends if the hardware supports scatter-gather and checksumming.
4518 This allows socklnd to do client-write and server-read more
4519 efficiently and reduce CPU utilization from skbuf copying.
4522 Frequency : only if NFS exporting from client
4524 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4525 Details : If setting attributes on a file created under NFS that had
4526 never been opened it would be possible to oops the client
4527 if the file had no objects.
4530 Frequency : always for liblustre
4532 Description: liblustre client does MDS+OSTs setattr RPC for each write
4533 Details : When doing a write from a liblustre client, the client
4534 incorrectly issued an RPC to the MDS and each OST the file was
4535 striped over in order to update the timestamps. When writing
4536 with small chunks and many clients this could overwhelm the MDS
4537 with RPCs. In all cases it would slow down the write because
4538 these RPCs are unnecessary.
4540 Severity : enhancement
4542 Description: allow number of MDS service threads to be changed at module load
4543 Details : It is now possible to change the number of MDS service threads
4544 running. Adding "options mds mds_num_threads={N}" to the MDS's
4545 /etc/modprobe.conf will set the number of threads for the next
4546 time Lustre is restarted (assuming the "mds" module is also
4547 reloaded at that time). The default number of threads will
4548 stay the same, 32 for most systems.
4553 Description: OST crash if filesystem is unformatted or corrupt
4554 Details : If an OST is started on a device that has never been formatted
4555 or if the filesystem is corrupt and cannot even mount then the
4556 error handling cleanup routines would dereference a NULL pointer.
4561 Description: NULL pointer deref in llap_from_page.
4562 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4563 incorrect) mapping if the page was truncated/reclaimed while it was
4564 searched for. Check for this condition and skip such pages when
4565 doing readahead. Introduce extra check to llap_from_page() to
4566 verify page->mapping->host is non-NULL (so page is not anonymous).
4569 Frequency : Sometimes when using sys_sendfile
4571 Description: "page not covered by a lock" warnings from ll_readpage
4572 Details : sendfile called ll_readpage without right page locks present.
4573 Now we introduced ll_file_sendfile that does necessary locking
4574 around call to generic_file_sendfile() much like we do in
4578 Frequency : with certain MDS communication failures at client mount time
4580 Description: NULL pointer deref after failed client mount
4581 Details : a client connection request may delayed by the network layer
4582 and not be sent until after the PTLRPC layer has timed out the
4583 request. If the client fails the mount immediately it will try
4584 to clean up before the network times out the request. Add a
4585 reference from the request import to the obd device and delay
4586 the cleanup until the network drops the request.
4589 Frequency : occasionally during client (re)connect
4591 Description: assertion failure during client (re)connect
4592 Details : processing a client connection request may be delayed by the
4593 client or server longer than the client connect timeout. This
4594 causes the client to resend the connection request. If the
4595 original connection request is replied in this interval, the
4596 client may trip an assertion failure in ptlrpc_connect_interpret()
4597 which thought it would be the only running connect process.
4600 Frequency : only with obd_echo servers and clients that are rebooted
4602 Description: kernel BUG accessing uninitialized data structure
4603 Details : When running an obd_echo server it did not start the ping_evictor
4604 thread, and when a client was evicted an uninitialized data
4605 structure was accessed. Start the ping_evictor in the RPC
4606 service startup instead of the OBD startup.
4608 Severity : enhancement
4609 Bugzilla : 10193 (patchless)
4610 Description: Remove dependency on various unexported kernel interfaces.
4611 Details : No longer need reparent_to_init, exit_mm, exit_files,
4612 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4615 Frequency : rare (only users of deprecated and unsupported LDAP config)
4617 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4618 Details : LDAP apparently contains 'lustreName' attributes instead of
4619 'name'. A simple remapping of the name is sufficient.
4622 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4624 Description: waiting_locks_callback trips kernel BUG if client is evicted
4625 Details : Running with the dump_on_timeout debug flag turned on makes
4626 it possible that the waiting_locks_callback() can try to dump
4627 the Lustre kernel debug logs from an interrupt handler. Defer
4628 this log dumping to the expired_lock_main() thread.
4630 Severity : enhancement
4632 Description: Support NFS exporting on 2.6 kernels.
4633 Details : Implement non-rawops metadata methods for NFS server to use without
4634 changing NFS server code.
4637 Frequency : very rare (synthetic metadata workload only)
4639 Description: two racing renames might cause an MDS thread to deadlock
4640 Details : Running the "racer" program may cause one MDS thread to rename
4641 a file from being the source of a rename to being the target of
4642 a rename at exactly the same time that another thread is doing
4643 so, and the second thread has already enqueued these locks after
4644 doing a lookup of the target and is trying to relock them in
4645 order. Ensure that we don't try to re-lock the same resource.
4648 Frequency : only very large systems with liblustre clients
4650 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4651 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4652 all OSTs in parallel. This allows the eviction of stale liblustre
4653 clients to proceed much faster than if they were done in series,
4654 and also offers similar improvements for other set_info RPCs.
4659 Description: excessive CPU usage during initial read phase on client
4660 Details : During the initial read phase on a client, it would agressively
4661 retry readahead on the file, consuming too much CPU and impacting
4662 performance (since 1.4.5.8). Improve the readahead algorithm
4663 to avoid this, and also improve some other common cases (read
4664 of small files in particular, where "small" is files smaller than
4665 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4670 Description: MDS crash when receiving packet with unknown intent.
4671 Details : Do not LBUG in unknown intent case, just return -EFAULT
4673 Severity : enhancement
4674 Bugzilla : 9293, 9385
4675 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4676 Details : Do not serialize getattr (non-intent version) and statfs.
4679 Frequency : occasional, when OST network is overloaded/intermittent
4681 Description: client evicted by OST after bulk IO timeout
4682 Details : If a client sends a bulk IO request (read or write) the OST
4683 may evict the client if it is unresposive to its data GET/PUT
4684 request. This is incorrect if the network is overloaded (takes
4685 too long to transfer the RPC data) or dropped the OST GET/PUT
4686 request. There is no need to evict the client at all, since
4687 the pinger and/or lock callbacks will handle this, and the
4688 client can restart the bulk request.
4691 Frequency : Always when mmapping file with no objects
4693 Description: client crashes when mmapping file with no objects
4694 Details : Check that we actually have objects in a file before doing any
4695 operations on objects in ll_vm_open, ll_vm_close and
4701 Description: Request leak when working with deleted CWD
4702 Details : Introduce advanced request refcount tracking for requests
4703 referenced from lustre intent.
4705 Severity : Enhancement
4707 Description: Cache open file handles on client.
4708 Details : MDS now will return special lock along with openhandle, if
4709 requested and client is allowed to hold openhandle, even if unused,
4710 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4711 closing files for every read/write openration.
4713 Severity : Enhancement
4715 Description: Cache open negative dentries on client when possible.
4716 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4717 negative dentries on lock revocation.
4722 Description: Remounting a client read-only wasn't possible with a zconf mount
4723 Details : It wasn't possible to remount a client read-only with llmount.
4725 Severity : enhancement
4726 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4727 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4728 MPICH is included to add Lustre-specific ADIO interfaces.
4729 This is based closely on the UFS ADIO layer and only differs
4730 in file creation, in order to allow the OST striping to be set.
4731 This is user-contributed code and not supported by CFS.
4736 Description: extended inode attributes (immutable, append-only) work improperly
4737 when 2.4 and 2.6 kernels are used on client/server or vice versa
4738 Details : Introduce kernel-independent values for these flags.
4740 Severity : enhancement
4743 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4744 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4745 granularity. Now, Fractional MB granularity is be supported,
4746 this is very useful for low memory system.
4748 Severity : enhancement
4750 Description: Getattr by fid
4751 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4752 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4753 have locks and avoids getattr rpc requests for it on every stat.
4756 Frequency : Always, for filesystems larger than 2TB
4758 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4759 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4760 filesystems larger than 2TB to be mounted with mballoc enabled.
4763 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4765 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4766 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4767 filesystems larger than 2TB is mounted then the kernel will
4768 silently corrupt the start of the filesystem. CONFIG_LBD is
4769 enabled for all CFS-supported kernels, but the possibility of
4770 this happening with a modified kernel config exists.
4772 Severity : enhancement
4774 Description: add client O_DIRECT support for 2.6 kernels
4775 Details : It is now possible to do O_DIRECT reads and writes to files
4776 in the Lustre client mountpoint on 2.6 kernel clients.
4778 Severity : enhancement
4780 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4781 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4782 parallel, not waiting for response from every OST before sending
4783 a rpc to the next OST.
4788 Description: setattr vs write race when updating file timestamps
4789 Details : Client processes that update a file timestamp into the past
4790 right after writing to the file (e.g. tar) it is possible that
4791 the updated file modification time can be reset to the current
4792 time due to a race between processing the setattr and write RPC.
4794 Severity : enhancement
4796 Description: Bring 'lfs find' closer in line with regular Linux find.
4797 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4798 -print0 options and obtains all the needed info through the lustre
4801 Severity : enhancement
4803 Description: support up to 1024 configured devices on one node
4804 Details : change obd_dev array from statically allocated to dynamically
4805 allocated structs as they are first used to reduce memory usage
4810 Description: Flush dirty partially truncated pages during truncate
4811 Details : Immediatelly flush partially truncated pages in filter_setattr,
4812 this way we completely avoid having any pages in page cache on OST
4813 and can retire ugly workarounds during writes to flush such pages.
4818 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4819 Details : protect inode from truncation within vfs_unlink() context
4820 just take a reference before calling vfs_unlink() and release it
4821 when parent's i_sem is free.
4826 Description: last_id value checked outside lock on OST caused LASSERT failure
4827 Details : If there were multiple MDS->OST object precreate requests in
4828 flight, it was possible that the OST's last object id was checked
4829 outside a lock and incorrectly tripped an assertion. Move checks
4830 inside locks, and discard old precreate requests.
4833 Frequency : always, if extents are used on OSTs
4835 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4836 Details : OSTs using the extents format would not zero the high 16 bits of
4837 the index physical block number. This is not a problem for any
4838 OST filesystems smaller than 16TB, and no kernels support ext3
4839 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4840 new/modified files) and can be fixed for existing filesystems
4841 with e2fsprogs-1.39-cfs1.
4846 Description: import connection selection may be incorrect if timer wraps
4847 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4848 import connections to be ignored if the 32-bit jiffies counter
4849 wraps. Use a 64-bit jiffies counter.
4852 Frequency : very large clusters immediately after boot
4854 Description: LNET request buffers exhausted under heavy short-term load
4855 Details : If a large number of client requests are generated on a service
4856 that has previously never seen so many requests it is possible
4857 that the request buffer growth cannot keep up with the spike in
4858 demand. Instead of dropping incoming requests, they are held in
4859 the LND until the RPC service can accept more requests.
4862 Frequency : Sometimes during replay
4864 Description: Assertion failure in ll_local_open after replay.
4865 Details : If replay happened on an open request reply before we were able
4866 to set replay handler, reply will become not swabbed tripping the
4867 assertion in ll_local_open. Now we set the handler right after
4868 recognising of open request
4871 Frequency : very rare
4873 Description: kernel reports "badness in vsnprintf"
4874 Details : Reading from the "recovery_status" /proc file in small chunks
4875 may cause a negative length in lprocfs_obd_rd_recovery_status()
4876 call to vsnprintf() (which is otherwise harmless). Exit early
4877 if there is no more space in the output buffer.
4879 Severity : enhancement
4881 Description: clear OBD RPC statistics by writing to them
4882 Details : It is now possible to clear the OBD RPC statistics by writing
4883 to the "stats" file.
4888 Description: Client mtime is not the same on different clients after utimes
4889 Details : In some cases, the client was using the utimes() syscall on
4890 a file cached on another node. The clients now validate the
4891 ctime from the MDS + OSTs to determine which one is right.
4896 Description: Inability to activate failout mode
4897 Details : lconf script incorrectly assumed that in pythong string's numeric
4898 value is used in comparisons.
4901 Frequency : always with multiple stripes per file
4903 Description: Inefficient object allocation for mutli-stripe files
4904 Details : When selecting which OSTs to stripe files over, for files with
4905 a stripe count that divides evenly into the number of OSTs,
4906 the MDS is always picking the same starting OST for each file.
4907 Return the OST selection heuristic to the original design.
4912 Description: mount failures may take full timeout to return an error
4913 Details : Under some heavy load conditions it is possible that a
4914 failed mount can wait for the full obd_timeout interval,
4915 possibly several minutes, before reporting an error.
4916 Instead return an error as soon as the status is known.
4918 Frequency : quota enabled and large files being deleted
4920 Description: releasing more than 4GB of quota at once hangs OST
4921 Details : If a user deletes more than 4GB of files on a single OST it
4922 will cause the OST to spin in an infinite loop. Release
4923 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4928 Description: statfs data retrieved from /proc may be stale or zero
4929 Details : When reading per-device statfs data from /proc, in the
4930 {kbytes,files}_{total,free,avail} files, it may appear
4931 as zero or be out of date.
4934 Frequency : systems with MD RAID1 external journal devices
4936 Description: lconf's call to blkid is confused by RAID1 journal devices
4937 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4938 of returning all block devices that match the journal UUID.
4941 Frequency : always, for aggregate stripe size over 4GB
4943 Description: assertion fails when trying to use 4GB stripe size
4944 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4945 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4948 Frequency : always on ppc64
4950 Description: the first write on an ext3 filesystem with mballoc got stuck
4951 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4952 perform endianness conversion.
4954 ------------------------------------------------------------------------------
4956 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4958 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4959 INTEROPERATE with older versions automatically. Please read the
4960 user documentation before upgrading any part of a live system.
4961 * WARNING: Lustre networking configuration changes are required with
4962 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4965 * Support for kernels:
4966 2.6.9-22.0.2.EL (RHEL 4)
4967 2.6.5-7.244 (SLES 9)
4968 2.6.12.6 vanilla (kernel.org)
4971 Severity : enhancement
4972 Bugzilla : 7981/8208
4973 Description: Introduced Lustre Networking (LNET)
4974 Details : LNET is new networking infrastructure for Lustre, it includes
4975 a reorganized network configuration mode (see the user
4976 documentation for full details) as well as support for routing
4977 between different network fabrics. Lustre Networking Devices
4978 (LNDs) for the supported network fabrics have also been
4979 created for this new infrastructure.
4981 Severity : enhancement
4982 Description: Introduced Access control lists
4983 Details : clients can set ACLs on files and directories in order to have
4984 more fine-grained permissions than the standard Unix UGO+RWX.
4985 The MDS must be started with the "-o acl" mount option.
4987 Severity : enhancement
4988 Description: Introduced filesystem quotas
4989 Details : Administrators may now establish per-user quotas on the
4992 Severity : enhancement
4994 Description: Configuration change for the XT3
4995 The PTLLND is now used to run Lustre over Portals on the XT3
4996 The configure option(s) --with-cray-portals are no longer used.
4997 Rather --with-portals=<path-to-portals-includes> is used to
4998 enable building on the XT3. In addition to enable XT3 specific
4999 features the option --enable-cray-xt3 must be used.
5004 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
5005 Details : A race between allocating a new llcd and re-getting the llcd_lock
5006 allowed another thread to grab newly-allocated llcd.
5008 Severity : enhancement
5010 Description: 2.6 OST async journal commit and locking fix to improve performance
5011 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
5012 2.6 kernels are now async as they already were in 2.4 kernels so
5013 that they can commit concurrently with the network bulk transfer.
5014 For block-allocated files the filter allocation semaphore is held
5015 to avoid filesystem fragmentation during allocation. BKL lock
5016 removed for 2.6 xattr operations where it is no longer needed.
5021 Description: lconf incorrectly determined whether two IP networks could talk
5022 Details : In some more complicated routing and multiple-network
5023 configurations, lconf will avoid trying to make a network
5024 connection to a disjoint part of the IP space. It was doing the
5025 math incorrectly for one set of cases.
5030 Description: Fix for potential infinite loop processing records in an llog.
5031 Details : If an llog record is corrupted/zeroed, it is possible to loop
5032 forever in llog_process(). Validate the llog record length
5033 and skip the remainder of the block on error.
5036 Frequency : occasional (liblustre only)
5038 Description: liblustre could not open files whose last component is a symlink
5039 Details : sysio_path_walk() would incorrectly pass the open intent to
5040 intermediate path components.
5043 Frequency : rare (liblustre only with non-standard tuning)
5044 Bugzilla : 7201 (7350)
5045 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
5046 Details : llu_lookup_finish_locks() tries to set lock data on a lock
5047 after it has been released, only do this for referenced locks
5049 Severity : enhancement
5051 Description: specifying an (invalid) directory default stripe_size of -1
5052 would reset the directory default striping
5053 Details : stripe_size -1 was used internally to signal directory stripe
5054 removal, now use "all default" to signal dir stripe removal
5055 as a directory striping of "all default" is not useful
5058 Frequency : common for large clusters running liblustre clients
5060 Description: doing an ls when liblustre clients are running is slow
5061 Details : sending a glimpse AST to a liblustre client waits for every AST
5062 to time out, as liblustre clients will not respond. Since they
5063 cannot cache data we refresh the OST lock LVB from disk instead.
5065 Severity : enhancement
5067 Description: doing an ls at the same time as file IO can be slow
5068 Details : enqueue and other "small" requests can be blocked behind many
5069 large IO requests. Create a new OST IO portal for non-IO
5070 requests so they can be processed faster.
5073 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
5075 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
5076 created file with mode 000
5077 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
5078 allow an ftruncate() as a non-root user to a file with mode 000.
5079 Lustre now respects this flag to disable mode checks when
5080 truncating a file owned by the user
5083 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
5086 Description: Revoking locks from clients that went dead or catatonic might take
5088 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
5089 cancellation of such locks instant on servers without waiting for
5090 any reply from clients. Clients drops these locks when cancel
5091 notification from server is received without replying.
5094 Frequency : liblustre-only, when liblustre client dies or becomes busy
5096 Description: Doing ls on Linux clients can take a long time with active
5098 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
5099 granting such locks to it in the first place if possible. Locks
5100 are taken by proxy on the OST during the read or write and
5101 dropped immediately afterward. Add connect flags handling, do
5102 not grant locks to liblustre clients for glimpse ASTs.
5104 Severity : enhancement
5106 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
5107 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5108 sequential reads and behaves badly with applications doing random
5109 reads. Improve it by reading ahead at least read region, and
5110 avoiding excessive large RPC for small reads.
5112 Severity : enhancement
5114 Description: Creating more than 1000 files for a single job may cause a load
5115 imbalance on the OSTs if there are also a large number of OSTs.
5116 Details : qos_prep_create() uses an OST index reseed value that is an
5117 even multiple of the number of available OSTs so that if the
5118 reseed happens in the middle of the object allocation it will
5119 still utilize the OSTs as uniformly as possible.
5124 Description: OST or MDS may oops in ping_evictor_main()
5125 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5126 but doesn't restart at beginning of obd_exports_timed list
5129 Severity : enhancement
5131 Description: improve by-nid export eviction on the MDS and OST
5132 Details : allow multiple exports with the same NID to be evicted at one
5133 time without re-searching the exports list.
5136 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5138 Description: MDS may oops in groups_free()
5139 Details : in rare race conditions a newly allocated group_info struct is
5140 freed again, and this can be NULL. The 2.4 compatibility code
5141 for groups_free() checked for a NULL pointer, but 2.6 did not.
5144 Frequency : common for liblustre clients doing little filesystem IO
5145 Bugzilla : 9352, 7313
5146 Description: server may evict liblustre clients accessing contended locks
5147 Details : if a client is granted a lock or receives a completion AST
5148 with a blocking AST already set it would not reply to the AST
5149 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5152 Frequency : lfs setstripe, only systems with more than 160 OSTs
5154 Description: unable to set striping with a starting offset beyond OST 160
5155 Details : llapi_create_file() incorrectly limited the starting stripe
5156 index to the maximum single-file stripe count.
5159 Frequency : LDAP users only
5161 Description: lconf did not handle in-kernel recovery with LDAP properly
5162 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5164 Severity : enhancement
5166 Description: bind OST threads to NUMA nodes to improve performance
5167 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5168 node and do their allocations there to localize memory access
5170 Severity : enhancement
5172 Description: llmount can determine client NID directly from Myrinet (GM)
5173 Details : the client NID code from gmnalnid was moved directly into
5174 llmount, removing the need to use this or specifying the
5175 client NID explicitly when mounting GM clients with zeroconf
5178 Frequency : if client is started with down MDS
5180 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5181 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5182 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5183 ptlrpc_import_delayed_req() because we didn't check if the
5184 request was interrupted, and we also didn't break out of the
5185 event loop if there was no timeout
5190 Description: data loss during non-page-aligned writes to a single file from
5191 both multiple nodes and multiple threads on one node at same time
5192 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5193 inconsistency led to false short-reads, that were cached and later
5194 used by ->prepare_write() to fill in partially written page,
5195 leading to data loss.
5198 Frequency : always, if lconf --abort_recovery used
5200 Description: lconf --abort_recovery fails with 'Operation not supported'
5201 Details : lconf was attempting to abort recovery on the MDT device and not
5204 Severity : enhancement
5206 Description: remove cleanup logs
5207 Details : replace lconf-generated cleanup logs with lustre internal
5208 cleanup routines. Eliminates the need for client-cleanup and
5211 Severity : enhancement
5213 Description: add support for EAs (user and system) on lustre filesystems
5214 Details : it is now possible to store extended attributes in the Lustre
5215 client filesystem, and with the user_xattr mount option it
5216 is possible to allow users to store EAs on their files also
5218 Severity : enhancement
5220 Description: Add possibility (config option) to show minimal available OST free
5222 Details : When compiled with --enable-mindf configure option, statfs(2)
5223 (and so, df) will return least minimal free space available from
5224 all OSTs as amount of free space on FS, instead of summary of
5225 free spaces of all OSTs.
5227 Severity : enhancement
5229 Description: do not expand extent locks acquired on OST-side
5230 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5231 by server: they are not cached anyway.
5234 Frequency : when mmap is used/binaries executed from Lustre
5236 Description: Unmmap pages before throwing them away from read cache.
5237 Details : llap_shrink cache now attempts to unmap pages before discarding
5238 them (if unmapping failed - do not discard). SLES9 kernel has
5239 extra checks that trigger if this unmapping is not done first.
5244 Description: lconf didn't resolve symlinks before checking to see whether a
5245 given mountpoint was already in use
5248 Frequency : when migrating failover services
5249 Bugzilla : 6395, 9514
5250 Description: When migrating a subset of services from a node (e.g. failback
5251 from a failover service node) the remaining services would
5252 time out and evict clients.
5253 Details : lconf --force (implied by --failover) sets the global obd_timeout
5254 to 5 seconds in order to quickly disconnect, but this caused
5255 other RPCs to time out too quickly. Do not change the global
5256 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5258 Severity : enhancement
5259 Frequency : if MDS is started with down OST
5260 Bugzilla : 9439,5706
5261 Description: Allow startup/shutdown of an MDS without depending on the
5262 availability of the OSTs.
5263 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5264 Add appropriate locking and lov-osc refcounts for safe
5265 cleaning. Add osc abort_inflight calls in case the
5266 synchronize never started.
5269 Frequency : occasional (Cray XT3 only)
5271 Description: root not authorized to access files in CRAY_PORTALS environment
5272 Details : The client process capabilities were not honoured on the MDS in
5273 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5274 been accessed by an authorized user then root was able to access
5275 the file on the local client also. The root user capabilities
5276 are now allowed on the MDS, as this environment has secure UID.
5279 Frequency : occasional
5281 Description: ldiskfs "too long searching" message happens too often
5282 Details : A debugging message (otherwise harmless) prints too often on
5283 the OST console. This has been reduced to only happen when
5284 there are fragmentation problems on the filesystem.
5289 Description: Division by zero in statfs when all OSCs are inactive
5290 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5291 lov_statfs divides by value returned by lov_get_stripecnt().
5295 Bugzilla : 9489, 3273
5296 Description: First write from each client to each OST was only 4kB in size,
5297 to initialize client writeback cache, which caused sub-optimal
5298 RPCs and poor layout on disk for the first written file.
5299 Details : Clients now request an initial cache grant at (re)connect time
5300 and so that they can start streaming writes to the cache right
5301 away and always do full-sized RPCs if there is enough data.
5302 If the OST is rebooted the client also re-establishes its grant
5303 so that client cached writes will be honoured under the grant.
5308 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5309 Details : Now I/O RPCs go to different portal number and (presumably) fast
5310 lock requests (and glimses) and other RPCs get their own service
5311 threads pool that should be able to service those RPCs
5314 Severity : enhancement
5316 Description: Ability to exchange lustre version between client and servers and
5317 issue warnings at client side if client is too old. Also for
5318 liblustre clients there is ability to refuse connection of too old
5320 Details : New 'version' field is added to connect data structure that is
5321 filled with version info. That info is later checked by server and
5325 Frequency : rare, liblustre only.
5326 Bugzilla : 9296, 9581
5327 Description: Two simultaneous writes from liblustre at offset within same page
5328 might proceed at the same time overwriting eachother with stale
5330 Details : I/O lock withing llu_file_prwv was released too early, before data
5331 actually was hitting the wire. Extended lock-holding time until
5332 server acknowledges receiving data.
5335 Frequency : extremely rare. Never observed in practice.
5337 Description: avoid generating lustre_handle cookie of 0.
5338 Details : class_handle_hash() generates handle cookies by incrementing
5339 global counter, and can hit 0 occasionaly (this is unlikely, but
5340 not impossible, because initial value of cookie counter is
5341 selected randonly). Value of 0 is used as a sentinel meaning
5342 "unassigned handle" --- avoid it. Also coalesce two critical
5343 sections in this function into one.
5345 Severity : enhancement
5347 Description: allow liblustre clients to delegate truncate locking to OST
5348 Details : To avoid overhead of locking, liblustre client instructs OST to
5349 take extent lock in ost_punch() on client's behalf. New connection
5350 flag is added to handle backward compatibility.
5352 Severity : enhancement
5353 Bugzilla : 4928, 7341, 9758
5354 Description: allow number of OST service threads to be specified
5355 Details : a module parameter allows the number of OST service threads
5356 to be specified via "options ost ost_num_threads={N}" in the
5357 OSS's /etc/modules.conf or /etc/modprobe.conf.
5361 Bugzilla : 6146, 9635, 9895
5362 Description: servers crash with bad pointer in target_handle_connect()
5363 Details : In rare cases when a client is reconnecting it was possible that
5364 the connection request was the last reference for that export.
5365 We would temporarily drop the export reference and get a new
5366 one, but this may have been the last reference and the export
5367 was just destroyed. Get new reference before dropping old one.
5369 Severity : enhancement
5370 Frequency : if client is started with failover MDS
5372 Description: Allow multiple MDS hostnames in the mount command
5373 Details : Try to read the configuration from all specified MDS
5374 hostnames during a client mount in case the "primary"
5377 Severity : enhancement
5379 Description: Stop sending data to evicted clients as soon as possible.
5380 Details : Check if the client we are about to send or are sending data to
5381 was evicted already. (Check is done every second of waiting,
5382 for which l_wait_event interface was extended to allow checking
5383 of exit condition at specified intervals).
5386 Frequency : rare, normally only when NFS exporting is done from client
5388 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5390 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5393 Frequency : always, if srand() is called before liblustre initialization
5395 Description: Liblustre uses system PRNG disturbing its usage by user application
5396 Details : Introduce internal to lustre fast and high-quality PRNG for
5397 lustre usage and make liblustre and some other places in generic
5398 lustre code to use it.
5400 Severity : enhancement
5401 Bugzilla : 9477, 9557, 9870
5402 Description: Verify that the MDS configuration logs are updated when xml is
5403 Details : Check if the .xml configuration logs are newer than the config
5404 logs stored on the MDS and report an error if this is the case.
5405 Request --write-conf, or allow starting with --old_conf.
5407 Severity : enhancement
5409 Description: Handle symlinks in the path when checking if Lustre is mounted.
5410 Details : Resolve intermediate symlinks when checking if a client has
5411 mounted a filesystem to avoid duplicate client mounts.
5416 Description: lconf can hit an error exception but still return success.
5417 Details : The lconf command catches the Command error exception at the top
5418 level script context and will exit with the associated exit
5419 status, but doesn't ensure that this exit status is non-zero.
5424 Description: failure of ptlrpc thread startup can cause oops
5425 Details : Starting a ptlrpc service thread can fail if there are a large
5426 number of threads or the server memory is very fragmented.
5427 Handle this without oopsing.
5430 Frequency : always, only if liblustre and non-default acceptor port was used
5432 Description: liblustre cannot connect to servers with non-default acceptor port
5433 Details : tcpnal_set_default_params() was not called and was therefore
5434 ignoring the environment varaible TCPNAL_PORT, as well as other
5435 TCPNAL_ environment variables
5440 Description: two objects could be created on the same OST for a single file
5441 Details : If an OST is down, in some cases it was possible to create two
5442 objects on a single OST for a single file. No problems other
5443 than potential performance impact and spurious error messages.
5447 Bugzilla : 5681, 9562
5448 Description: Client may oops in ll_unhash_aliases
5449 Details : Client dcache may become inconsistent in race condition.
5450 In some cases "getcwd" can fail if the current directory is
5456 Description: Inode refcounting problems in NFS export code
5457 Details : link_raw functions used to call d_instantiate without obtaining
5458 extra inode reference first.
5462 Bugzilla : 9942, 9903
5463 Description: Referencing freed requests leading to crash, memleaks with NFS.
5464 Details : We used to require that call to ll_revalidate_it was always
5465 followed by ll_lookup_it. Also with revalidate_special() it is
5466 possible to call ll_revalidate_it() twice for the same dentry
5467 even if first occurence returned success. This fix changes semantic
5468 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5469 reference on a request referred from the intent.
5470 ll_intent_release() then releases such a request.
5473 Frequency : rare, normally benchmark loads only
5475 Description: unlinked inodes were kept in memory on the client
5476 Details : If a client is repeatedly creating and unlinking files it
5477 can accumulate a lot of stale inodes in the inode slab cache.
5478 If there is no other client load running this can cause the
5479 client node to run out of memory. Instead flush old inodes
5480 from client cache that have the same inode number as a new inode.
5483 Frequency : SLES9 2.6.5 kernel and long filenames only
5484 Bugzilla : 9969, 10379
5485 Description: utime reports stale NFS file handle
5486 Details : SLES9 uses out-of-dentry names in some cases, which confused
5487 the lustre dentry revalidation. Change it to always use the
5491 Frequency : rare, unless heavy write-truncate concurrency is continuous
5492 Bugzilla : 4180, 6984, 7171, 9963, 9331
5493 Description: OST becomes very slow and/or deadlocked during object unlink
5494 Details : filter_destroy() was holding onto the parent directory lock
5495 while truncating+unlinking objects. For very large objects this
5496 may block other threads for a long time and slow overall OST
5497 responsiveness. It may also be possible to get a lock ordering
5498 deadlock in this case, or run out of journal credits because of
5499 the combined truncate+unlink. Solution is to do object truncate
5500 first in one transaction without parent lock, and then do the
5501 final unlink in a new transaction with the parent lock. This
5502 reduces the lock hold time dramatically.
5505 Frequency : rare, 2.4 kernels only
5507 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5508 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5509 before unmounting their filesystem in order to avoid deadlock.
5510 The kernel_locked() function in 2.4 kernels only checks whether
5511 the kernel lock is held, not whether it is this process that is
5512 holding it as 2.6 kernels do.
5517 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5518 Details : The client ptlrpc code may be trying to reconnect to a down
5519 server before a previous connection attempt has timed out.
5520 Increase the reconnect interval to be longer than the connection
5521 timeout interval to avoid sending duplicate connections to
5525 Frequency : echo_client brw_test command
5527 Description: fix echo_client to work with OST preallocated code
5528 Details : OST preallocation code (5137) didn't take echo_client IO path
5529 into account: echo_client calls filter methods outside of any
5530 OST thread and, hence, there is no per-thread preallocated
5531 pages and buffers to use. Solution: hijack pga pages for IO. As
5532 a byproduct, this avoids unnecessary data copying.
5536 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5537 Description: Client can oops in mdc_commit_close() after open replay
5538 Details : It was possible for the MDS to return an open request with no
5539 transaction number in mds_finish_transno() if the client was
5540 evicted, but without actually returning an error. Clients
5541 would later try to replay that open and may trip an assertion
5542 Simplify the client close codepath, and always return an error
5543 from the MDS in case the open is not successful.
5546 Frequency : rare, 2.6 OSTs only
5548 Description: OST may deadlock under high load on fragmented files
5549 Details : If there was a heavy load and highly-fragmented OST filesystems
5550 it was possible to have all the OST threads deadlock waiting on
5551 allocation of biovecs, because the biovecs were not released
5552 until the entire RPC IO was completed. Instead, release biovecs
5553 as soon as they are complete to ensure forward IO progress.
5555 Severity : enhancement
5557 Description: Support for specifying external journal device at mount
5558 Details : If an OST or MDS device is formatted with an external journal
5559 device, this device major/minor is stored in the ext3 superblock
5560 and may not be valid for failover. Allow detecting and
5561 specifying the external journal at mount time.
5566 Description: Mounting an MDS with pending unlinked files may cause oops
5567 Details : target_finish_recovery() calls mds_postrecov() which returned
5568 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5569 considers this an error and immediately begins cleaning up the
5570 lov, just after starting the mds_lov process
5572 Severity : enhancement
5574 Description: Implement 'lfs df' to report actual free space on per-OST basis
5575 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5576 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5577 usage of objects; '-h' to report in human readable format.
5579 ------------------------------------------------------------------------------
5581 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5588 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5589 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5590 at startup during orphan unlinking. Instead, skip buddy update
5591 before setup, as it will be regenerated after recovery is complete.
5596 Description: If an OST is inactive, its locks might reference stale inodes.
5597 Details : lov_change_cbdata() must iterate over all namespaces, even if
5598 they are inactive to clear inode references from the lock.
5600 Severity : enhancement
5601 Frequency : occasional, if non-standard max_dirty_mb used
5603 Description: Client will block write RPCs if not enough grant
5604 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5605 then the client will block writes while waiting for another RPC
5606 to complete instead of consuming its dirty limit. With change
5607 we get improved performance when max_dirty_mb is small.
5609 Severity : enhancement
5610 Bugzilla : 3389, 6253
5611 Description: Add support for supplementary groups on the MDS.
5612 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5613 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5614 lookups for user supplementary groups into a cache.
5618 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5619 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5620 for liblustre O_EXCL is enforced by MDS.
5624 Description: readdir never returns NULL in liblustre.
5625 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5626 dentry in struct dirent.
5630 Frequency : liblustre only, depends on application IO pattern
5631 Description: liblustre clients evicted if not contacting servers
5632 Details : Don't put liblustre clients into the ping_evictor list, so
5633 they will not be evicted by the pinger ever.
5635 Severity : enhancement
5637 Description: Add ability to evict clients by NID from MDS.
5638 Details : By echoing "nid:$NID" string into
5639 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5640 $NID would be instantly evicted from this MDS and from all active
5641 OSTs connected to it.
5645 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5646 Details : lookup_it_finish() used to query file size from OSTs that was not
5651 Description: service threads change working directory to that of init
5652 Details : Starting lustre service threads may pin the working directory
5653 of the parent thread, making that filesystem busy. Threads
5654 now change to the working directory of init to avoid this.
5658 Frequency : during shutdown only
5659 Description: shutdown with a failed MDS or OST can cause unmount to hang
5660 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5665 Frequency : 2.6 only
5666 Description: chmod/chown may include an extra supplementary group
5667 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5668 supplementary group and if none is specified this is used.
5671 Bugzilla : 5479 (6816)
5673 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5674 Details : If lookup is in progress on a file that is unlinked we might try
5675 to revalidate the inode and fail in revalidate after lookup is
5676 complete and ll_file_open() enqueues the open again but
5677 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5680 Frequency : always, if lconf --abort_recovery used
5682 Description: lconf --abort_recovery fails with 'Operation not supported'
5683 Details : lconf was attempting to abort recovery on the MDT device and not
5686 ------------------------------------------------------------------------------
5688 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5693 Frequency : rare (only unsupported configurations with a node running as an
5695 Bugzilla : 6514, 5137
5696 Description: Mounting a Lustre file system on a node running as an OST could
5698 Details : OSTs now preallocates memory needed to write out data at
5699 startup, instead of when needed, to avoid having to
5700 allocate memory in possibly low memory situations.
5701 Specifically, if the file system is mounted on on OST,
5702 memory pressure could force it to try to write out data,
5703 which it needed to allocate memory to do. Due to the low
5704 memory, it would be unable to do so and the node would
5705 become unresponsive.
5707 Severity : enhancement
5709 Description: Addition of lconf --service command line option
5710 Details : lconf now accepts a '--service <arg>' option, which is
5711 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5713 Severity : enhancement
5715 Description: Failover mode is now the default for OSTs.
5716 Details : By default, OSTs will now run in failover mode. To return to
5717 the old behaviour, add '--failout' to the lmc line for OSTs.
5719 Severity : enhancement
5721 Description: Health checks are now provided for MDS and OSTs
5722 Details : Additional detailed health check information on MSD and OSTs
5723 is now provided through the procfs health_check value.
5726 Frequency : occasional, depends on IO load
5728 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5729 after numerous create/delete cycles
5730 Details : The ext3 inode allocation policy would not allocate new inodes
5731 very well on the OSTs because there are no new directories
5732 being created. Instead we look for groups with free space if
5733 the parent directories are nearly full.
5738 Description: Network or server problems during mount may cause partially
5739 mounted clients instead of returning an error.
5740 Details : The config llog parsing code may overwrite the error return
5741 code during mount error handling, returning success instead
5747 Description: MDS can fail to allocate large reply buffers
5748 Details : After long uptimes the MDS can fail to allocate large reply
5749 buffers (e.g. zconf client mount config records) due to memory
5750 fragmentation or consumption by the buffer cache. Preallocate
5751 some large reply buffers so that these replies can be sent even
5752 under memory pressure.
5756 Frequency : rare (liblustre)
5757 Description: fsx running with liblustre complained that using truncate() to
5758 extend the file doesn't work. This patch corrects that issue.
5759 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5760 ATTR_SIZE and lsm use in llu_setattr_raw.
5764 Frequency : rare, only 2.6 kernels
5765 Description: Unusual file access patterns on the MDS may result in inode
5766 data being lost in very rare circumstances.
5767 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5768 code in the 2.6 kernel caused the inode and/or EA data not to
5769 be read from disk, causing single-file corruption.
5773 Frequency : rare, only 2.6 filesystems using extents
5774 Description: Heavy concurrent write and delete load may cause data corruption.
5775 Details : It was possible under high-load situations to have an extent
5776 metadata block in the block device cache from a just-unlinked
5777 file overwrite a newly-allocated data block. We now unmap any
5778 metadata buffers that alias just-allocated data blocks.
5782 Frequency : filesystems with default stripe_count larger than 77
5783 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5784 Details : lconf specifies an inode size of 4096 bytes when the default
5785 stripe_count is larger than 77. This conflicts with the default
5786 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5787 this case to avoid pinning too much memory for large EAs.
5789 ------------------------------------------------------------------------------
5791 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5796 Frequency : rare (extremely heavy IO load with hundreds of clients)
5798 Description: Client is evicted, gets IO error writing to file
5799 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5800 caused clients to be evicted for AST timeouts. The fixes in
5801 bug 5192 mean we no longer need to have such short AST timeouts
5802 so ldlm_timeout has been increased.
5805 Frequency : occasional during --force or --failover shutdown under load
5806 Bugzilla : 5949, 4834
5807 Description: Server oops/LBUG if stopped with --force or --failover under load
5808 Details : a collection of import/export refcount and cleanup ordering
5809 issues fixed for safer force cleanup
5812 Frequency : only filesystems larger than 120 OSTs
5813 Bugzilla : 5990, 6223
5814 Description: lfs getstripe would oops on a very large filesystem
5815 Details : lov_getconfig used kfree on vmalloc'd memory
5818 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5819 Bugzilla : 6242, 6243
5820 Description: reading from files that had been truncated to a non-zero size
5821 but never opened returned no data
5822 Details : ll_file_read() reads zeros from no-object files to EOF
5827 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5828 Details : There's little harm in aborting MDS/OSS recovery and letting it
5829 try again, so I removed the LASSERT and return an error instead.
5831 Severity : enhancement
5833 Description: New debugging infrastructure for tracking down data corruption
5834 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5835 (b) cover more of the client-side code path, and (c) try to narrow
5836 down where problems occurred
5840 Bugzilla : 3819, 4364, 4397, 6313
5841 Description: Racing close and eviction MDS could cause assertion in mds_close
5842 Details : It was possible to get multiple mfd references during close and
5843 client eviction, leading to one thread referencing a freed mfd.
5845 Severity: : enhancement
5846 Bugzilla : 3262, 6359
5847 Description: Attempts to reconnect to servers are now more aggressive.
5848 Details : This builds on the enhanced upcall-less recovery that was added
5849 in 1.4.2. When trying to reconnect to servers, clients will
5850 now try each server in the failover group every 10 seconds. By
5851 default, clients would previously try one server every 25 seconds.
5856 Description: After recovery, certain operations trigger a failed
5857 assertion on a client.
5858 Details : Failing over an mds, using lconf -d --failover, while a
5859 client was doing a readdir() call would cause the client to
5860 LBUG after recovery completed and the readdir() was resent.
5862 Severity : enhancement
5864 Description: Default groups are now added by lconf
5865 Details : You can now run lconf --group <servicename> without having to
5866 manually add groups with lmc.
5869 Frequency : occasional
5871 Description: Nodes with an elan id of 0 trigger a failed assertion
5874 Frequency : always when accessing e.g. tty/console device nodes
5876 Description: tty and some other devices nodes cannot be used on lustre
5877 Details : file's private_data field is used by device data and lustre
5878 values in there got lost. New field was added to struct file to
5879 store fs-specific private data.
5882 Frequency : when exporting Lustre via NFS
5884 Description: NFSD failed occasionally when looking up a path component
5885 Details : NFSD is looking up ".." which was broken in ext3 directories
5886 that had grown large enough to become hashed.
5889 Frequency : Clusters with multiple interfaces not on the same subnet
5891 Description: Nodes will repeatedly try to reconnect to an interface which it
5892 cannot reach and report an error to the log.
5893 Details : Extra peer list entries will be created by lconf with some peers
5894 unreachable. lconf now validates the peer before adding it.
5897 Frequency : Only if a default stripe is set on the filesystem root.
5899 Description: Setting a default stripe on the filesystem root prevented the
5900 filesystem from being remounted.
5901 Details : The client was sending extra request flags in the root getattr
5902 request and did not allocate a reply buffer for the dir EA.
5905 Frequency : occasional, higher if lots of files are accessed by one client
5906 Bugzilla : 6159, 6097
5907 Description: Client trips assertion regarding lsm mismatch/magic
5908 Details : While revalidating inodes the VFS looks up inodes with ifind()
5909 and in rare cases can find an inode that is being freed.
5910 The ll_test_inode() code will free the lsm during ifind()
5911 when it finds an existing inode and then the VFS later attaches
5912 this free lsm to a new inode.
5916 Bugzilla : 6422, 7030
5917 Description: MDS deadlock between mkdir and client eviction
5918 Details : Creating a new file via mkdir or mknod (starting a transaction
5919 and getting the ns lock) can deadlock with client eviction
5920 (gets ns lock and trying to finish a synchronous transaction).
5923 Frequency : occasional
5924 Description: While starting a server, the fsfilt_ext3 module could not be
5926 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5927 kernels. Previously, lconf would still use the ext3 name
5928 when trying to load modules. Now, it will correctly use
5929 ext3 on 2.4 and ldiskfs on 2.6.
5931 Severity : enhancement
5932 Description: The default stripe count has been changed to 1
5933 Details : The interpretation of the default stripe count (0, to lfs
5934 or lmc) has been changed to mean striping across a single
5935 OST, rather than all available. For general usage we have
5936 found a stripe count of 1 or 2 works best.
5938 Severity : enhancement
5939 Description: Add support for compiling against Cray portals.
5940 Details : Conditional compiling for some areas that are different
5944 Frequency : occasional
5945 Bugzilla : 6409, 6834
5946 Description: Creating files with an explicit stripe count may lead to
5947 a failed assertion on the MDS
5948 Details : If some OSTs are full or unavailable, creating files may
5949 trigger a failed assertion on the MDS. Now, Lustre will
5950 try to use other servers or return an error to the
5954 Frequency : occasional
5956 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5957 caused lock timeout during readahead (since 1.4.2).
5958 Details : Processes doing readahead might match a lock that hasn't been
5959 granted yet if there are overlapping and conflicting lock
5960 requests. The readahead process waits on ungranted lock
5961 (original lock is CBPENDING), while OST waits for that process
5962 to cancel CBPENDING read lock and eventually evicts client.
5964 Severity : enhancement
5966 Description: Initial enabling of flock support for clients
5967 Details : Implements fcntl advisory locking and file status functions.
5968 This feature is provided as an optional mount flag (default
5969 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5970 locking are implemented yet, and those that are are not guaranteed
5971 to be completely correct in production environments.
5972 mount -t lustre -o [flock|noflock] ...
5975 Frequency : occasional
5977 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5978 assertion in the ext3 JBD (journaling) layer.
5979 Details : The b_committed_data struct is protected by the big kernel lock
5980 in 2.4 kernels, serializing journal_commit_transaction() and
5981 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5982 access to this struct. In 2.6 kernels there is finer grained
5983 locking to improve SMP performance of the JBD layer.
5987 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5992 Description: merge and cleanup kernel patches.
5995 -----------------------------------------------------------------------------
5997 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5999 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
6000 versions of Lustre. Please update all servers and clients to
6001 version 1.4.2 or later at the same time. You must also run
6002 "lconf --write-conf {config}.xml" on the MDS while it is stopped
6003 to update the configuration logs.
6005 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
6006 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
6007 - swab llog records of type '0' so we get proper header size/idx (5861)
6008 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
6009 - fix rename of one directory over another leaking an inode (5953)
6010 - avoid SetPageDirty on 2.6 (5981)
6011 - don't re-add just-being-destroyed locks to the waiting list (5653)
6012 - when creating new directories, inherit the parent's custom
6013 striping settings if present parent (3048)
6014 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
6015 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
6016 - don't hold client locks on temporary worklist from l_lru (5666)
6017 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
6018 - automatically evict dead clients (5921)
6019 - Update file size properly in create+truncate+fstat case (6196)
6020 - Do not unhash mountpoint dentries, do not allow removal of
6022 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
6023 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
6024 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
6025 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
6026 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
6027 - Correctly handle reads of files with no objects (6243)
6028 - lctl recover will also mark a device active if deactivate used (5933)
6030 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
6031 - allow --write-conf on an MDS with different nettype than client (5619)
6032 - don't write config llogs to MDS for mounts not from that MDS (5617)
6033 - lconf should create multiple TCP connections from a client (5201)
6034 - init scripts are now turned off by default; run chkconfig --on
6035 lustre and chkconfig --on lustrefs to use them
6036 - upcalls are no longer needed for clients to recover to failover
6038 - add --abort-recovery option to lconf to abort recovery on device
6040 - add support for an arbitrary number of OSTs (3026)
6041 - Quota support protocol changes.
6042 - forward compatibility changes to wire structs (6007)
6043 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
6044 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
6045 - improved "lustre status" script
6046 - initialize blocksize for non-regular files (6062)
6047 - added --disable-server and --disable-client configure options (5782)
6048 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
6049 - Vanilla 2.4.29 support
6050 - increase maximum number of obd devices to 520 (6242)
6051 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
6052 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
6053 SLES 9 2.6.5 kernels
6054 - limit stripes per file to 160 (the maximum EA size) (6093)
6056 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6059 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
6060 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
6061 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
6062 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
6063 - small ext3 extents cleanups and fixes (5733)
6064 - improved mballoc code, several small races and bugs fixed (5733, 5638)
6065 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
6066 - avoid needless client->OST connect, fix handle mismatch (5317)
6067 - fix DLM error path that led to out-of-sync client, long delays (5779)
6068 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
6069 - fix several locking issues related to i_size (5492,5624,5654,5672)
6070 - don't move pending lock onto export if it is already evicted (5683)
6071 - fix kernel oops when creating .foo in unlinked directory (5548)
6072 - fix deadlock in obdfilter statistics vs. object create (5811)
6073 - use time_{before,after} to avoid timer jiffies wrap (5882)
6074 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
6075 - Do not leak request if server was not able to process it (5154)
6076 - If mds_open unable to find parent dir, make that negative lookup(5154)
6077 - don't create new directories with extent-mapping (5909, 5936)
6079 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
6080 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
6081 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
6082 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
6083 - support for RHEL4 kernel on i686 (5773)
6084 - provide error messages when incompatible logs are encountered (5898)
6086 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
6087 * version 1.4.0.10 (1.4.1 release candidate 1)
6089 - don't keep a lock reference when lock is not granted (4238)
6090 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6091 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
6092 - drop import inflight refcount on signal_completed_replay error (5255)
6093 - unlock page after async write error during send (3677)
6094 - handle missing objects in filter_preprw_read properly (5265)
6095 - no transno return for symlink open, don't save no-trasno open (3440)
6096 - don't try to complete elan receive that already failed (4012)
6097 - free RPC server reply state on error (5406)
6098 - clean up thread from ptlrpc_start_thread() on error (5160)
6099 - readahead could read extra page into cache that wasn't ejected (5388)
6100 - prevent races in class_attach/setup/cleanup/detach (5260)
6101 - don't dereference de->d_inode after l_dput of de (5458)
6102 - use "int" for stripe value returned from lock_to_stripe (5544)
6103 - mballoc allocation and error-checking fixes in 2.6 (5504)
6104 - block device patches to fix I/O request sizes in 2.6 (5482)
6105 - look up hostnames for IB nals (5602)
6106 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
6107 - don't start multiple acceptors for the same port (5277)
6108 - fix incorrect LASSERT in mds_getattr_name (5635)
6109 - export a proc file for general "ping" checking (5628)
6110 - fix "lfs check" to not block when the MDS is down (5628)
6112 - service request history (4965)
6113 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6114 - create an "evict_client" /proc entry on OSTs, like the MDS has
6115 - fix mount usage message, return errors per mount(8) (5168)
6116 - change grep [] to grep "[]" in tests so they work in more UMLs
6117 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6118 - remove ancient LOV_MAGIC_V0 EA support (5047)
6119 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6120 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6121 - allow readahead allocations to fail when low on memory (5383)
6122 - mmap locking landed again, after considerable improvement (2828)
6123 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6125 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6128 - send OST transaction number in read/write reply to free req (4966)
6129 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6130 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6131 - the watchdog thread now runs as interruptible (5246)
6132 - flock/lockf fixes (but it's still disabled, pending 5135)
6133 - don't use EXT3 constants in llite code (5094)
6134 - memory shortage at startup could cause assertion (5176)
6136 - reorganization of lov code
6137 - single portals codebase
6139 - add extents/mballoc support (5025)
6140 - direct I/O reads in the obdfilter (4048)
6141 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6143 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6146 - send OST transaction number in read/write reply to free req (4966)
6147 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6148 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6149 - the watchdog thread now runs as interruptible (5246)
6150 - handle missing objects in filter_preprw_read properly (5265)
6151 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6152 - drop import inflight refcount on signal_completed_replay error (5255)
6153 - unlock page after async write error during send (3677)
6154 - return original error code on reconstructed replies (3761)
6155 - no transno return for symlink open, don't save no-trasno open (3440)
6157 - add pid to ldlm debugging output (4922)
6158 - bump the watchdog timeouts -- we can't handle 30sec yet
6159 - extra debugging for orphan dentry/inode bug (5259)
6161 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6164 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6165 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6166 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6167 - fix NULL dereference in /proc/sys/portals/routes (4827)
6168 - allow failed mdc_close() operations to be interrupted (4561)
6169 - stop precreate on OST before MDS would time out on it (4778)
6170 - don't send partial-page writes before EOF from client (4410)
6171 - discard client grant for sub-page writes on large-page clients (4520)
6172 - don't free dentries not owned by NFS code, check generation (4806)
6173 - fix lsm leak if mds_create_objects() fails (4801)
6174 - limit debug_daemon file size, always print CERROR messages (4789)
6175 - use transno after validating reply (3892)
6176 - process timed out requests if import state changes (3754)
6177 - update mtime on OST during writes, return in glimpse (4829)
6178 - add mkfsoptions to LDAP (4679)
6179 - use ->max_readahead method instead of zapping global ra (5039)
6180 - don't interrupt __l_wait_event() during strace
6182 - add software watchdogs to catch hung threads quickly (4941)
6183 - make lustrefs init script start after nfs is mounted
6184 - fix CWARN/ERROR duplication (4930)
6185 - return async write errors to application if possible (2248)
6186 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6187 - print NAL number in %x format (4645)
6188 - update barely-supported suse-2.4.21-171 series (4842)
6189 - support for sles 9 %post scripts
6190 - support for building 2.6 kernel-source packages
6191 - support for sles km_* packages
6193 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6196 - ignore -ENOENT errors in osc_destroy (3639)
6197 - notify osc create thread that OSC is being cleaned up (4600)
6198 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6199 - reconstruct ost_handle() like mds_handle() (4657)
6200 - create a new thread to do import eviction to avoid deadlock (3969)
6201 - let lconf resolve symlinked-to devices (4629)
6202 - don't unlink "objects" from directory with default EA (4554)
6203 - hold socknal file ref over connect in case target is down (4394)
6204 - allow more than 32000 subdirectories in a single directory (3244)
6205 - fix blocks count for O_DIRECT writes (3751)
6206 - OST returns ENOSPC from object create when no space left (4539)
6207 - don't send truncate RPC if file size isn't changing (4410)
6208 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6209 - bind to privileged port in socknal and tcpnal (3689)
6211 - rate limit CERROR/CWARN console message to avoid overload (4519)
6212 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6213 - basic mmap support (3918)
6214 - kernel patch series update from b1_4 (4711)
6216 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6219 - avoid crash during MDS cleanup with OST shut down (2775)
6220 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6221 - don't use bad inodes on the MDS (3744)
6222 - dynamic object preallocation to improve recovery speed (4236)
6223 - don't hold spinlock over lock dumping or change debug flags (4401)
6224 - don't zero obd_dev when it is force cleaned (3651)
6225 - print grants to console if they go negative (4431)
6226 - "lctl deactivate" will stop automatic recovery attempts (3406)
6227 - look for existing locks in ldlm_handle_enqueue() (3764)
6228 - don't resolve lock handle twice in recovery avoiding race (4401)
6229 - revalidate should check working dir is a directory (4134)
6231 - don't always mark "slow" obdfilter messages as errors (4418)
6233 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6236 - don't close LustreDB during write_conf until it is done (3860)
6237 - fix typo in lconf for_each_profile (3821)
6238 - allow dumping logs from multiple threads at one time (3820)
6239 - don't allow multiple threads in OSC recovery (3812)
6240 - fix debug_size parameters (3864)
6241 - fix mds_postrecov to initialize import for llog ctxt (3121)
6242 - replace config semaphore with spinlock (3306)
6243 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6244 - don't allow enqueue to complete on a destroyed export (3822)
6245 - down write_lock before checking llog header bitmap (3825)
6246 - recover from lock replay timeout (3764)
6247 - up llog sem before sending rpc (3652)
6248 - reduce ns lock hold times when setting kms (3267)
6249 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6250 - fix NULL deref and obd_dev leak on setup error (3312)
6251 - replace some LBUG about llog ops with error handling (3841)
6252 - don't match INVALID dentries from d_lookup and spin (3784)
6253 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6254 - fix invalid assertion in ptlrpc_set_wait (3880)
6256 - add libwrap support for the TCP acceptor (3996)
6257 - add /proc/sys/portals/routes for non-root route listing (3994)
6258 - allow setting MDS UUID in .xml (2580)
6259 - print the stack of a process that LBUGs (4228)
6261 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6264 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6265 - make sure to unset replay flag from failed open requests (3440)
6266 - if default stripe count is 0, use OST count for inode size (3636)
6267 - update parent mtime/ctime on client for create/unlink (2611)
6268 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6269 - free recovery state on server during a forced cleanup (3571)
6270 - unregister_reply for resent reqs (3063)
6271 - loop back devices mounting and status check on 2.6 (3563)
6272 - fix resource-creation race that can provoke i_size == 0 (3513)
6273 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6274 - more debugging for page-accounting assertion (3746)
6275 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6276 - avoid deadlock after precreation failure (3758)
6277 - fix race and lock order deadlock in orphan handling (3450, 3750)
6278 - add validity checks when grabbing inodes from l_ast_data (3599)
6280 - add /proc/.../recovery_status to obdfilter (3428)
6281 - lightweight CDEBUG infrastructure, debug daemon (3668)
6282 - change default OSC RPC parameters to be better on small clusters
6283 - turn off OST read cache for files smaller than 32MB
6284 - install man pages and include them in rpms (3100)
6285 - add new init script for (un)mounting lustre filesystems (2593)
6286 - run chkconfig in %post for init scripts (3701)
6287 - drop scimac NAL (unmaintained)
6289 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6292 - clean kiobufs before and after use (3485)
6293 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6294 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6295 - change init script to fail more clearly if not run as root (1528)
6296 - allow clients to reconnect during replay (1742)
6297 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6298 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6299 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6300 - fix cond_resched (3554)
6301 - only evict unfinished clients after recovery (3515)
6302 - allow bulk resend, prevent data loss (3570)
6303 - dynamic ptlrpc request buffer allocation (2102)
6304 - don't allow unlinking open directory if it isn't empty (2904)
6305 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6306 - remove extraneous obd dereference causing LASSERT failure (3334)
6307 - don't use get_cycles() when creating temp. files on the mds (3156)
6308 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6309 - handle EEXIST for set-stripe, set proper directory name (3336)
6311 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6312 - fix ksocknal_fmb_callback() error messages (2918)
6314 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6317 - don't copy lvb into (possibly NULL) reply on error (2983)
6318 - don't deref dentry after dput, don't free lvb on error (2922)
6319 - use the kms to determine writeback rpc length (2947)
6320 - increment oti_logcookies when osc is inactive (2948)
6321 - update client's i_blocks count via lvb messages (2543)
6322 - handle intent open/close of special files properly (1557)
6323 - mount MDS with errors=remount-ro, like obdfilter (2009)
6324 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6325 - don't use cancelling-locks' kms values (2947)
6326 - use highest lock extent for kms, not last one (2925)
6327 - don't dereference ERR_PTR() dentry in error handling path (3107)
6328 - fix thread race in portals_debug_dumplog() (3122)
6329 - create lprocfs device entries at setup instead of at attach (1519)
6330 - common AST error handler, don't evict client on completion race (3145)
6331 - zero nameidata in detach_mnt in 2.6 (3118)
6332 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6333 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6334 - initialize RPC timeout timer earlier for 2.6 (3219)
6335 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6336 - print nal/nid for unknown nid (3258)
6337 - additional checks for oscc recovery before doing precreate (3284)
6338 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6339 - don't crash in mdc_close for bad permissions on open (3285)
6340 - zero i_rdev for non-device files (3147)
6341 - clear page->private before handing to FS, better assertion (3119)
6342 - tune the read pipeline (3236)
6343 - fix incorrect decref of invalidated dentry (2350)
6344 - provide read-ahead stats and refine rpc in flight stats (3328)
6345 - don't hold journal transaction open across create RPC (3313)
6346 - update atime on MDS at close time (3265)
6347 - close LDAP connection when recovering to avoid server load (3315)
6348 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6349 - don't leak open file on MDS after open resend (3325)
6350 - serialize filter_precreate and filter_destroy_precreated (3329)
6351 - loop device shouldn't call sync_dev() for nul device (3092)
6352 - clear page cache after eviction (2766)
6353 - resynchronize MDS->OST in background (2824)
6354 - refuse to mount the same filesystem twice on same mountpoint (3394)
6355 - allow llmount to create routes for mounting behind routers (3320)
6356 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6357 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6358 - fix rare problem with rename on htree directories (3417)
6360 - allow default OST striping configuration per directory (1414)
6361 - fix compilation for qswnal for 2.6 kernels (3125)
6362 - increase maximum number of MDS request buffers for large systems
6363 - change liblustreapi to be useful for external progs like lfsck (3098)
6364 - increase local configuration timeout for slow disks (3353)
6365 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6367 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6370 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6371 - don't overwrite extent policy data in reply if lock was blocked (2901)
6372 - drop filter export grants atomically with removal from device (2663)
6373 - del obd_self_export from work_list in class_disconnect_exports (2908)
6374 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6375 - swab reply message in mdc_close, other PPC fixes (2464)
6376 - fix destroying of named logs (2325)
6377 - overwrite old logs when running lconf --write_conf (2264)
6378 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6379 - fix race in target_handle_connect (2898)
6380 - mds_reint_create() should take same inode create lock (2926)
6381 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6382 - don't close files for self_export to avoid uninitialized obd (2936)
6383 - allow MDS with the same name as client node (2939)
6384 - hold dentry reference for closed log files for unlink (2325)
6385 - reserve space for all logs during transactions (2059)
6386 - don't evict page beyond end of stripe extent (2925)
6387 - don't oops on a deleted current working directory (2399)
6388 - handle hard links to targets without a parent properly (2517)
6389 - don't dereference NULL lock when racing during eviction (2867)
6390 - don't grow lock extents when lots of conflicting locks (2919)
6392 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6395 - account for cache space usage on clients to avoid data loss (974)
6396 - lfsck support in lustre kernel code (2349)
6397 - reduce journal credits needed for BRW writes (2370)
6398 - orphan handling to avoid losing space on client/server crashes
6399 - ptlrpcd can be blocked, stopping ALL progress (2477)
6400 - use lock value blocks to assist in proper KMS, faster stat (1021)
6401 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6402 - recovery for initial connections (2355)
6403 - fixes for mds_cleanup_orphans (1934)
6404 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6405 - block all file creations until orphan recovery completes (1901)
6406 - client remove rq_connection from request struct (2423)
6407 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6408 - recovery timer race (2670)
6409 - mdc_close recovey bug (2532)
6410 - ptlrpc cleanup bug (2710)
6411 - mds timeout on local locks (2588)
6412 - namespace lock held during RPCs (2431)
6413 - handle interrupted sync write properly (2503)
6414 - don't try to handle a message that hasn't been replied to (2699)
6415 - client assert failure during cleanup after abort recovery (2701)
6416 - leak mdc device after failed mount (2712)
6417 - ptlrpc_check_set allows timedout requests to complete (2714)
6418 - wait for inflight reqs when ptlrpcd finishes (2710)
6419 - make sure unregistered services are removed from the srv_list
6420 - reset bulk XID's when resending them (caught by 1138 test)
6421 - unregister_bulk after timeout
6422 - fix lconf error (2694)
6423 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6424 - readahead locks pages, leaves pending causing memory pressure (2673)
6425 - increase OST request buffers to 4096 on large machines (2729)
6426 - fix up permission of existing directories in simple_mkdir (2661)
6427 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6428 - don't assign transno to errored transactions (2742)
6429 - don't delete objects on OST if given a bogus objid from MDS (2751)
6430 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6431 - if rq_no_resend, then timeout request after recovery (2432)
6432 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6433 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6434 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6435 - MDS recovery shouldn't delete objects that it has given out (2730)
6436 - if enqueue arrives after completion, don't clobber LVB (2819)
6437 - don't unlock pages twice when trigger_group_io returns error (2814)
6438 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6439 - don't write pages to disk if there was an error (1450)
6440 - don't ping imports that have recovery disabled (2676)
6441 - take buffered bytes into account when balancing socknal conn (2817)
6442 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6443 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6444 - remove little-endian swabbing of llog records (1987)
6445 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6446 - retry reposting request buffers if they fail (1191)
6447 - grow extent at grant time to avoid granting a revoked lock (2809)
6448 - lock revoke doesn't evict page if covered by a second lock (2765)
6449 - disable VM readahead to avoid reading outside lock extents (2805)
6451 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6452 - updated kernel patches for hp-2.4.20 kernel (2681)
6454 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6457 - fix truncated write corruption (2366)
6458 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6460 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6461 - ensure len(uuid) < 37 in lmc (1171)
6462 - fix ia64 OOPS in llog_test (2255)
6463 - zero end of page at obdfilter for partial page writes (2648)
6464 - don't leave stale dentries around after renames (bug 2428)
6465 - fix timeouts when evicting a client with a single lock held (2642)
6466 - set deadline for the initial HELLO message to drain (2634)
6467 - print out dotted-quad IP addresses in the socknal (2302)
6469 - additional debugging for MDS client eviction problem (2443)
6470 - fix mkfsoptions support for osts (2603, 2604)
6472 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6475 - add series for the vanilla 2.6.0 kernel
6476 - add series for the vanilla 2.4.24 kernel
6477 - add series for a cray x86/64 UL kernel drop
6478 - fix xattr patches for the vanilla 2.4.19 series
6480 - generate true UUIDs in lmc (1171)
6481 - have portals stack dumping break in UML (2466)
6482 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6483 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6484 - wake callback waiting threads on client eviction (2460)
6485 - Add --ptldebug and --subsystem to lmc (1719)
6486 - update assertion to allow safe interrupt allocation
6487 - set rq_no_resend for cancel requests (2432)
6488 - recalculate ptlrpcd timeout after resend (2494)
6489 - call vfs_rmdir when removing pending directories (2368)
6490 - fix renaming a file to itself (2429)
6491 - lmc creates a default one-stripe lov (2454)
6492 - expand procfs space to handle large clusters (2326)
6493 - increase UML stack to avoid overflow
6494 - update lconf's list of debug and subsystem masks
6495 - fix lfs find --obd (2510)
6496 - /proc tunable for disabling filter read caching (2591)
6497 - stop rpm packages from altering slapd.conf (2301)
6498 - disable nagle in the socknal under 0conf (2578)
6499 - choose mds inode size based on stripe count (2572)
6500 - fix kernel-source rpm problems (2516)
6502 - add --disable-doc to avoid pdf generation (2421)
6503 - update documentation, tests, type-os, comments
6504 - avoid format warnings on ia64
6505 - remove the TOE NAL
6506 - tiny code cleanups by removing unused fields
6508 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6511 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6512 - remove the most blatant lies from BUILDING file (2371)
6513 - change default debug level to reasonable production setting
6514 - reduce client side cache size to reduce cache flush time
6515 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6516 - make TCP zerocopy and pinger support enabled by default (2476)
6517 - sync writes completed after process exits caused crashes (2319)
6518 - maintain correct mount count on the MDS (2356)
6519 - backout 1557, because 2316 wasn't really fixed
6520 - better file I/O statistics gathering in /proc
6521 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6522 - another kernel patch to fix zero-copy TCP function export
6523 - don't take duplicate lock when processing re-sent getattr (2420)
6524 - lctl uses obd_self_export instead of creating new conn (2353)
6525 - MDS/OST recovery case which requires object creation asserted (2425)
6526 - move lfs from /usr/sbin to /usr/bin in packages
6527 - fix race between mds_client_add and mds_client_free (2417)
6528 - use kmalloc instead of slabs in portals (2430)
6529 - don't create duplicate records when a failover MDS is present (2442)
6530 - remove unnecessary mount age check (2332)
6531 - don't remove directory inodes from locks prematurely (2451)
6532 - don't break if MDS service name is the same as hostname (2103)
6533 - fix races in client write RPC generation when cache full (2482)
6535 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6538 - remove now-unused request->rq_obd (278)
6539 - if an allocation fails, print out how much memory we've used (1933)
6540 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6541 - add the "configurable stack size" patch to most series files (1256)
6542 - ability to write large log records, for 100+ OST configs (2306)
6543 - fix NULL deref when filter_prep fails (2314)
6544 - fix operator precedence error in filter_sync
6545 - dynamic allocation of socknal TX descriptors (2315)
6546 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6547 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6548 - ensure that CWARN messages in Portals always get to the syslog
6549 - __init/__exit are not for prototype decls (ldlm_init/exit)
6550 - x86-64 compile warning fixes
6551 - fix gateway LMC keyword conflict (2318)
6552 - fix MDS lock inversions in getattr/reint paths (1844)
6553 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6554 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6555 - DLM race condition prevented some lock evictions (2328)
6556 - ENOMEM detection and retry on socknal sends (2230)
6557 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6558 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6559 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6560 - better error messages when a client is rejected during recovery (1505)
6561 - avoid cancelling locks which were never granted, after failure (2330)
6562 - fix i_sem/journal inversion in mds_client_add (2333)
6563 - fix truncate/getattr lock cycle deadlock (2334)
6564 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6565 - fix two rare exit paths which could leak an l_lock() ref (2321)
6566 - fencepost error in MDS/OST orphan recovery (2226)
6567 - make log record alignment 8 bytes (1988)
6568 - lstripe now fails when requested offset > ost_count (2237)
6569 - ensure that all kernel series have a complete list.h (1607)
6570 - fix crashes in special-file operations (2316)
6571 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6573 - allow configurable automake binary, for testing new versions
6574 - small update to the lfs documentation
6576 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6578 * fix negative export reference count in fsfilt_sync (2312)
6580 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6581 * release candidate 0.9.1
6583 - orphans are moved into the PENDING directory for possible recovery
6584 - replayed opens now open by fid for orphan/rename safety (1042)
6585 - last close of an orphan inode generates a transno (683)
6586 - chdir() and mount() now pin the directory entry (1020)
6587 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6588 - discard very old requests without processing them (1502)
6589 - remove some common, well-understood CERRORs (1505)
6590 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6591 - clear "grant" flags in OST replies until OST grant code lands (1644)
6592 - fix read performance by not clobbering i_blksize on client (1598)
6593 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6594 - make LRU size a /proc tunable, clears locks when reduced (707)
6595 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6596 - server locks take references on exports now (1558)
6597 - build fixes for 2.4.20-rh trees (1663)
6598 - return an error from lov_create if all OSCs are inactive (1751)
6599 - fix import levels when a reconnect happens without a timeout (1597)
6600 - exit early from mds_open if we get a lookup error (1749)
6601 - partial page read at EOF wouldn't wait for disk before sending (1642)
6602 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6603 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6604 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6605 - write/truncate lock inversion (1639)
6606 - Don't auto-load obdclass, portals modules during cleanup (1495)
6607 - fix timestamps from jumping to "now" (1763)
6608 - extra journal assertions (1648)
6609 - add an extra multiunlink test (1771)
6610 - fix read_record/write_record API (1776)
6611 - fix leak of offset_extent, possible incorrect i_size later (1772)
6612 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6613 - Debugging for the kqswnal_get_idle_tx problems (1820)
6614 - Allow recovery to be attempted multiple times (1536)
6615 - Write out MDS last_rcvd file after it is first created (1600)
6616 - Fix tx_descriptor leak in failed transmit situations (1827)
6617 - ext3 journaling fixes for assertion failure after IO error (1871)
6618 - class_export_put() on freed export after completion AST error (1896)
6619 - Fix revalidate looping in VFS (1322)
6620 - Don't access a freed export during MDS_REINT timeout (1521)
6621 - Add open-unlink recovery support on the MDS (1673,1764)
6622 - Return an error if no MDS data was read from last_rcvd (1946)
6623 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6624 - Don't setup a disk device that doesn't match exported UUID (317)
6625 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6626 - avoid committing NULL handle in force close
6627 - local.sh is now a one-stripe LOV configuration
6628 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6629 - fix ext3 htree duplicate directory entry corruption (1516)
6630 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6631 - update to new LOV EA format (2097)
6632 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6634 2003-06-15 Phil Schwan <phil@clusterfs.com>
6637 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6638 - per-import/export recovery handling (958, 931, 959)
6639 - multiple last-rcvd slots, for serving multiple FSes (949)
6640 - connections are again shared between multiple imp/exports (963, 964)
6641 - "umount -f" would hang if any requests needed to be sent (393, 978)
6642 - avoid pinning large req buffer by copying for queued messages (989)
6643 - add "uuid" to "lctl device" command to help upcalls (991)
6644 - "open" RPCs with transnos would confuse recovery counters (1037)
6645 - do proper endian conversion of all wire messages (288, 340, 891)
6646 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6647 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6648 - fix ptlrpc_connection leak in target_handle_connect (1174)
6649 - fix import refcounting bug in OST and MDS cleanup (1134)
6650 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6651 - fix very unlikely obd_types race condition (501)
6652 - remove osc_open hack for echo_client (1187)
6653 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6654 - a failure in read_inode2 leads to deadlock (1139)
6655 - cancel ack-locks as soon as transaction is committed (1072)
6656 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6657 - make sure to commitrw after any preprw to avoid deadlock (1162)
6658 - failing to execute a file in a lustre FS would lock inode (1203)
6659 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6660 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6661 - obd_uuid2tgt would walk off the end of the list (1255)
6662 - on IA64 the osc would give portals incorrect bulk size (1258)
6663 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6664 - fix lock inversion caused by new llite matching code (1282)
6665 - limit the number of dirty pages on a client to 10MB (1286)
6666 - timed out locks were not being corrected cancelled (1289)
6667 - fix O_DIRECT above 4GB on IA-32 (1292)
6668 * major user-visible changes
6669 - fail out/fail over policy now controlled by the upcall (993)
6671 - add OBD_PING to check server availability and failure (954)
6672 - lustre messages are now sent in sending host order (288, 340, 891)
6673 - add eadatalen to MDS getattr reply (340)
6674 - OST read replies may contain second buffer, with per-page status (593)
6676 2003-03-11 Phil Schwan <phil@clusterfs.com>
6679 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6680 - failed open()s could cause deadlock; fixed (867, 869)
6681 - stop cancelling OST locks when files are closed (481)
6682 - overlapping XID spaces caused network corruption (851, 853)
6683 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6684 - setattr_raw added, to do single-RPC, server-side setattrs
6685 - lmc/lconf syntax change for OST UUIDs
6686 - fix crashy race condition between ptlrpc_free_req and osc_close
6687 - don't use request in mdc_enqueue if we hit a timeout (889)
6688 - don't set the inode i_size for regular files from the MDS (896)
6689 - handle out of order completion AST (842)
6690 - don't LBUG if a lock request times out after receiving AST (913)
6691 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6692 - if a bad lock AST arrives, send an error instead of dropping entirely
6693 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6694 - fix leak in bulk IO when only partially completed (899, 900, 926)
6695 - fix O_DIRECT for ia64 (55)
6696 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6697 - C-z after timeout could hang a process forever; fixed (977)
6699 - client-side I/O cache (678, 924, 929, 941, 970)
6701 - READPAGE and SETATTRs which don't take server-side locks get
6704 2003-02-11 Phil Schwan <phil@clusterfs.com>
6707 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6708 - Fix fsfilt_extN_readpage() to read a full page of directory
6709 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6710 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6711 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6712 - Most intent operations were removed, in favour of a new RPC mode
6713 that does a single RPC to the server and bypasses most of the VFS
6714 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6715 - Aggressively cancel local locks on DLM servers
6716 - mds_reint_unlink sends EA to the client if it's the last nlink.
6717 client uses that EA to unlink OST objects.
6718 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6719 - recursive symlinks were fixed (439)
6720 - fixed NULL deref in DEBUG_REQ
6721 - filter_update_lastobjid no longer calls sync, which annoyed extN
6722 - fixed multi-client small-writes to a single file problem (445)
6723 - fixed mtime updates during file writes (607)
6724 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6725 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6726 - fixed Linux OST size reporting problem (444, 656)
6727 - OST now updates object mtime with writes or setattr (607, 619)
6728 - client verifies file size before zeroing page past EOF (445)
6729 - OST now writes last allocated objid to disk with allocation (108)
6730 - LOV on echo now works (409)
6732 - mds_reint_unlink sends a new buffer, with the EA included. this
6733 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6734 set if a regular file was being unlinked, and it was the last link
6735 - use PtlGet from the target for bulk writes (315)
6736 - OST now updates object mtime with writes or setattr (607, 619)
6737 - LDLM now has a grant-time callback to revalidate locked items, if
6739 - Many MDS operations were reorganized to combat race conditions
6741 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6742 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6744 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6747 - Fully reactivate OST imports after reconnection (512, others)
6748 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6749 - More graceful error handling for truncating on dead OST (515)
6750 - Don't error out unless we're actually accessing dead stripes (474)
6751 - Fix garbage sizes when stripes are missing (410)
6752 - LRU counters were broken, causing constant lock purge (433, 432)
6753 - garbage on read from stripes with failed OSTs (441)
6754 - mark OSCs as active before reconnecting during recovery (438)
6755 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6756 - lfind did not preserve OST order in output (443)
6757 - symlinks cause hung clients, incorrect data (439)
6758 - stop dereferencing request after dropping refcount (457)
6759 - don't LASSERT(spin_is_locked) on non-SMP (455)
6760 - fixes for many rename() bugs
6761 - fstat didn't correctly synchronize attributes (399)
6762 - server must handle lock cancellation during blocking AST prep (487)
6763 - bulk descriptors were free()d too soon (511)
6764 - fix paths in lconf, which would load incorrect modules (451, 507)
6765 - fix confusing lconf 'host not found' error message (386)
6766 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6767 - fix race condition in mdc_blocking_ast() for inode access (526)
6768 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6769 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6770 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6771 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6772 - fix lstripe in several ways (kernel side) (527)
6773 - fix request leak in ldlm_cli_enqueue (262)
6774 - incorrect OSC was marked inactive after OST failure
6775 - call mds_fs_cleanup before unmounting filesystem (524)
6776 - fix races between taking ns_lock and ldlm_lock_change_resource
6777 - fix races updating LOV export open file list
6778 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6779 - fix recovery NULL deref in ldlm_cli_cancel_unused
6780 - fix some DLM races by using new hash table for lock handles (419)
6781 - permit the client to specify desired inodes, at replay
6782 - duplicate requests when we queue them for replay reintegration
6783 - fix last_rcvd offset calculation
6784 - sync after each recovered transaction, so we always make progress
6785 - never, not always, ERESTART requests without transnos
6786 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6788 - skip replay if the MDS says that the client is already connected
6789 - don't check for a recovery-enabled export to match lctl's UUID
6790 - don't INC_USE_COUNT for phantom exports
6791 - don't crash when cleaning up phantom exports (567)
6792 - don't double-finish or set replay data for errored mdc_open requests
6793 - abort requests when they time out, so we don't get old replies
6794 - send/receive replies for AST messages again
6795 - if the client says that it doesn't have the lock, cancel it on the
6797 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6798 mark it as destroyed, it will all work out when decref is called
6799 - fix module use counts (22, 581)
6801 - ASTs now expect a reply (server cancels lock on error reply)
6803 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6806 - fix many simultaneous client startup (392)
6807 - fix dentry->d_it clobbering
6808 - credentials weren't being shipped for readdir/getattr operations
6809 - remove invalid assertions triggered during some concurrent MD
6811 - proper Lustre versions added (336, 389)
6812 - fix memory leak for create error case (398)
6813 - fix LOV locking bug that would get cli/srv out of sync
6814 - fix echo client over LOV (409)
6815 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6816 - fix double-O_EXCL intent crash (424)
6817 - avoid sending multiple lock CANCELs (352)
6819 - MDS can do multi-client recovery (modulo bugs in new code)
6821 - many updates, edits, cleanups
6823 2002-11-18 Phil Schwan <phil@clusterfs.com>
6826 - fix null d_it dereference (346)
6827 - fix full OST/dbench hang (333)
6828 - fix permission problem with file removal (286)
6829 - fix removal of OSCs from LOV when they fail
6830 - fix NULL deref during bulk timeout (214)
6831 - fix problems related to multiple filesystems on one MDS (241)
6832 - fixed serious subtle metadata locking bugs
6833 - free locks on clients when inodes are removed due to memory
6835 - fix inode pointer in lock data (285)
6836 - partial support for multiple MDS on a single host (241)
6837 - data locks weren't cancelled at clear_inode time (290, 311)
6838 - intent locks could lead to unbounded lock growth (205)
6839 - added a maximum lock count, an LRU list, and a flusher
6840 - fix multiple rename (365)
6841 - properly abstracted the echo client
6842 - OSC locked 1 byte too many; fixed
6843 - rewrote brw callback code:
6844 - fixed recovery bugs related to LOVs (306)
6845 - fixed too-many-pages-in-one-write crash (191)
6846 - fixed (again) crash in sync_io_timeout (214)
6847 - probably fixed callback-related race (385)
6849 - Add capability to MDS protocol
6850 - LDLM cancellations and callbacks on different portals
6852 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6855 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6856 - defer open object destruction to close time (601981, 138)
6857 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6858 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6859 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6860 - don't call obd_setattr in ll_file_release for destroyed objects
6861 * protocol change to lustre_msg: move |version| and add |flags|
6862 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6863 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6864 * added replay of create, unlink, link and rename operations during
6865 MDS failover; recovery should be much more robust now
6866 * remove failed OSCs from LOVs (only lov_create uses this so far)
6867 * the lustre-HOWTO was brought (more) up to date (582544)
6869 2002-10-23 Phil Schwan <phil@clusterfs.com>
6872 - in-use dentries weren't being reused properly (617851)
6873 - prevent multiple LDLM setup (599178)
6874 - fix LOV size calculations for truncate (617853)
6875 - fix client handling of MDS intent errors (POSIX)
6876 - fix permission bug in lovstripe.c test (624321)
6877 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6878 - truncate past end of file could corrupt data
6879 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6880 - a race in recovery could return ETIMEDOUT to apps (623947)
6881 - building outside the source directory was fixed
6882 * the lustre-HOWTO was brought (more) up to date (582544)
6883 * major progress was made on recovery functionality
6885 2002-10-10 Phil Schwan <phil@clusterfs.com>
6888 - recovery deadlock fix
6889 - rm -rf causes LBUG fix (617817)
6890 - file open by multiple tasks fix (618962)
6891 - directory permissions bugs (602707 and 620007)
6892 - journal_stop fixed with locking (611313)
6893 - O_APPEND failures resolved (618273, perhaps 614459)
6894 - lconf PATH fix (619770)
6895 - IA64 build fix (621450)
6896 - RPC buffer sizes scale with amount of memory
6898 2002-10-01 Phil Schwan <phil@clusterfs.com>
6901 - locks would be cancelled without throwing away data pages,
6902 resulting in inconsistent data (605627)
6903 - inode attributes were not always being refreshed (605627, 612449)
6904 - lconf now continues to cleanup after lctl reports an error
6905 - MDS now enforces user permissions (602707)
6906 - lprocfs cleanup fixed, but not yet enabled (614157)
6907 - fixed infinite server hang, should a client not respond to an AST
6908 - avoid going into recovery if user calls readlink() with a buffer
6909 that's too small (613941)
6910 - AST RPCs no longer require replies (614867) -- this may be changed
6911 - don't crash server if client sends an IOV that's too big (611336)
6912 - fixed lock conversion deadlock (611892)
6913 - fixed the following of symlinks (614622)
6914 * recovery: the server can remove locks from a client that dies, other
6915 clients can make progress
6916 * more extN patch fixes
6917 * compile-time configurable ptlrpc buffer allocations
6919 - collaborative read cache document
6920 - Lustre Lite Performance CDR document-in-progress
6922 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6925 - fix typo in patch-2.4.18
6927 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6930 - clear ptlrpc request each time in handle_incoming_request()
6931 - unlink of files now destroys the object on the OST
6933 2002-09-19 Peter Braam <braam@clusterfs.com>
6935 * add hard link support
6936 * change obdfile creation method
6937 * kernel patch changed
6939 2002-09-19 Peter Braam <braam@clusterfs.com>
6942 - stack overflow bug in extN fixed
6944 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6946 * documentation updates
6947 - add man pages for config tools
6948 - update tests/README to describe testing with new config tools
6949 - finish metadata API descriptions
6950 * bug fixes and cleanups
6951 - statfs workaround for 16TB limit
6952 - LOV stripe allocation improved, can stripe on subset of OSTs
6953 - LOV file size/IO offset was wrong for files > 4GB in size
6954 - object EA data was being dropped, caused files to be unreadable
6955 - memory overflow with non-LOV OST caused memory corruption
6956 - fixed regression tests to work with new config tools, obdfilter
6957 - fixed bug when directory size became larger than 1 block
6958 - fixed bug (for single client case) when PWD was deleted
6959 - invalidate local directory pages when doing intent-based ops
6960 - avoid LDLM oops when lock callback contained bad data
6962 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6964 * documentation updates
6965 * bug fixes and cleanups
6966 - configuration tools
6969 - 64-bit compile warnings
6970 - 64-bit internal statfs data
6972 * test_brw on persistent OST devices
6974 * lprocfs (disabled)
6976 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6978 * documentation updates
6979 * bug fixes and cleanups
6980 * configuration tools
6982 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6985 * many small fixes to 0_5_4
6986 * io/network handling
6987 * thinkos in MDS operations
6989 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6992 * crucial basic fixes to 0.5.3
6993 * IOR, Iozone work over Elan
6996 2002-08-07 Phil Schwan <phil@clusterfs.com>
6997 * version 0_5_3, our first alpha
6998 * we use the new Portals iovs
6999 * documentation updates
7000 * bug fixes and cleanups
7001 * small changes in the DLM wire protocol
7003 2002-07-25 Peter J. Braam <braam@clusterfs.com>
7004 * version 0_5_1 with some initial stability,
7005 * locking on MD and file I/O.
7006 * documentation updates
7007 * several bug fixes since 0.5.0
7008 * small changes in wire protocol
7010 2002-07-18 Phil Schwan <phil@clusterfs.com>
7012 * delivered as Lustre Light Alpha
7013 * fixed a crash after handling invalid MDS requests
7014 * fixed directory pages for architectures with non-4k pages sizes
7016 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
7017 * release version v0_4_4
7018 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
7020 * Fixes a number of interruption problems with OST operations.
7021 * Update documentation for portals header changes
7022 * Move all wire protocol structs/defines to lustre_idl.h
7023 * Fixes symlink length bug.
7024 * Add tcpdump to repository.
7026 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
7027 * release version v0_4_3
7028 * Fixes statfs for inodes on extN.
7029 * Fixes bug in runtests which would delete /etc/hosts.
7030 * Use 64-bit object IDs wherever possible (not into VFS though)
7031 Remove ost_get_info, which is unused by lustre, and out of date.
7033 2002-07-03 Peter Braam <braam@clusterfs.com>
7034 * release version v0_4_2 Fixes a lookup error (type not passed)
7035 * move forward to head of Portals
7036 * move forward to latest Lustre kernel
7038 2002-06-25 Peter Braam <braam@clusterfs.com>
7039 * release version v0_4_1. Hopefully stable on single node use.