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 5.14.0-162.18.1.el9 (RHEL9.1)
7 4.18.0-425.10.1.el8 (RHEL8.7)
8 * Other server kernels known to build and work at some point (others may also work):
9 3.10.0-862.14.4.el7 (RHEL7.5)
10 3.10.0-957.27.2.el7 (RHEL7.6)
11 3.10.0-1062.18.1.el7 (RHEL7.7)
12 3.10.0-1127.19.1.el7 (RHEL7.8)
13 3.10.0-1160.83.1.el7 (RHEL7.9)
14 4.18.0-80.11.2.el8 (RHEL8.0)
15 4.18.0-147.8.1.el8 (RHEL8.1)
16 4.18.0-193.28.1.el8 (RHEL8.2)
17 4.18.0-240.22.1.el8 (RHEL8.3)
18 4.18.0-305.25.1.el8 (RHEL8.4)
19 4.18.0-348.23.1.el8 (RHEL8.5)
20 4.18.0-372.32.1.el8 (RHEL8.6)
21 4.4.120-92.70 (SLES12 SP2)
22 4.4.180-94.100 (SLES12 SP3)
23 4.4.0-131 (Ubuntu 16.04)
24 4.15.0-32 (Ubuntu 18.04)
25 5.4.0-48 (Ubuntu 20.04)
26 vanilla linux 5.4.0 (ZFS + ldiskfs)
27 vanilla linux 5.4.21 (ZFS + ldiskfs)
28 vanilla linux 5.4.136 (ZFS + ldiskfs)
29 5.10.0-60.79.0.103.oe2203 (openEuler 22.03 LTS)
30 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
31 * Client primary kernels built and tested during release cycle:
32 5.14.0-162.18.1.el9 (RHEL9.1)
33 4.18.0-425.10.1.el8 (RHEL8.7)
34 5.4.0-37 (Ubuntu 20.04)
35 5.14.21-150400.24.46 (SLES15 SP4)
36 * Other clients known to build on these kernels at some point (others may also work):
37 3.10.0-862.14.4.el7 (RHEL7.5)
38 3.10.0-957.27.2.el7 (RHEL7.6)
39 4.14.0-49.13.1.el7a (RHEL7.5)
40 4.14.0-115.2.2.el7a (RHEL7.6)
41 3.10.0-1062.18.1.el7 (RHEL7.7)
42 3.10.0-1127.19.1.el7 (RHEL7.8)
43 3.10.0-1160.83.1.el7 (RHEL7.9)
44 4.18.0-80.11.2.el8 (RHEL8.0)
45 4.18.0-147.8.1.el8 (RHEL8.1)
46 4.18.0-193.28.1.el8 (RHEL8.2)
47 4.18.0-240.22.1.el8 (RHEL8.3)
48 4.18.0-305.25.1.el8 (RHEL8.4)
49 4.18.0-348.23.1.el8 (RHEL8.5)
50 4.18.0-372.32.1.el8 (RHEL8.6)
51 5.14.0-70.30.1.el9 (RHEL9.0)
52 4.4.120-92.70 (SLES12 SP2)
53 4.4.180-94.100 (SLES12 SP3)
54 4.12.14-95.48 (SLES12 SP4)
55 4.12.14-122.133 (SLES12 SP5)
56 4.12.14-197.75 (SLES15 SP1)
57 5.3.18-24.96 (SLES15 SP2)
58 5.3.18-150300.59.93 (SLES15 SP3)
59 4.4.0-131 (Ubuntu 16.04)
60 4.15.0-48 (Ubuntu 18.04)
61 5.8.0-53 (Ubuntu 20.04.2 HWE)
62 5.11.0-31 (Ubuntu 20.04.3 HWE)
63 5.11.0 (vanilla kernel.org)
64 5.10.0 (openEuler 22.03 LTS)
65 * Recommended e2fsprogs version: 1.46.6-wc1 or newer
66 * Recommended ZFS version: 2.1.5
67 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
68 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
69 stack overflow. For more information, please refer to bugzilla 17630.
70 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
71 the re-exporting nodes to be mounted with "32bitapi" mount option
73 --------------------------------------------------------------------------------
78 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
79 for currently supported client and server kernel versions.
80 * Server primary kernels built and tested during release cycle:
81 4.18.0-348.2.1.el8 (RHEL8.5)
82 * Other server kernels known to build and work at some point (others may also work):
83 3.10.0-862.14.4.el7 (RHEL7.5)
84 3.10.0-957.27.2.el7 (RHEL7.6)
85 3.10.0-1062.18.1.el7 (RHEL7.7)
86 3.10.0-1127.19.1.el7 (RHEL7.8)
87 3.10.0-1160.49.1.el7 (RHEL7.9)
88 4.18.0-80.11.2.el8 (RHEL8.0)
89 4.18.0-147.8.1.el8 (RHEL8.1)
90 4.18.0-193.28.1.el8 (RHEL8.2)
91 4.18.0-240.22.1.el8 (RHEL8.3)
92 4.18.0-305.25.1.el8 (RHEL8.4)
93 4.4.120-92.70 (SLES12 SP2)
94 4.4.180-94.100 (SLES12 SP3)
95 4.4.0-131 (Ubuntu 16.04)
96 4.15.0-32 (Ubuntu 18.04)
97 5.4.0-48 (Ubuntu 20.04)
98 vanilla linux 5.4.0 (ZFS + ldiskfs)
99 vanilla linux 5.4.21 (ZFS + ldiskfs)
100 vanilla linux 5.4.136 (ZFS + ldiskfs)
101 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
102 * Client primary kernels built and tested during release cycle:
103 4.18.0-348.2.1.el8 (RHEL8.5)
104 5.4.0-37 (Ubuntu 20.04)
105 5.3.18-59.27 (SLES15 SP3)
106 * Other clients known to build on these kernels at some point (others may also work):
107 3.10.0-862.14.4.el7 (RHEL7.5)
108 3.10.0-957.27.2.el7 (RHEL7.6)
109 4.14.0-49.13.1.el7a (RHEL7.5)
110 4.14.0-115.2.2.el7a (RHEL7.6)
111 3.10.0-1062.18.1.el7 (RHEL7.7)
112 3.10.0-1127.19.1.el7 (RHEL7.8)
113 3.10.0-1160.49.1.el7 (RHEL7.9)
114 4.18.0-80.11.2.el8 (RHEL8.0)
115 4.18.0-147.8.1.el8 (RHEL8.1)
116 4.18.0-193.28.1.el8 (RHEL8.2)
117 4.18.0-240.22.1.el8 (RHEL8.3)
118 4.18.0-305.25.1.el8 (RHEL8.4)
119 4.4.120-92.70 (SLES12 SP2)
120 4.4.180-94.100 (SLES12 SP3)
121 4.12.14-95.48 (SLES12 SP4)
122 4.12.14-122.91 (SLES12 SP5)
123 4.12.14-197.75 (SLES15 SP1)
124 5.3.18-24.96 (SLES15 SP2)
125 4.4.0-131 (Ubuntu 16.04)
126 4.15.0-48 (Ubuntu 18.04)
127 5.8.0-53 (Ubuntu 20.04.2 HWE)
128 5.11.0-31 (Ubuntu 20.04.3 HWE)
129 5.11.0 (vanilla kernel.org)
130 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
131 * Recommended ZFS version: 2.0.7
132 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
133 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
134 stack overflow. For more information, please refer to bugzilla 17630.
135 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
136 the re-exporting nodes to be mounted with "32bitapi" mount option
138 --------------------------------------------------------------------------------
142 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
143 for currently supported client and server kernel versions.
144 * Server primary kernels built and tested during release cycle:
145 3.10.0-1062.18.1.el7 (RHEL7.7)
146 3.10.0-1127.19.1.el7 (RHEL7.8)
147 3.10.0-1160.6.1.el7 (RHEL7.9)
148 * Other server kernels known to build and work at some point (others may also work):
149 3.10.0-862.14.4.el7 (RHEL7.5)
150 3.10.0-957.27.2.el7 (RHEL7.6)
151 4.18.0-80.11.2.el8 (RHEL8.0)
152 4.18.0-147.8.1.el8 (RHEL8.1)
153 4.18.0-193.28.1.el8 (RHEL8.2)
154 4.18.0-240.1.1.el8 (RHEL8.3)
155 4.4.120-92.70 (SLES12 SP2)
156 4.4.180-94.100 (SLES12 SP3)
157 4.4.0-131 (Ubuntu 16.04)
158 4.15.0-32 (Ubuntu 18.04)
159 5.4.0-48 (Ubuntu 20.04)
160 vanilla linux 5.4.0 (ZFS + ldiskfs)
161 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
162 * Client primary kernels built and tested during release cycle:
163 3.10.0-1062.18.1.el7 (RHEL7.7)
164 3.10.0-1127.19.1.el7 (RHEL7.8)
165 3.10.0-1160.6.1.el7 (RHEL7.9)
166 4.12.14-95.48 (SLES12 SP4)
167 4.12.14-122.20 (SLES12 SP5)
168 4.12.14-197.75 (SLES15 SP1)
169 5.3.18-24.24 (SLES15 SP2)
170 4.15.0-48 (Ubuntu 18.04)
171 * Other clients known to build on these kernels at some point (others may also work):
172 3.10.0-862.14.4.el7 (RHEL7.5)
173 3.10.0-957.27.2.el7 (RHEL7.6)
174 4.14.0-49.13.1.el7a (RHEL7.5)
175 4.14.0-115.2.2.el7a (RHEL7.6)
176 4.18.0-80.11.2.el8 (RHEL8.0)
177 4.18.0-147.8.1.el8 (RHEL8.1)
178 4.18.0-193.28.1.el8 (RHEL8.2)
179 4.18.0-240.1.1.el8 (RHEL8.3)
180 4.4.120-92.70 (SLES12 SP2)
181 4.4.180-94.100 (SLES12 SP3)
182 4.4.0-131 (Ubuntu 16.04)
183 5.4.0-37 (Ubuntu 20.04)
184 5.4.0 (vanilla kernel.org)
185 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
186 * Recommended ZFS version: 2.0.0
187 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
188 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
189 stack overflow. For more information, please refer to bugzilla 17630.
190 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
191 the re-exporting nodes to be mounted with "32bitapi" mount option
193 --------------------------------------------------------------------------------
197 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
198 for currently supported client and server kernel versions.
199 * Server primary kernels built and tested during release cycle:
200 3.10.0-1062.1.1.el7 (RHEL7.7)
201 * Other server kernels known to build and work at some point (others may also work):
202 3.10.0-862.14.4.el7 (RHEL7.5)
203 3.10.0-957.27.2.el7 (RHEL7.6)
204 4.18.0-80.7.1.el8 (RHEL8.0)
205 4.4.120-92.70 (SLES12 SP2)
206 4.4.180-94.100 (SLES12 SP3)
207 4.4.0-131 (Ubuntu 16.04)
208 4.15.0-32 (Ubuntu 18.04)
209 vanilla linux 4.6.7 (ZFS only)
210 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
211 * Client primary unpatched kernels built and tested during release cycle:
212 3.10.0-1062.1.1.el7 (RHEL7.7)
213 4.12.14-95.29 (SLES12 SP4)
214 4.15.0-48 (Ubuntu 18.04)
215 * Other clients known to build on unpatched kernels at some point (others may also work):
216 3.10.0-862.14.4.el7 (RHEL7.5)
217 3.10.0-957.27.2.el7 (RHEL7.6)
218 4.14.0-49.13.1.el7a (RHEL7.5)
219 4.14.0-115.2.2.el7a (RHEL7.6)
220 4.18.0-80.7.1.el8 (RHEL8.0)
221 4.4.120-92.70 (SLES12 SP2)
222 4.4.180-94.100 (SLES12 SP3)
223 4.4.0-131 (Ubuntu 16.04)
224 vanilla linux 4.15.0 (ZFS only)
225 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
226 * Recommended ZFS / SPL version: 0.7.13
227 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
228 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
229 stack overflow. For more information, please refer to bugzilla 17630.
230 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
231 the re-exporting nodes to be mounted with "32bitapi" mount option
232 * Note for Ubuntu 14.04.5:
233 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
234 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
236 --------------------------------------------------------------------------------
240 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
241 for currently supported client and server kernel versions.
242 * Server known to build on the following kernels (others may also work):
243 2.6.32-696.18.7.el6 (RHEL6.9)
244 3.10.0-862.14.4.el7 (RHEL7.5)
245 3.10.0-957.el7 (RHEL7.6)
246 4.4.120-92.70 (SLES12 SP2)
247 4.4.162-94.69 (SLES12 SP3)
248 4.4.0-131 (Ubuntu 16.04)
249 4.15.0-32 (Ubuntu 18.04)
250 vanilla linux 4.6.7 (ZFS only)
251 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
252 * Client known to build on unpatched kernels (others may also work):
253 2.6.32-696.18.7.el6 (RHEL6.9)
254 2.6.32-754.6.3.el6 (RHEL6.10)
255 3.10.0-862.14.4.el7 (RHEL7.5)
256 3.10.0-957.el7 (RHEL7.6)
257 4.14.0-49.13.1.el7a (RHEL7.5)
258 4.14.0-115.2.2.el7a (RHEL7.6)
259 4.4.120-92.70 (SLES12 SP2)
260 4.4.162-94.69 (SLES12 SP3)
261 4.4.0-131 (Ubuntu 16.04)
262 4.15.0-32 (Ubuntu 18.04)
263 vanilla linux 4.15.0 (ZFS only)
264 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
265 * Recommended ZFS / SPL version: 0.7.9
266 * Tested with ZFS / SPL version: 0.7.9
267 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
268 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
269 stack overflow. For more information, please refer to bugzilla 17630.
270 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
271 the re-exporting nodes to be mounted with "32bitapi" mount option
272 * Note for Ubuntu 14.04.5:
273 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
274 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
276 --------------------------------------------------------------------------------
278 2018-04-03 Intel Corporation
280 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
281 for currently supported client and server kernel versions.
282 * Server known to build on patched kernels:
283 2.6.32-431.29.2.el6 (RHEL6.5)
284 2.6.32-504.30.3.el6 (RHEL6.6)
285 2.6.32-573.26.1.el6 (RHEL6.7)
286 2.6.32-642.15.1.el6 (RHEL6.8)
287 2.6.32-696.18.7.el6 (RHEL6.9)
288 3.10.0-693.21.1.el7 (RHEL7.4)
289 3.0.101-0.47.71 (SLES11 SP3)
290 3.0.101-107 (SLES11 SP4)
291 3.12.74-60.64.40 (SLES12 SP1)
292 4.4.114-92.64 (SLES12 SP2)
293 4.4.114-94.11 (SLES12 SP3)
294 3.13.0-101 (Ubuntu 14.04, ZFS only)
295 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
296 4.4.0-116 (Ubuntu 16.04)
297 vanilla linux 4.6.7 (ZFS only)
298 * Client known to build on unpatched kernels:
299 2.6.32-431.29.2.el6 (RHEL6.5)
300 2.6.32-504.30.3.el6 (RHEL6.6)
301 2.6.32-573.26.1.el6 (RHEL6.7)
302 2.6.32-642.15.1.el6 (RHEL6.8)
303 2.6.32-696.18.7.el6 (RHEL6.9)
304 3.10.0-693.21.1.el7 (RHEL7.4)
305 3.0.101-0.47.71 (SLES11 SP3)
306 3.0.101-107 (SLES11 SP4)
307 3.12.74-60.64.40 (SLES12 SP1)
308 4.4.114-92.64 (SLES12 SP2)
309 4.4.114-94.11 (SLES12 SP3)
310 3.13.0-101 (Ubuntu 14.04)
311 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
312 4.4.0-116 (Ubuntu 16.04)
314 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
315 * Recommended ZFS / SPL version: 0.7.6
316 * Tested with ZFS / SPL version: 0.7.6
317 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
318 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
319 stack overflow. For more information, please refer to bugzilla 17630.
320 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
321 the re-exporting nodes to be mounted with "32bitapi" mount option
322 * Note for Ubuntu 14.04.5:
323 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
324 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
326 --------------------------------------------------------------------------------
328 2017-07-05 Intel Corporation
330 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
331 for currently supported client and server kernel versions.
332 * Server known to build on patched kernels:
333 2.6.32-431.29.2.el6 (RHEL6.5)
334 2.6.32-504.30.3.el6 (RHEL6.6)
335 2.6.32-573.26.1.el6 (RHEL6.7)
336 2.6.32-642.15.1.el6 (RHEL6.8)
337 2.6.32-696.3.1.el6 (RHEL6.9)
338 3.10.0-514.21.1.el7 (RHEL7.3)
339 3.0.101-0.47.71 (SLES11 SP3)
340 3.0.101-100 (SLES11 SP4)
341 3.12.74-60.64.40 (SLES12 SP1)
342 4.4.59-92.17 (SLES12 SP2)
343 vanilla linux 4.6.7 (ZFS only)
344 * Client known to build on unpatched kernels:
345 2.6.32-431.29.2.el6 (RHEL6.5)
346 2.6.32-504.30.3.el6 (RHEL6.6)
347 2.6.32-573.26.1.el6 (RHEL6.7)
348 2.6.32-642.15.1.el6 (RHEL6.8)
349 2.6.32-696.3.1.el6 (RHEL6.9)
350 3.10.0-514.21.1.el7 (RHEL7.3)
351 3.0.101-0.47.71 (SLES11 SP3)
352 3.0.101-100 (SLES11 SP4)
353 3.12.74-60.64.40 (SLES12 SP1)
354 4.4.59-92.17 (SLES12 SP2)
356 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
357 * Recommended ZFS / SPL version: 0.6.5.9
358 * Tested with ZFS / SPL version: 0.6.5.9
359 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
360 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
361 stack overflow. For more information, please refer to bugzilla 17630.
362 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
363 the re-exporting nodes to be mounted with "32bitapi" mount option
365 --------------------------------------------------------------------------------
367 12-07-2016 Intel Corporation
369 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
370 for currently supported client and server kernel versions.
371 * Server known to build on patched kernels:
372 2.6.32-431.29.2.el6 (RHEL6.5)
373 2.6.32-504.30.3.el6 (RHEL6.6)
374 2.6.32-573.26.1.el6 (RHEL6.7)
375 2.6.32-642.11.1.el6 (RHEL6.8)
376 3.10.0-514.el7 (RHEL7.3)
377 3.0.101-0.47.71 (SLES11 SP3)
378 3.0.101-80 (SLES11 SP4)
379 3.12.62-60.64.8 (SLES12 SP1)
380 vanilla linux 4.5.7 (ZFS only)
381 * Client known to build on unpatched kernels:
382 2.6.32-431.29.2.el6 (RHEL6.5)
383 2.6.32-504.30.3.el6 (RHEL6.6)
384 2.6.32-573.26.1.el6 (RHEL6.7)
385 2.6.32-642.11.1.el6 (RHEL6.8)
386 3.10.0-514.el7 (RHEL7.3)
387 3.0.101-0.47.71 (SLES11 SP3)
388 3.0.101-80 (SLES11 SP4)
389 3.12.62-60.64.8 (SLES12 SP1)
391 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
392 * Recommended ZFS / SPL version: 0.6.5.8
393 * Tested with ZFS / SPL version: 0.6.5.8
394 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
395 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
396 stack overflow. For more information, please refer to bugzilla 17630.
397 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
398 the re-exporting nodes to be mounted with "32bitapi" mount option
400 --------------------------------------------------------------------------------
402 02-29-2016 Intel Corporation
404 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
405 for currently supported client and server kernel versions.
406 * Server known to build on patched kernels:
407 2.6.32-431.29.2.el6 (RHEL6.5)
408 2.6.32-504.30.3.el6 (RHEL6.6)
409 2.6.32-573.12.1.el6 (RHEL6.7)
410 3.10.0-327.3.1.el7 (RHEL7.2)
411 3.0.101-0.47.71 (SLES11 SP3)
412 3.0.101-68 (SLES11 SP4)
413 vanilla linux 4.2.1 (ZFS only)
414 * Client known to build on unpatched kernels:
415 2.6.32-431.29.2.el6 (RHEL6.5)
416 2.6.32-504.30.3.el6 (RHEL6.6)
417 2.6.32-573.12.1.el6 (RHEL6.7)
418 3.10.0-327.3.1.el7 (RHEL7.2)
419 3.0.101-0.47.71 (SLES11 SP3)
420 3.0.101-68 (SLES11 SP4)
423 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
424 * Recommended ZFS / SPL version: 0.6.4.2
425 * Tested with ZFS / SPL version: 0.6.4.2
426 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
427 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
428 stack overflow. For more information, please refer to bugzilla 17630.
429 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
430 the re-exporting nodes to be mounted with "32bitapi" mount option
432 --------------------------------------------------------------------------------
434 03-10-2015 Intel Corporation
436 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
437 for currently supported client and server kernel versions.
438 * Server known to build on patched kernels:
439 2.6.32-431.29.2.el6 (RHEL6.5)
440 2.6.32-504.8.1.el6 (RHEL6.6)
441 3.0.101-0.46 (SLES11 SP3)
442 * Client known to build on unpatched kernels:
443 2.6.32-431.29.2.el6 (RHEL6.5)
444 2.6.32-504.8.1.el6 (RHEL6.6)
445 3.10.0-123.20.1.el7 (RHEL7)
446 3.0.101-0.46 (SLES11 SP3)
447 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
448 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
449 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
450 stack overflow. For more information, please refer to bugzilla 17630.
451 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
452 the re-exporting nodes to be mounted with "32bitapi" mount option
454 Severity : enhancement
456 Description: control OST-index in IDIF via ROCOMPAT flag.
457 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
458 last_rcvd file. For new formatted OST device, it will be auto set;
459 for the case of upgrading from old OST device, you can enable it
460 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
461 enabled, for new created OST-object, its IDIF-in-LMA will contain
462 the OST-index; for the existing OST-object, the OSD will convert
463 old format IDIF as new format IDIF with OST-index stored in the
464 LMA EA when accessing such OST-object or via OI scrub. Once such
465 flag is enabled, it cannot be reverted back, so the system cannot
466 be downgraded to the orignal incompatible version.
468 --------------------------------------------------------------------------------
470 07-30-2014 Intel Corporation
472 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
473 for currently supported client and server kernel versions.
474 * Server known to build on patched kernels:
475 2.6.32-431.20.3.el6 (RHEL6.5)
476 3.0.101-0.31 (SLES11 SP3)
477 * Client known to build on unpatched kernels:
478 2.6.32-431.20.3.el6 (RHEL6.5)
479 3.0.101-0.31 (SLES11 SP3)
482 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
483 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
484 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
485 stack overflow. For more information, please refer to bugzilla 17630.
486 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
487 the re-exporting nodes to be mounted with "32bitapi" mount option
489 --------------------------------------------------------------------------------
491 10-31-2013 Intel Corporation
493 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
494 * Server builds for kernels:
495 2.6.32-358.18.1.el6 (RHEL6.4)
496 * Client builds for unpatched kernels:
497 2.6.32-358.18.1.el6 (RHEL6.4)
498 3.0.92-0.8 (SLES11 SP3)
500 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
501 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
502 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
503 stack overflow. For more information, please refer to bugzilla 17630.
504 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
505 the re-exporting nodes to be mounted with "32bitapi" mount option
507 --------------------------------------------------------------------------------
509 05-30-2013 Intel Corporation
511 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
512 * Server builds for kernels:
513 2.6.32-358.6.2.el6 (RHEL6.4)
514 * Client builds for unpatched kernels:
515 2.6.32-358.6.2.el6 (RHEL6.4)
516 3.0.74-0.6.6 (SLES11 SP2)
518 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
519 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
520 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
521 stack overflow. For more information, please refer to bugzilla 17630.
522 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
523 the re-exporting nodes to be mounted with "32bitapi" mount option
525 --------------------------------------------------------------------------------
527 9-30-2012 Whamcloud, Inc.
529 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
530 * Server support for kernels:
531 2.6.32-279.5.1.el6 (RHEL6)
532 * Client support for unpatched kernels:
533 2.6.32-279.5.1.el6 (RHEL6)
535 2.6.18-238.19.1.el5 (RHEL5)
536 2.6.16.60-0.69.1 (SLES10)
537 * Recommended e2fsprogs version: 1.41.90.wc4
538 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
539 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
540 stack overflow. For more information, please refer to bugzilla 17630.
541 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
542 the re-exporting nodes to be mounted with "32bitapi" mount option
544 Severity : enhancement
546 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
548 Severity : enhancement
550 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
552 Severity : enhancement
554 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
556 Severity : enhancement
558 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
560 Severity : enhancement
562 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
564 --------------------------------------------------------------------------------
566 04-30-2012 Whamcloud, Inc.
568 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
569 * Server support for kernels:
570 2.6.32-220.4.2.el6 (RHEL6)
571 * Client support for unpatched kernels:
572 2.6.32-220.4.2.el6 (RHEL6)
574 2.6.18-238.19.1.el5 (RHEL5)
575 2.6.16.60-0.69.1 (SLES10)
576 * Recommended e2fsprogs version: 1.41.90.wc4
577 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
578 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
579 stack overflow. For more information, please refer to bugzilla 17630.
580 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
581 the re-exporting nodes to be mounted with "32bitapi" mount option
583 Severity : enhancement
585 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
587 Severity : enhancement
589 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
591 Severity : enhancement
593 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
595 Severity : enhancement
597 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
599 --------------------------------------------------------------------------------
601 09-30-2011 Whamcloud, Inc.
603 * Server support for kernels:
604 2.6.18-238.19.1.el5 (RHEL5)
605 2.6.32-131.6.1.el6 (RHEL6)
606 * Client support for unpatched kernels:
607 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
608 2.6.18-238.19.1.el5 (RHEL5)
609 2.6.32-131.6.1.el6 (RHEL6)
610 2.6.29.4-167.fc11 (FC11)
611 2.6.16.60-0.69.1 (SLES10)
613 * Recommended e2fsprogs version: 1.41.90.wc3
614 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
615 Lustre filesystem with 4K stack may cause a stack overflow. For more
616 information, please refer to bugzilla 17630.
617 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
618 mounted with "32bitapi" mount option
620 Severity : enhancement
621 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
623 Severity : enhancement
624 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
626 Severity : enhancement
627 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
630 Frequency : only with SLES10
632 Description: Use OFED "KMP" provided by Novell
633 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
634 Lustre is now built against this package. That means you need to
635 install the ofed-kmp package from Novell for the patchless client.
636 Note that the ofed-kmp that Novell ships may not exactly match the
637 kernel version but should still be compatible.
639 Severity : enhancement
641 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
642 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
644 --------------------------------------------------------------------------------
646 2010-07-15 Oracle, Inc.
648 * Server support for kernels:
649 2.6.18-164.11.1.el5 (RHEL 5)
650 2.6.18-164.11.1.0.1.el5 (OEL 5)
651 * Client support for unpatched kernels:
652 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
653 2.6.18-164.11.1.el5 (RHEL 5),
654 2.6.18-164.11.1.0.1.el5 (OEL 5)
655 2.6.16.60-0.42.8 (SLES 10),
657 2.6.29.4-167.fc11 (FC11)
658 * Recommended e2fsprogs version: 1.41.10-sun2
660 Severity : enhancement
662 Description: Distribute lfs_migrate script with lustre
663 Details : The lfs_migrate mentioned in the manual is now included in the
664 lustre tree in lustre/scripts and will distributed in lustre RPMs.
666 Severity : enhancement
668 Description: Failover nid list added to import proc file.
670 Severity : enhancement
672 Description: "som_preview" mount option required to use SOM
673 Details : The clients will need to use som_preview mount option
677 Description: Lustre does not do 1MB IOs to HW RAID
678 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
680 Severity : enhancement
682 Description: lfs quota output is not very convenient for awk/sed-parsing
683 Details : Some positions in lfs quota output table could be empty or
684 non-empty which made it hard to parse it with scripts, now
685 a dash is put instead of space where there is not supposed
690 Description: Update to OFED 1.5.1
691 Details : OFED 1.5.1 will be used in environments that don't use
692 vendor supplied OFED.
696 Description: Remove lnet. prefix from lctl params display
697 Details : Also renamed lnet.memused to lnet_memused to
702 Description: Use conf_param -d to remove permanant settings.
706 Description: Add list_param -R
707 Details : lctl list_param now supports recursive mode.
710 Bugzilla : 21829, 22312
711 Description: Fix llobdstat.
712 Details : 1. Now updating the write_bytes counter
713 2. Port script fixes from 1.8.
717 Description: decrease the usage of memory on clients.
718 Details : 1. On clients, recycle dentries and inodes unused.
719 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
724 Description: llobdstat fix and enhancement
725 Details : add a counter to set a limit to how many samples will be returned
726 fix a wildcard in the path to limit to obdfilter stats only
730 Description: lfs check only allowed for root
731 Details : Code cleanups and added a regresion test of non-root use of
736 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
737 Details : Fix hr->hr_index overload.
741 Description: Fix FIEMAP functionality.
745 Description: FC11 patchless client support added
749 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
750 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
754 Description: New acc-sm script and various test-framework enhancments.
758 Description: Previously landed patch was reverted.
762 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
766 Description: Add ldiskfs maxdirsize mount option.
770 Description: Fix mkfs.lustre support for 16TB LUNs.
774 Description: Print mmp_check_interval
775 Details : print mmp_check_interval and make it possible to abort
776 mount operation in case it takes too long.
780 Description: Add cascading_rw to lustre mpi tests.
784 Description: Store SOM data in EA.
785 Details : Also several SOM related cleanups.
789 Description: Add refcount to md_open_data
793 Description: Use refcount on lov_request_set to prevent memory corruption.
797 Description: Statahead: use parent dir's i_mutex for synchronization.
801 Description: loadgen improvements.
805 Description: Enable peer health detection in ksocklnd and ko2iblnd.
809 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
813 Description: Fix race on obd_devcie:md_stats
817 Description: Fix SLES 11 client builds.
818 Details : Resolved uninitalized pointer gcc warnings.
822 Description: Allow non-root access for "lfs check".
823 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
827 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
829 Severity : enhancement
831 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
832 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
836 Description: Sanity test_27p fails.
837 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
838 Sanity tests 27 were cleaned up and improved.
842 Description: Unify req format on client/servers
843 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
844 buffers and swabbers approach..
849 Description: drop join file stuff.
851 Severity : enhancement
853 Description: Add 2.6.27 and SLES11 (patchless client) support
857 Description: Fix in ptlrpc_expire_one_request() to print the signed time
860 Severity : enhancement
862 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
865 Severity : enhancement
867 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
869 Severity : enhancement
871 Description: correctly handle big reply message.
872 Details : send LNet event if reply is bigger than buffer and adjust this buffer
877 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
879 Severity : enhancement
881 Description: Add support for OEL5.
885 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
887 Severity : enhancement
889 Description: "weak-modules" support
890 Details : Implement "weak-modules" support which enables kernel modules
891 to be used with any kernel that implements the same kABI. In
892 order to achieve this modules are now installed in
893 /lib/modules/$(uname -r)/updates/kernel on all distributions.
897 Description: client could not reconnect to OST because of an active request.
898 Details : The client now retries to reconnect to the same server, if a
899 connect request failed with EBUSY or -EAGAIN. Test case 35b
900 conf-sanity.sh has been created.
904 Description: Don't allow make backward step on assignin osc next id.
905 Details : race between allocation next id and ll_sync thread can be cause
906 of set wrong osc next id and can be kill valid ost objects.
910 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
914 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
918 Description: Error handling in osc_statfs_interpret() has been improved.
919 Details : Check in osc_statfs_interpret() for EBADR.
923 Description: Do not send statfs() requests to OSTs disabled by administrator.
924 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
926 Severity : enhancement
928 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
930 Severity : enhancement
932 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
934 Severity : enhancement
936 Description: Enhancement for lfs(1) command to use numeric uid/gid.
940 Description: Improved cleanup procedure for conf-sanity test_46a.
944 Description: Do not start bulk transfer, if client tried to reconnect.
945 Details : Do not start bulk transfer, if client tried to reconnect,
946 but failed due to active rpcs. Check for exp_abort_active_req
947 has been implemented.
951 Description: Prevent inconsistences between linux and lustre mount structures.
952 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
953 Make the sleep interruptible.
955 Severity : enhancement
957 Description: provide server to client comms path
958 Details : server to client communications path via new LDLM_SET_INFO rpc
962 Description: truncate starts GFP_FS allocation under transaction causing deadlock
963 Details : ldiskfs_truncate calls grab_cache_page which may start page
964 allocation under an open transaction. This may lead to
965 calling prune_icache with consequent lustre reentrance.
969 Description: send correctly lsm on open replay
970 Details : MDS is trust to LSM size on replay open, but client can set wrong size
975 Description: enable client interface failover
976 Details : When a child reconnects from another NID, properly update export
977 nid hash position and ldlm reverse import.
979 Severity : enhancement
981 Description: Communicate OST degraded/readonly state via statfs to MDS
982 Details : Flags in the statfs returned from OSTs indicate whether the
983 OST is in a degraded RAID state, or if the filesystem has
984 turned read-only after a filesystem error is detected.
989 Description: don't panic if EPROTO was hit when reading symlink
990 Details : correctly handling request reference in error cases.
992 Severity : enhancement
994 Description: Add LustreNetLink, a kernel-userspace communcation path.
996 Severity : enhancement
998 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1003 Description: access to llog context before init.
1004 Details : move handling CATALOGS file at osc layer and forbid access to llog
1005 context before init.
1008 Frequency : rare, only when client is writing to OST on local node
1010 Description: Avoid deadlock for local client writes
1011 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1012 memory freeing context. This allows OST threads to set the
1013 PF_MEMALLOC flag on task structures in order to allocate memory
1014 from reserved pools and complete IO.
1015 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1016 so that the OST threads generate memory pressure and allow
1017 inactive pages to be reclaimed.
1019 Severity : enhancement
1021 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1026 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1027 Details : move ldlm namespace creation in setup phase to avoid grab
1028 _lprocfs_lock with cli_sem held.
1032 Description: Temporarily disable grant shrink.
1033 Details : Disable the feature for debugging.
1036 Frequency : only when formatting test filesystems
1038 Description: Unable to run several mkfs.lustre on loop devices at the same time
1039 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1040 formatting. The solution is to proper handle the error.
1042 Severity : enhancement
1044 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1046 Severity : enhancement
1048 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1053 Description: resolve race between obd_disconnect and class_disconnect_exports
1054 Details : if obd_disconnect will be called to already disconnected export he
1055 forget release one reference and osc module can't unloaded.
1057 Severity : enhancement
1059 Description: Allow tuning service thread via /proc
1060 Details : For each service a new
1061 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1062 created that can be used to set min/max thread counts, and get the
1063 current number of running threads.
1068 Descriptoin: don't return error if have partially created objects for file.
1069 Details : lov_update_create_set uses set->set_success as index for created
1070 objects, so if some requests will be failed, they will have hole
1071 at end of array and qos_shrink_lsm for allocate correct lsm.
1073 Severity : enhancement
1075 Description: Update OFED support to 1.4.1
1080 Description: Reduce small size read RPC
1081 Details : Set read-ahead limit for every file and only do read-ahead when
1082 available read-ahead pages are bigger than 1M to avoid small size
1085 Severity : enhancement
1087 Description: move AT tunable parameters for more consistent usage
1088 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1090 Severity : enhancement
1092 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1093 Details : allow skip disconnected ost for send statfs request and hide error
1099 Description: fix racy locking of mballoc block bitmaps causing BUG
1100 Details : The locking of the mballoc buddy bitmap and the in-memory
1101 block bitmap was using two different spin locks in some
1102 cases. This made it possible to incorrectly access the
1103 mballoc bitmap while another process was modifying it,
1104 causing a sanity assertion to fail. While no on-disk corruption
1105 was reported, there was some risk of this happening.
1108 Frequency : rare, on llog test 6
1110 Descriptoin: don't allow connect to already connected import
1111 Details : allowing connect to already connected import is hide connecting problem.
1116 Descriptoin: allow kill process which wait statahead result
1117 Details : for some reasons 'ls' can stick in waiting result from statahead,
1118 in this case need way for kill this process.
1120 Severity : enhancement
1122 Description: Add state history info file, enhance import info file
1123 Details : Track import connection state changes in a new osc/mdc proc file;
1124 add overview-type data to the osc/mdc import proc file.
1126 Severity : enhancement
1128 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1130 Severity : enhancement
1132 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1137 Descriptoin: OST grant too much space to client even there are not enough space.
1138 Details : Client will shrink its grant cache to OST if there are no write
1139 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1140 this grant cache if there are already not enough avaible space
1141 (left_space < total_clients * 32M).
1144 Frequency : start MDS on uncleanly shutdowned MDS device
1146 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1147 Details : stay in waiting mds<>ost recovery finished produce random bugs
1148 due race between two ll_sync thread for one lov target. send
1149 ACTIVATE event only if connect realy finished and import have
1153 Frequency : rare, connect and disconnect target at same time
1155 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1156 Details : don't call obd_disconnect under lov_lock. this long time
1157 operation and can block ptlrpcd which answer to connect request.
1162 Descriptoin: don't lose wakeup for imp_recovery_waitq
1163 Details : recover_import_no_retry or invalidate_import and import_close can
1164 both sleep on imp_recovery_waitq, but we was send only one wakeup
1168 Frequency : always with long access acl
1170 Descriptoin: mds can't pack reply with long acl.
1171 Details : mds don't control size of acl but they limited by reint/getattr
1174 Severity : enhancement
1176 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1178 Severity : enhancement
1180 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1183 Frequency : start MDS on uncleanly shutdowned MDS device
1185 Descriptoin: aborting recovery hang on MDS
1186 Details : don't throttle destroy RPCs for the MDT.
1189 Frequency : on remount
1191 Description: external journal device not working after the remount
1192 Details : clear dev_rdonly flag for external journal devices in
1198 Description: shutdown vs evict race
1199 Details : client_disconnect_export vs connect request race.
1200 if client will evicted at this time - we start invalidate
1201 thread without referece to import and import can be freed
1205 Frequency : rare, need acl's on inode.
1207 Description: client can't handle ost additional correctly
1208 Details : if ost was added after client connected to mds client can have
1209 hit lnet_try_match_md ... to big messages to wide striped files.
1210 in this case need teach client to handle config events about add
1211 lov target and update client max ea size at that event.
1213 Severity : enhancement
1215 Description: Changelogs
1216 Details : Changelogs are a lightweight mechanism to track filesystem
1217 metadata and namespace changes. The changelog is recorded
1218 permanently on the MDTs, and is periodically "consumed" / purged
1219 when records are no longer needed.
1221 Severity : enhancement
1223 Description: compact fld format with extents
1224 Details : Store range of seq rather than every seq in FLD. Seq
1225 controller update FLD rather than clients. In Case of CMD, mdt0
1226 has FLD, all other metadata server act as non persistent proxy
1227 for FLD queries and cache fld entries in fld cache.
1232 Description: don't skip ost target if they assigned to file
1233 Details : Drop slow OSCs if we can, but not for requested start idx.
1234 This means "if OSC is slow and it is not the requested
1235 start OST, then it can be skipped, otherwise skip it only
1236 if it is inactive/recovering/out-of-space.
1240 Description: more cleanup in mds_lov
1241 Details : not send LOV EA under replay, we can't know about they size at this
1242 time. Don't allow client connect to mds before any ost connected,
1243 for avoid problems with LOV EA size and returning EIO to client.
1245 Severity : enhancement
1247 Description: Interoperability at server side (Disk interoperability)
1249 Severity : enhancement
1251 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1253 Severity : enhancement
1255 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1257 Severity : enhancement
1259 Description: New client IO stack (CLIO).
1261 Severity : enhancement
1263 Description: Commit on sharing. Eliminate inter-client dependencies between
1264 uncommitted transactions by doing transaction commits.
1265 Thereby clients may recovery independently.
1268 Frequency : Create a symlink file with a very long name
1270 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1271 Details : If there is no extra space in the request for early cancels,
1272 ldlm_req_handles_avail() returns 0 instead of a negative value.
1274 Severity : enhancement
1276 Description: Add /proc entry for import status
1277 Details : The mdc, osc, and mgc import directories now have
1278 an import directory that contains useful import data for debugging
1279 connection problems.
1281 Severity : enhancement
1283 Description: Re-disable certain /proc logging
1284 Details : Enable and disable client's offset_stats, extents_stats and
1285 extents_stats_per_process stats logging on the fly.
1288 Frequency : Only on FC kernels 2.6.22+
1290 Description: oops in statahead
1291 Details : Do not drop reference count for the dentry from VFS when lookup,
1292 VFS will do that by itself.
1294 Severity : enhancement
1296 Description: Generic /proc file permissions
1297 Details : Set /Proc file permissions in a more generic way to enable non-
1298 root users operate on some /proc files.
1302 Description: Hitting mdc_commit_close() ASSERTION
1303 Details : Properly handle request reference release in
1304 ll_release_openhandle().
1308 Description: quota recovery deadlock during mds failover
1309 Details : This patch includes att18982, att18236, att18237 in bz14840.
1311 1. fix osts hang when mds does failover with quotaon
1312 2. prevent watchdog storm when osts threads wait for the
1317 Frequency : only patchless client
1318 Description: add workaround for race between add/remove dentry from hash
1320 Severity : enhancement
1322 Description: Allow OST glimpses to return PW locks
1326 Description: LBUG when llog conf file is full
1327 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1331 Description: Prevent import from entering FULL state when server in recovery
1335 Description: service mount cannot take device name with ":"
1336 Details : Only when device name contains ":/" will mount treat it as
1342 Description: replace ptlrpcd with the statahead thread to interpret the async
1343 statahead RPC callback
1347 Frequency : on recovery
1348 Description: I/O failures after umount during fail back
1349 Details : if client reconnected to restarted server we need join to recovery
1350 instead of find server handler is changed and process self
1351 eviction with cancel all locks.
1353 Severity : enhancement
1355 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1357 Severity : enhancement
1359 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1361 Severity : enhancement
1363 Description: Upcall on Lustre log has been dumped
1364 Details : Allow for a user mode script to be called once a Lustre log has
1365 been dumped. It passes the filename of the dumped log to the
1366 script, the location of the script can be specified via
1367 /proc/sys/lnet/debug_log_upcall.
1372 Description: avoid idr_remove called for id which is not allocated.
1373 Details : Move assigment s_dev for clustered nfs to end of initialization,
1374 to avoid problem with error handling.
1379 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1380 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1381 into CONN_USED_HASH and this prodice warning when put connection
1382 again in unused hash.
1385 Severity : enhancement
1387 Description: File striping can now be set to use an arbitrary pool of OSTs.
1389 Severity : enhancement
1391 Description: Export bytes_read/bytes_write count on OSC/OST.
1395 Description: Early reply size mismatch, MGC loses connection
1396 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1397 the connect flags are properly negotiated.
1402 Description: quotas are not honored with O_DIRECT
1403 Details : all writes with the flag O_DIRECT will use grants which leads to
1404 this problem. Now using OBD_BRW_SYNC to guard this.
1408 Description: add quota statistics
1409 Details : 1. sort out quota proc entries and proc code.
1410 2. add quota statistics
1412 Severity : enhancement
1414 Description: enable quota support for HEAD.
1418 Description: Properly propagate oinfo flags from lov to osc for statfs
1419 Details : restore missing copy oi_flags to lov requests.
1421 Severity : enhancement
1423 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1424 : l_getgroups(8), lst(8), routerstat(8)
1425 Details : included man pages for llobdstat(8), llstat(8),
1426 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1428 Severity : enhancement
1430 Description: configure's --enable-quota should check the
1431 : kernel .config for CONFIG_QUOTA
1432 Details : configure is terminated if --enable-quota is passed but
1433 : no quota support is in kernel
1437 Description: Remove portals compatibility
1438 Details : Remove portals compatibility, not interoperable with releases
1443 Description: Resolve device initialization race
1444 Details : Prevent proc handler from accessing devices added to the
1445 obd_devs array but yet be intialized.
1447 Severity : enhancement
1449 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1451 Severity : enhancement
1453 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1458 Description: Using wrong pointer in osc_brw_prep_request
1459 Details : Access to array[-1] can produce panic if kernel compiled with
1460 CONFIG_PAGE_ALLOC enabled
1464 Description: Client runs out of low memory
1465 Details : Consider only lowmem when counting initial number of llap pages
1469 Description: *optional* service tags registration
1470 Details : if the "service tags" package is installed on a Lustre node
1471 When the filesystem is mounted, a local-node service tag will
1472 be created. See http://inventory.sun.com/ for more information
1473 about the Service Tags asset management system.
1477 Description: Kernel BUG tries to release flock
1478 Details : Lustre does not destroy flock lock before last reference goes
1479 away. So always drop flock locks when client is evicted and
1480 perform unlock regardless of successfulness of speaking to MDS.
1484 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1488 Description: sanity test 65a fails if stripecount of -1 is set
1489 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1494 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1495 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1498 Severity : enhancement
1500 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1504 Description: do not process already freed flock
1505 Details : flock can possibly be freed by another thread before it reaches
1506 to ldlm_flock_completion_ast.
1510 Description: LBUG during stress test
1511 Details : Need properly lock accesses the flock deadlock detection list.
1515 Description: oops in page fault handler
1516 Details : kernel page fault handler can return two special 'pages' in error
1517 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1521 Description: timeout with invalidate import.
1522 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1523 should be handled by ptlrpcd. This produce long age waiting and
1524 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1526 Severity : enhancement
1528 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1532 Description: enable MGS and MDT services start separately
1533 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1534 a co-located MGS without starting the MGS, which is a complement
1535 to 'nosvc' mount option.
1539 Frequency : after recovery
1540 Description: precreate to many object's after del orphan.
1541 Details : del orphan st in oscc last_id == next_id and this triger growing
1542 count of precreated objects. Set flag LOW to skip increase count
1543 of precreated objects.
1547 Frequency : rare, on clear nid stats
1548 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1549 Details : when clean nid stats sometimes try destroy live entry,
1550 and this produce panic in free.
1554 Description: Stack overflow during MDS log replay
1555 ease stack pressure by using a thread dealing llog_process.
1559 Description: wait until IO finished before start new when do lock cancel.
1560 Details : VM protocol want old IO finished before start new, in this case
1561 need wait until PG_writeback is cleared until check dirty flag and
1562 call writepages in lock cancel callback.
1564 Severity : enhancement
1566 Description: using special macro for print time and cleanup in includes.
1570 Description: mds_mfd_close() ASSERTION(rc == 0)
1571 Details : In mds_mfd_close(), we need protect inode's writecount change
1572 within its orphan write semaphore to prevent possible races.
1576 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1580 Frequency : rare, on shutdown ost
1581 Description: don't hit live lock with umount ost.
1582 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1583 use shrink_dcache_sb instead.
1587 Description: don't panic with use echo client
1588 Details : echo client pass NULL as client nid pointer and this produce null
1589 pointer dereference.
1593 Description: fix build on ppc32
1594 Details : compile code with -m64 flag produce wrong object file for ppc32.
1598 Description: add message levels for liblustreapi
1602 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1603 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1604 will result in an -EAGAIN and in the exit call path we call
1605 alloc_rr() which will with increasing aggressiveness attempt to
1606 aquire precreated objects on the minimum number of required OSCs.
1610 Description: Use old size assignment to avoid deadlock
1611 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1612 the scheduling while holding a spinlock. We do not need locking
1613 for size in ll_update_inode() because size is only updated from
1614 the MDS for directories or files without objects, so there is no
1615 other place to do the update, and concurrent access to such inodes
1616 are protected by the inode lock.
1620 Description: resolve "_IOWR redefined" build error on SLES10
1624 Description: dump the memory debugging after all modules are unloaded to
1625 suppress false negative in conf_sanity test 39
1627 Severity : enhancement
1629 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1634 Description: fix SLES kernel versioning
1635 Details : the kernel version for our SLES 10 kernel did not include a "-"
1636 before the "smp" at the end. while this was not a problem in
1637 general, it did mean that software trying to use the kernel
1638 version to try to detect a vendor specific kernel would fail.
1639 this was most evident by the OFED build scripts.
1643 Description: Don't update lov_desc members until making sure they are valid
1644 Details : When updating lov_desc members via proc fs, need fix their
1645 validities before doing the real update.
1649 Description: don't put request into delay list while invalidate in flight.
1650 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1651 import in flight. this produce timeout for invalidate and sometimes
1652 can cause stale data.
1656 Frequency : on ppc only
1657 Description: not convert ost objects for directory because it's not exist.
1658 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1662 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1664 Severity : enhancement
1666 Description: Update to support 2.6.22.14 vanilla kernel.
1670 Frequency : rare, on recovery
1671 Description: read procfs can produce deadlock in some situation
1672 Details : Holding lprocfs lock which send rpc can produce block for destroy
1673 obd objects and this also block reconnect with -EALREADY. This
1674 isn't fix all lprocfs bugs - but make it rare.
1676 Severity : enhancement
1678 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1681 Frequency : frequent on X2 node
1683 Description: mdc_set_open_replay_data LBUG
1684 Details : Set replay data for requests that are eligible for replay.
1688 Description: lustre_mgs: operation 101 on unconnected MGS
1689 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1690 MGC, and late on MGC cannot successfully connect to MGS and a lot
1691 of the error messages complaining that MGS is not connected.
1695 Frequency : on network error
1696 Description: panic with double free request if network error
1697 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1698 but it's true only for synchronus enqueue, for async enqueue
1699 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1702 Severity : enhancement
1704 Description: client-side metadata stat-ahead during readdir(directory readahead)
1705 Details : perform client-side metadata stat-ahead when the client detects
1706 readdir and sequential stat of dir entries therein
1709 Frequency : on start mds
1711 Description: Implement get_info(last_id) in obdfilter.
1714 Frequency : occasional
1716 Description: Correctly check stale fid, not start epoch if ost not support SOM
1717 Details : open with flag O_CREATE need set old fid in op_fid3 because
1718 op_fid2 was overwritten with new generated fid, but mds can answer
1719 with one of these two fids and both is not stale. Setattr
1720 incorrectly started an epoch and assume will be called
1721 done_writing, but without SOM done_writing ever being called.
1724 Frequency : rare, depends on device drivers and load
1726 Description: MDS or OSS nodes crash due to stack overflow
1727 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1728 In some cases, in conjunction with device drivers that use a lot
1729 of stack the MDS (or possibly OSS) service threads could overflow
1730 the stack. One change which was identified to consume additional
1731 stack has been reworked to avoid the extra stack usage.
1734 Frequency : occasional
1736 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1737 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1738 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1739 oscc_has_objects and see OSC has no objests and call
1740 oscc_internal_create to resend create request.
1742 Severity : enhancement
1744 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1746 Severity : enhancement
1748 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1751 Frequency : very rare
1753 Description: Fix replay if there is an un-replied request and open
1754 Details : In some cases, older replay request will revert the
1755 mcd->mcd_last_xid on MDS which is used to record the client's
1756 latest sent request.
1758 Severity : enhancement
1760 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1762 Severity : enhancement
1764 Description: Add rhel5 support to HEAD.
1766 Serverity : enhancement
1768 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1773 Description: Don't allow skipping OSTs if index has been specified.
1774 Details : Don't allow skipping OSTs if index has been specified, make
1775 locking in internal create lots better.
1779 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1780 Details : access to bitfield in structure is always rounded to long
1781 and this produce problem with not atomic change any bit.
1785 Description: Lustre make rpms failed.
1786 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1787 builds Lustre rpms from tarball.
1790 Frequency : rare on shutdown ost
1792 Description: If llog cancel was not send before clean_exports phase, this can
1793 produce deadlock in llog code.
1794 Details : If llog thread has last reference to obd and call class_import_put
1795 this produce deadlock because llog_cleanup_commit_master wait when
1796 last llog_commit_thread exited, but this never success because was
1797 called from llog_commit_thread.
1801 Description: allow userland application know is lost one of stripes.
1802 Details : fill lvb_blocks with error code on ost and return it to
1803 application if error flag found.
1807 Description: NULL lov_tgts causing MDS oops
1808 Details : more safe checks for NULL lov_tgts for avoid oops.
1810 Severity : enhancement
1812 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1816 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1817 Details : If an OST is down the MDS will hang indefinitely in
1818 obd_statfs_rqset() waiting for the statfs data. While for
1819 MDS QOS usage of statfs, it should not stuck in waiting.
1821 Severity : enhancement
1823 Description: remote_acl support
1824 Details : Support ACL-based permission check for remote user.
1825 Support setfacl/getfacl for remote user with the utils
1826 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1827 the system "{s,g}etfacl" utils.
1829 Severity : enhancement
1831 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1833 Severity : enhancement
1835 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1839 Description: make mgs_setparam() handle fsname containing dash
1840 Details : fsname containing a dash does not work with lctl conf_param
1842 Severity : enhancement
1844 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1846 Severity : enhancement
1848 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1852 Frequency : rare, at shutdown
1853 Description: access already free / zero obd_namespace.
1854 Details : if client_disconnect_export was called without force flag set,
1855 and exist connect request in flight, this can produce access to
1856 NULL pointer (or already free pointer) when connect_interpret
1857 store ocd flags in obd_namespace.
1861 Frequency : only at startup
1862 Description: not alloc memory with spinlock held.
1863 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1864 if any spinlock held.
1866 Severity : enhancement
1868 Description: make lustre randomly fail allocating memory
1869 Details : Make lustre randomly failed allocating memory for testing purpose.
1871 Severity : enhancement
1873 Description: lost problems with lov objid file
1874 Details : Fixes some scability and access to not inited memory problems
1875 in work with lov objdid file.
1880 Description: lfs find does not continue on file error
1881 Details : Continue other files processing when a file/dir is absent.
1885 Description: Inconsistent usage of lustre_pack_reply()
1886 Details : Standardize the usage of lustre_pack_reply() such that it
1887 always generate a CERROR on failure.
1890 Frequency : occasional
1892 Description: MDS hang or stay in waiting lock
1893 Details : If client receive lock with CBPENDING flag ldlm need send lock
1894 cancel as separate rpc, to avoid situation when cancel request
1895 can't processed due all i/o threads stay in wait lock.
1899 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1903 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1907 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1911 Description: Update to RHEL4 latest kernel.
1913 Severity : enhancement
1915 Description: Build SLES10 patchless client fails
1916 Details : The configure was broken by run ./configure with
1917 --with-linux-obj=.... argument for patchless client. When the
1918 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1919 can't search header adequately. Use absolute path such as
1920 -I($LINUX)/include instead.
1924 Description: interrupt oig_wait produce painc on resend.
1925 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1926 private set, and this produce situation when rq_ptlrpcd_data not
1927 copyed to new allocated request and triggered LBUG on assert
1928 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1929 ptlrpcd set if request is changed and can be safety changed to use
1934 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1935 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1936 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1937 these macros and replace them with their difinitions instead.
1941 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1942 Details : Update which_patch & target file for SLES10 latest kernel.
1945 Description: Remove client patches from SLES 10 kernel.
1946 Details : This causes SLES 10 clients to behave as patchless clients
1947 even on a Lustre-patched (server) kernel.
1949 Severity : enhancement
1951 Description: self-adjustable client's lru lists
1952 Details : use adaptive algorithm for managing client cached locks lru
1953 lists according to current server load, other client's work
1954 pattern, memory activities, etc. Both, server and client
1955 side namespaces provide number of proc tunables for controlling
1958 Severity : enhancement
1960 Description: light-weight GSS support
1961 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1962 and reduce performance overhead.
1964 Severity : enhancement
1966 Description: Linux keyring support
1967 Details : Support using service of Linux keyring for Lustre GSS internal
1968 context refresh/cache mechanism.
1972 Description: Fix errors in lfs documentation
1973 Details : Fixes man pages
1977 Description: don't use GFP_* in generic Lustre code.
1978 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1982 Description: obdclass is limited by single OBD_ALLOC(idarray)
1983 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1987 Description: warnings with build patchless client with vanila 2.6.19 and up
1988 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1989 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1993 Frequency : Only for SLES
1994 Description: sanity_quota fail test_1
1995 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1996 file, which makes TSTID[2] inunique.
2000 Frequency : Only for SLES9
2001 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2002 Details : Update target/ChangeLog/which_patch .
2006 Frequency : Only for test_10 in sanity-quota.sh
2007 Description: a bug in quota test code
2008 Details : checking a test flag in a wrong place
2012 Frequency : Only for test_13 in sanity-quota.sh
2013 Description: a bug in quota test script
2014 Details : shouldn't check overfull allocated quota
2018 Description: jbd statistics
2019 Details : Port older jbd statistics patch for sles10
2021 Severity : enhancement
2023 Description: Add printing inode info into message about error in writepage.
2028 Description: reply_lock_interpret crash due to race with it and lock cancel.
2029 Details : Do not replay locks that are being cancelled. Do not reference
2030 locks by their address during replay, just by their handle.
2034 Frequency : When flocks are used.
2035 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2036 Details : Flock locks might destroy just granted lock if it could be merged
2037 with another existing flock, this is done in completion handler,
2038 so teach ldlm_cli_enquque_fini that this is a valid case for
2044 Description: Oops in read and write path when failing to allocate lock.
2045 Details : Check if lock allocation failed and return error back.
2049 Description: lstripe command fails for valid OST index
2050 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2051 instead of lov->desc.ld_active_tgt_count.
2055 Description: sanity.sh failed test 103
2056 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2061 Description: df doesn't work properly if diskfs blocksize != 4K
2062 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2066 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2067 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2069 Severity : enhancement
2071 Description: Adaptive timeouts
2072 Details : RPC timeouts adapt to changing server load and network
2073 conditions to reduce resend attempts and improve recovery time.
2077 Description: llapi_file_create() does not allow some changes
2078 Details : add llapi_file_open() that allows specifying the mode and
2079 open flags, and also returns an open file handle.
2083 Description: merge and cleanup kernel patches.
2084 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2088 Description: Add journal checksum support.(Kernel part)
2089 Details : The journal checksum feature adds two new flags i.e
2090 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2091 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2092 indicates that the commit block contains the checksum for
2093 the blocks described by the descriptor blocks. Now commit
2094 record can be sent to disk without waiting for descriptor
2095 blocks to be written to disk. This behavior is controlled
2096 using JBD2_FEATURE_ASYNC_COMMIT flag.
2100 Description: OSS needs mutliple precreate threads
2101 Details : Add ability to start more than one create thread per OSS.
2105 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2106 Details : execute lfs setstripe on client
2110 Description: mds_obd_create error creating tmp object
2111 Details : When the user sets quota on root, llog will be affected and can't
2112 create files and write files.
2115 Frequency : Always on ia64 patchless client, and possibly others.
2117 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2118 Details : This allows the patchless client to be loaded on architectures
2119 without this export.
2123 Description: RedHat Update kernel for RHEL5
2124 Details : Add the kernel config file for RHEL5.
2128 Description: RedHat Update kernel for RHEL5
2129 Details : Modify the kernel config file more closer RHEL5.
2133 Description: Build failure against Centos5 (RHEL5)
2134 Details : Define PAGE_SIZE when it isn't present.
2138 Description: test_7 on sanity-quota.sh can't be used on separate machines
2139 Details : add facet to handle it
2143 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2144 Details : with NFS, the anon dentry's parent was set to itself in
2145 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2146 corresponding dentry other than use rec->ur_name.
2148 Severity : enhancement
2150 Description: lfs setstripe enhancement
2151 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2155 Description: enable data checksumming by default
2156 Details : enable checksum by default, allow --disable-checksum
2157 configure option and "-o nochecksum" mount option. Checksums
2158 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2162 Description: lustre support for RHEL5
2163 Details : Add support for RHEL5.
2167 Description: Client eviction due to failover config
2168 Details : after a connection loss, the lustre client should attempt to
2169 reconnect to the last active server first before trying the
2170 other potential connections.
2174 Description: when mds and osts use different quota unit(32bit and 64bit),
2175 quota will be released repeatly.
2176 Details : void sending multiple quota reqs to mds, which will keep the
2177 status between the reqs.
2181 Description: osts not allocated evenly to files
2182 Details : change the condition to increase offset_idx
2185 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2187 Description: Data corruption for OSTs that are formatted larger than 2TB
2189 Details : When generating the bio request for lustre file writes the
2190 sector number would overflow a temporary variable before being
2191 used for the IO. The data reads correctly from Lustre (which
2192 will overflow in a similar manner) but other file data or
2193 filesystem metadata may be corrupted in some cases.
2197 Description: Tune the kernel for good SCSI performance.
2198 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2199 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2204 Description: rewrite ext2-derived code in obdclass/uuid.c
2205 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2206 from scratch preserving functionality.
2210 Description: "lfs find -obd UUID" prints directories
2211 Details : "lfs find -obd UUID" will return all directory names instead
2212 of just file names. It is incorrect because the directories
2213 do not reside on the OSTs.
2217 Description: use i_size_read and i_size_write in 2.6 port
2218 Details : replace inode->i_size access with i_size_read/write()
2221 Frequency : when removing large files
2223 Description: scheduling issue during removal of large Lustre files
2224 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2225 It causes scheduling issues when removing large files (17TB in the
2229 Frequency : only with liblustre clients on XT3
2231 Description: evictions taking too long
2232 Details : allow llrd to evict clients directly on OSTs
2235 Frequency : only on ppc
2237 Description: /proc/fs/lustre/devices broken on ppc
2238 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2239 We should make sure the type of 'index' is loff_t and then cast
2240 explicitly as needed below. Do not assign an explicitly cast
2244 Frequency : only for rhel5
2246 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2247 Details : Modify the target file & which_kernel.
2249 Severity : enhancement
2251 Description: omit set fsid for export NFS
2252 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2253 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2257 Frequency : Always, for kernels after 2.6.16
2258 Description: Fix warning idr_remove called for id=.. which is not allocated.
2259 Details : Last kernels save old s_dev before kill super and not allow
2260 to restore from callback - restore it before call kill_anon_super.
2264 Description: buffer overruns could theoretically occur
2265 Details : llapi_semantic_traverse() modifies the "path" argument by
2266 appending values to the end of the origin string, and a
2267 overrun may occur. Adding buffer overrun check in liblustreapi.
2271 Description: Fix error on 'ls .' at the top of the Lustre mount.
2272 Details : Don't revalidate dentry if it is a root dentry.
2276 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2277 Details : Modify vm-tunables-rhel4.patch.
2281 Description: Kernel config for 2.6.18-vanilla.
2282 Details : Modify targets/2.6-vanilla.target.in.
2283 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2284 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2285 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2286 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2290 Description: improve handling recoverable errors
2291 Details : if request processig with error which can be recoverable on server
2292 request should be resend, otherwise page released from cache and
2297 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2298 Details : Modify target file & which_patch.
2299 A flaw was found in the IA32 system call emulation provided
2300 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2301 value could be stored in the %RAX register, which could trigger an
2302 out-of-bounds system call table access. An untrusted local user
2303 could exploit this flaw to run code in the kernel
2304 (ie a root privilege escalation). (CVE-2007-4573).
2308 Description: O_DIRECT bypasses client statistics.
2309 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2310 read_ahead_stats were updated. copied the stats section from
2311 osc_send_oap_rpc() into async_internal().
2315 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2319 Description: change order of libsysio includes
2320 Details : '#include sysio.h' should always come before '#include xtio.h'
2324 Description: handle "serious error: objid * already exists" more gracefully
2325 Details : If LAST_ID value on disk is smaller than the objects existing in
2326 the O/0/d* directories, it indicates disk corruption and causes an
2327 LBUG(). If the object is 0-length, then we should use the existing
2328 object. This will help to avoid a full fsck in most cases.
2330 Severity : enhancement
2332 Description: adapt the lustre_config script to support the upgrade case
2333 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2334 server targets to 1.6 in parallel.
2338 Description: To avoid grant space > avaible space when the disk is almost
2339 full. Without this patch you might see the error "grant XXXX >
2340 available" or some LBUG about grant, when the disk is almost
2342 Details : In filter_check_grant, for non_grant cache write, we should
2343 check the left space by if (*left > ungranted + bytes), instead
2344 of (*left > ungranted), because only we are sure the left space
2345 is enough for another "bytes", then the ungrant space should be
2346 increase. In client, we should update cl_avail_grant only there
2347 is OBD_MD_FLGRANT in the reply.
2351 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2352 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2353 described in CVE-2007-4573. This problem could allow untrusted
2354 local users to gain root access.
2357 Frequency : when using O_DIRECT and quotas
2359 Description: Incorrect file ownership on O_DIRECT output files
2360 Details : block usage reported by 'lfs quota' does not take into account
2361 files that have been written with O_DIRECT.
2366 Description: touch file failed when fs is not full
2367 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2368 otherwise we can get ENOSP while fs is not full.
2372 Description: parallel lock callbacks
2373 Details : Instead of sending blocking and completion callbacks as separated
2374 requests, adding them to a set and sending in parallel.
2377 Frequency : only for Cray XT3
2378 Bugzilla : 12829/13455
2379 Description: Changing primary group doesn't change the group lustre assigns to
2381 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2382 overridden with the primary group provided by the group upcall,
2383 whereas the supplied fsgid can be trusted if it is in the list of
2384 supplementary groups returned by the group upcall.
2386 Severity : enhancement
2388 Description: Allow masking D_WARNING, D_ERROR messages from console
2389 Details : Console messages can now be disabled via lnet.printk.
2393 Description: User code with malformed file open parameter crashes client node
2394 Details : Before packing join_file req, all the related reference should be
2395 checked carefully in case some malformed flags cause fake
2396 join_file req on client.
2400 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2401 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2402 mode is equal to the granted mode. However, it should also check
2403 that LDLM_FL_CP_REQD is not set.
2407 Description: Heavy nfs access might result in deadlocks
2408 Details : After ELC code landed, it is now improper to enqueue any mds
2409 locks under och_sem, because enqueue might want to decide to
2410 cancel open locks for same inode we are holding och_sem for.
2414 Description: Client eviction while running blogbench
2415 Details : A lot of unlink operations with concurrent I/O can lead to a
2416 deadlock causing evictions. To address the problem, the number of
2417 oustanding OST_DESTROY requests is now throttled to
2418 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2419 callbacks are processed in priority.
2423 Description: enable ACLs on MDS by default
2424 Details : ACLs must be enabled on MDS by default.
2427 Frequency : PPC/PPC64 only
2429 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2430 Details : fix duplicated definitions between asm-ppc64/types.h and
2431 lustre_types.h on PPC.
2434 Frequency : PPC/PPC64 only
2436 Description: asm-ppc/segment.h does not exist
2437 Details : fix compile issue on PPC.
2441 Description: better handle error messages in extents code
2444 Frequency : RHEL4 only
2446 Description: mkfs is very slow on IA64/RHEL4
2447 Details : A performance regression has been discovered in the MPT Fusion
2448 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2449 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2450 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2452 Severity : enhancement
2454 Description: SNMP support enhancement
2455 Details : Adding total number of sampled request for an MDS node in snmp
2458 Severity : enhancement
2460 Description: Optimize ldlm waiting list processing for PR extent locks
2461 Details : When processing waiting list for read extent lock and meeting read
2462 lock that is same or wider to it that is not contended, skip
2463 processing rest of the list and immediatelly return current
2464 status of conflictness, since we are guaranteed there are no
2465 conflicting locks in the rest of the list.
2469 Description: Time out and refuse to reconnect
2470 Details : When the failover node is the primary node, it is possible
2471 to have two identical connections in imp_conn_list. We must
2472 compare not conn's pointers but NIDs, otherwise we can defeat
2473 connection throttling.
2477 Description: port llog fixes from b1_6 into HEAD
2478 Details : Port llog reference couting and some llog cleanups from b1_6
2479 (bug 10800) into HEAD, for protect from panic and access to already
2480 free llog structures.
2484 Description: Detect stride IO mode in read-ahead
2485 Details : When a client does stride read, read-ahead should detect that and
2486 read-ahead pages according to the detected stride pattern.
2490 Description: data checksumming impacts single node performance
2491 Details : add support for several checksum algorithm. Currently, only CRC32
2492 and Adler-32 are supported. The checksum type can be changed on
2493 the fly via /proc/fs/lustre/osc/*/checksum_type.
2497 Description: use adler32 for page checksums
2498 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2503 Description: build for x2 fails
2504 Details : fix compile issue on Cray systems.
2508 Description: Properly match for duplicate locks
2509 Details : Due to different lock order from skiplists code, we need to
2510 traverse entire list for now
2513 Frequency : only on PPC/SLES10
2515 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2516 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2517 be defined. Add a hack in mkfs_lustre.c to work around this compile
2522 Description: LASSERT on MDS when client holding flock lock dies
2523 Details : ldlm pool logic depends on number of granted locks equal to
2524 number of released locks which is not true for flock locks, so
2525 just exclude such locks from consideration.
2529 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2530 Details : Use fsfilt_sync() for both the whole filesystem sync and
2531 individual file sync to eliminate dangerous inode locking
2532 with I_LOCK that can lead to a deadlock.
2536 Description: performance in 1.6.3
2537 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2538 sure that 1MB requests are not fragmented by the block layer.
2540 Severity : enhancement
2542 Description: organize the server-side client stats on per-nid basis
2543 Details : Change the structure of stats under obdfilter and mds to
2551 The "uuid"s file would list the uuids of _active_ exports.
2552 And the clear entry is to clear all stats and stale nids.
2554 Severity : enhancement
2556 Description: eliminate client locks in face of contention
2557 Details : file contention detection and lockless i/o implementation
2558 for contended files.
2562 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2565 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2567 Description: MDS may lose file striping (and hence file data) in some cases
2568 Details : If there are additional extended attributes stored on the MDS,
2569 in particular ACLs, SELinux, or user attributes (if user_xattr
2570 is specified for the client mount options) then there is a risk
2571 of attribute loss. Additionally, the Lustre file striping
2572 needs to be larger than default (e.g. striped over all OSTs),
2573 and an additional attribute must be stored initially in the
2574 inode and then increase in size enough to be moved to the
2575 external attribute block (e.g. ACL growing in size) for file
2580 Description: skiplist implementation simplification
2581 Details : skiplists are used to group compatible locks on granted list
2582 that was implemented as tracking first and last lock of each
2583 lock group the patch changes that to using doubly linked lists
2587 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2588 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2589 because LASSERT on both the data supplied by a client, and the
2590 data on disk is dangerous and incorrect.
2592 Severity : enhancement
2594 Description: Slow truncate/writes to huge files at high offsets.
2595 Details : Directly associate cached pages to lock that protect those pages,
2596 this allows us to quickly find what pages to write and remove
2597 once lock callback is received.
2601 Description: more ldlm soft lockups
2602 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2603 starve other threads from the resource lock for a long time in
2604 case of long waiting queue, so change the debug level from
2605 D_OTHER to the less frequently used D_INFO.
2607 Severity : enhancement
2609 Description: add -gid, -group, -uid, -user options to lfs find
2613 Description: Hung threads in invalidate_inode_pages2_range
2614 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2615 once one is completed. As a result, some RPCs are stuck in the
2616 queue and are never sent.
2620 Description: filter threads hungs on waiting journal commit
2621 Details : Cleanup filter group llog code, then only filter group llog will
2622 be only created in the MDS/OST syncing process.
2626 Description: Procfs and llog threads access destoryed import sometimes.
2627 Details : Sync the import destoryed process with procfs and llog threads by
2628 the import refcount and semaphore.
2630 Severity : enhancement
2632 Description: openlock cache of b1_6 port to HEAD
2637 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2638 Details : If insertion of an extent fails, then discard the inode
2639 preallocation and free data blocks else it can lead to duplicate
2644 Description: don't always update ctime in ext3_xattr_set_handle()
2645 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2646 In some cases the ctime should not be updated, for example for
2647 2.0->1.8 compatibility it is necessary to delete an xattr and it
2648 should not update the ctime.
2652 Bugzilla : 15713/16362
2653 Description: Assertion in iopen_connect_dentry in 1.6.3
2654 Details : looking up an inode via iopen with the wrong generation number can
2655 populate the dcache with a disconneced dentry while the inode
2656 number is in the process of being reallocated. This causes an
2657 assertion failure in iopen since the inode's dentry list contains
2658 both a connected and disconnected dentry.
2662 Description: assertion failure in ldlm_handle2lock()
2663 Details : fix a race between class_handle_unhash() and class_handle2object()
2664 introduced in lustre 1.6.5 by bug 13622.
2669 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2670 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2675 Description: oops in sd_iostats_seq_show()
2676 Details : unloading/reloading the scsi low level driver triggers a kernel
2677 bug when trying to access the sd iostat file.
2682 Description: Kernel panics during QLogic driver reload
2683 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2684 patch, causing memory corruption.
2689 Description: journal_dev option does not work in b1_6
2690 Details : pass mount option during pre-mount.
2692 Severity : enhancement
2694 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2695 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2696 extent information of a file. It can be used to map logical blocks
2697 in a file to physical blocks in the block device.
2701 Description: LDLM soft lockups - improvement
2702 Details : It is be possible to send the lock handle along with each read
2703 or write request because the client is already doing a lock match
2704 itself so there isn't any reason the OST should have to re-do that
2708 Frequency : only on Cray X2
2710 Description: X2 build failures
2711 Details : fix build failures on Cray X2.
2715 Description: xid & resent requests
2716 Details : Initialize RPC XID from clock at startup (randomly if clock is
2719 Severity : enhancement
2721 Description: Add lustre_start utility to start or stop multiple Lustre servers
2726 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2727 Details : In case of memory pressure, list_del() can be called twice on
2728 req->rq_history_list, causing a kernel oops.
2732 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2733 Details : fix stack overflow in the distributed lock manager by defering
2734 export eviction after a failed AST to the elt thread instead of
2735 handling it in the dlm interpret routine.
2739 Description: Convert some comments to new format.
2740 Details : Update documenting comments to match doxygen conventions.
2744 Description: Grammar fixes.
2745 Details : A couple of trivial spelling fixes.
2749 Description: OSD_COUNTERS-mandatory
2750 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2751 to simplify the code.
2755 Description: simplify cmm_device freeing logic.
2756 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2760 Description: Add lockdep support to dt_object_operations locking interface.
2761 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2762 parameter indicating lock ordering. Update mdd code to use new
2767 Description: Introduce failloc constants for lockless IO tests.
2768 Details : Add two new failloc constants to test lockless IO. Only one of
2769 them in implemented---another is checked in yet to be landed
2774 Description: Add lockdep support for inode mutex.
2775 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2776 in the situations where Lustre has to lock more than one inode
2781 Description: Add optional invariants checking support.
2782 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2783 configure switch. This macro is to be used for consistency and
2784 sanity checks that are too expensive to be left in `production'
2789 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2790 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2795 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2796 Details : Use spin_lock_nested() in (the only) situation where more than
2797 one ldlm_lock is locked simultaneously. Also, fix possible
2798 dead-lock in ldlm_lock_change_resource() by enforcing particular
2803 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2804 Details : Instead of specifying each ldlm_lock call-back through separate
2805 parameter, wrap them into struct ldlm_callback_suite.
2809 Description: Kill join_lru obd method and its callers.
2810 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2811 in memory---a requirement implemented through ->o_join_lru() obd
2812 method in HEAD. Remove this method and its users.
2816 Description: Add asynchronous ldlm ENQUEUE completion handler.
2817 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2818 case that stock ldlm_completion_ast() cannot handle as it waits
2819 until lock is granted. Introduce new ldlm_completion_ast_async()
2820 for this. Also comment ldlm_completion_ast().
2824 Description: ldlm_error <-> errno conversion.
2825 Details : Add functions to map (rather arbitrary) between LDLM error codes
2826 and standard errno values. CLIO needs this to prevent LDLM specific
2827 constants from escaping ldlm and osc.
2831 Description: Kill unused ldlm_handle2lock_ns() function.
2832 Details : Kill unused ldlm_handle2lock_ns() function.
2836 Description: Add lu_ref support to ldlm_lock
2837 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2838 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2839 and ldlm_resource. LDLM interface has to be changed, because code that
2840 releases a reference on a lock, has to "know" what reference this is.
2841 In the most frequent case
2843 lock = ldlm_handle2lock(handle);
2845 LDLM_LOCK_PUT(lock);
2847 no changes are required. When any other reference (received _not_ from
2848 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2849 instead of LDLM_LOCK_PUT().
2851 Arguably, changes are pervasive, and interface requires some discipline
2852 for proper use. On the other hand, it was very instrumental in finding
2853 a few leaked lock references.
2857 Description: Add ldlm_lock_addref_try().
2858 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2859 attempts to addref a lock that might be being canceled
2864 Description: Add ldlm_weigh_callback().
2865 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2866 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2867 hard-coded `number of pages' logic.
2871 Description: Add lockdep annotations to llog code.
2872 Details : Use appropriately tagged _nested() locking calls in the places
2873 where llog takes more than one ->lgh_lock lock.
2877 Description: Add loi_kms_set().
2878 Details : Wrap kms updates into a helper function.
2882 Description: Constify instances of struct lsm_operations.
2883 Details : Constify instances of struct lsm_operations.
2887 Description: lu_conf support.
2888 Details : On a server, a file system object is uniquely identified
2889 by a fid, which is sufficient to locate and load all object
2890 state (inode). On a client, on the other hand, more data are
2891 necessary instantiate an object. Change lu_object_find() and
2892 friends to take additional `lu_conf' argument describing object.
2893 Typically this includes layout information.
2897 Description: lu_context fixes.
2898 Details : Introduce new lu_context functions that are needed on the client
2899 side, where some system threads (ptlrpcd) are shared by multiple
2900 modules, and so cannot be stopped during module shutdown.
2904 Description: Add start and stop methods to lu_device_type_operations.
2905 Details : Introduce two new methods in lu_device_type_operations, that are
2906 invoked when first instance of a given type is created and last one
2907 is destroyed respectively. This is need by CLIO.
2911 Description: Add lu_ref support to struct lu_device.
2912 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2913 track leaked references.
2917 Description: Introduce lu_kmem_descr.
2918 Details : lu_kmem_descr and its companion interface allow to create
2919 and destroy a number of kmem caches at once.
2923 Description: Fix lu_object finalization race.
2924 Details : Fix a race between lu_object_find() finding an object and its
2925 concurrent finalization. This race is (most likely) not possible
2926 on the server, but might happen on the client.
2930 Description: Introduce lu_ref interface.
2931 Details : lu_ref is a debugging module allowing to track references to
2932 a given object. It is quite cpu expensive, and has to be
2933 explicitly enabled with --enable-lu_ref. See usage description
2938 Description: Factor lu_site procfs stats into a separate function.
2939 Details : Separate lu_site stats printing code into a separate function
2940 that can be reused on a client.
2944 Description: Constify instances of struct {lu,dt,md}_device_operations.
2945 Details : Constify instances of struct {lu,dt,md}_device_operations.
2949 Description: Introduce struct md_site and move meta-data specific parts of
2950 struct lu_site here.
2951 Details : Move md-specific fields out of struct lu_site into special struct
2952 md_site, so that lu_site can be used on a client.
2956 Description: Kill mdd_lov_destroy().
2957 Details : Remove unused mdd code.
2961 Description: Add st_block checking to multistat.c.
2962 Details : Add st_block checking to multistat.c.
2966 Description: Add lu_ref support to struct obd_device.
2967 Details : Add lu_ref tracking to obd_device.
2971 Description: Kill obd_set_fail_loc().
2972 Details : Remove unused code.
2976 Description: Add special type for ptlrpc_request interpret functions.
2977 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2978 there. Actual usage will be in CLIO.
2982 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2983 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2984 doesn't work with lockdep.
2988 Description: Add rwv.c test program.
2989 Details : New testing program exercising readv(2) and writev(2) (Qian).
2993 Description: Add sendfile.c test program.
2994 Details : New testing program exercising sendfile(2) (Jay).
2998 Description: Ratelimit a message that can be very frequent.
2999 Details : Ratelimit a memory allocation failure message that can
3004 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3005 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3009 Description: Liblustre build fixes.
3010 Details : Liblustre build fixes.
3014 Description: libcfs: add cfs_{need,cond}_resched() interface.
3015 Details : libcfs: add cfs_{need,cond}_resched() definition and
3016 implementations for Linux, NT, and liblustre.
3018 Severity : enhancement
3020 Description: More exported tunables for mballoc
3021 Details : Add support for tunable preallocation window and new tunables for
3022 large/small requests
3026 Description: Detect corruption of block bitmap and checking for preallocations
3027 Details : Checks validity of on-disk block bitmap. Also it does better
3028 checking of number of applied preallocations. When corruption is
3029 found, it turns filesystem readonly to prevent further corruptions.
3033 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3034 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3035 window to 0. If it does hit cache miss, and read-pattern is still
3036 stride-io mode, does not reset the stride window, but also does not
3037 increase the stride window length in this case.
3041 Frequency : only for big-endian servers
3042 Description: Check if system is big-endian while mounting fs with extents feature
3043 Details : Mounting a filesystem with extents feature will fail on big-endian
3044 systems since ext3-based ldiskfs is not supported on big-endian
3045 systems. This can be over-riden with "bigendian_extents" mount option.
3047 Severity : enhancement
3049 Description: The root squash functionality
3050 Details : A security feature, which is to prevent users from being able
3051 to mount lustre on their desktop, run as root, and delete
3052 all of the files in the filesystem. The goal is accomplished by
3053 remapping user id (UID) and group id (GID) of the root user to
3054 a UID and GID specified by the system administartor via Lustre
3055 configuration management server (MGS). The functionality also
3056 allows to specify sets of clients for which the remapping does
3061 Description: Excessive recovery window
3062 Details : With AT enabled, the recovery window can be excessively long (6000+
3063 seconds). To address this problem, we no longer use
3064 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3065 timeout no longer depends on the service time, it is set to
3066 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3067 time via pb_service_time.
3071 Description: Watchdog triggered on MDS failover
3072 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3073 the OSTs know that the MDS "UUID" can be reused for the same export
3074 from a different NID, so we do not need to wait for the export to be
3078 Frequency : rare, only if using MMP with Linux RAID
3080 Description: MMP doesn't work with Linux RAID
3081 Details : While using HA for Lustre servers with Linux RAID, it is possible
3082 that MMP will not detect multiple mounts. To make this work we
3083 need to unplug the device queue in RAID when the MMP block is being
3084 written. Also while reading the MMP block, we should read it from
3085 disk and not the cached one.
3087 Severity : enhancement
3089 Description: open file using fid
3090 Details : A file can be opened using just its fid, like
3091 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3094 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3097 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3098 Details : add proper locking for creation of jbd_4k slab cache
3102 Description: MMP check in ext3_remount() fails without displaying any error
3103 Details : When multiple mount protection fails during remount, proper error
3106 Severity : enhancement
3108 Description: Allow stripe size to be up to 4G-64k
3109 Details : Fix math logic to allow large stripe sizes.
3113 Description: add check for >8TB ldiskfs filesystems
3114 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3115 Don't allow >8TB ldiskfs filesystems to be mounted without
3116 force_over_8tb mount option
3121 Descriptoin: OSS complains "too many blocks" and crashes
3122 Details : In filter_commitrw_write():
3123 1. lquota_chkquota();
3124 2. obdo_from_inode();
3125 3. lquota_pending_commit();
3126 "2." has changed uid and gid in this case, then caused
3127 lquota_chkquota() and lquota_pending_commit() to use
3128 different uid and gid. That is the root of the bug.
3132 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3133 Details : Detect and print out the 'stripe count = -1' case properly.
3137 Description: Out or order replies might be lost on replay
3138 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3139 request with tid smaller than one currently being added, add it
3140 to the start, not end of the list.
3142 --------------------------------------------------------------------------------
3144 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3146 * Support for kernels:
3147 2.6.5-7.283 (SLES 9),
3148 2.6.9-55.EL (RHEL 4),
3149 2.6.16.46-0.14 (SLES 10),
3150 2.6.18.8 vanilla (kernel.org)
3151 * Client support for unpatched kernels:
3152 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3153 2.6.16 - 2.6.22 vanilla (kernel.org)
3154 * Due to recently discovered recovery problems, we do not recommend
3155 using patchless RHEL 4 clients with this or any earlier release.
3156 * Recommended e2fsprogs version: 1.39.cfs8
3157 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3158 * Starting with this release, the ldiskfs backing filesystem required
3159 by Lustre is now in its own package, lustre-ldiskfs. This package
3160 should be installed. It is versioned separately from Lustre and
3161 may be released separately in future.
3165 Description: block reactivating mgc import until all deactivates complete
3166 Details : Fix race when failing back MDT/MGS to itself (testing)
3169 Frequency : at statup only
3171 Description: mds_lov_synchronize race leads to various problems
3172 Details : simultaneous MDT->OST connections at startup can cause the
3173 sync to abort, leaving the OSC in a bad state.
3175 Severity : enhancement
3177 Description: add optional extra BUILD_VERSION info
3178 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3179 to override the lustre version.
3182 Frequency : 2.6.18 servers only
3184 Description: ll_kern_mount() doesn't release the module reference
3185 Details : The ldiskfs module reference count never drops down to 0
3186 because ll_kern_mount() doesn't release the module reference.
3191 Description: server LBUG when using old ost_num_threads parameter
3192 Details : Accept the old ost_num_threads parameter but warn that it
3193 is deprecated, and fix an off-by-one error that caused an LBUG.
3198 Description: Transient SCSI error results in persistent IO issue
3199 Details : iobuf->dr_error is not reinitialized to 0 between two
3203 Frequency : sometimes when underlying device returns I/O errors
3205 Description: OSTs not going read-only during write failures
3206 Details : OSTs are not remounted read-only when the journal commit threads
3207 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3208 instead of the ext3 wrappers.
3212 Description: poor connect scaling with increasing client count
3213 Details : Don't run filter_grant_sanity_check for more than 100 exports
3214 to improve scaling for large numbers of clients.
3217 Frequency : SLES10 only
3219 Description: sanity-quota.sh quotacheck failed: rc = -22
3220 Details : Quotas cannot be enabled on SLES10.
3223 Frequency : liblustre clients only
3225 Description: getdirentries does not give error when run on compute nodes
3226 Details : getdirentries does not fail when the size specified as an argument
3227 is too small to contain at least one entry
3229 Severity : enhancement
3231 Description: Add LNET router traceability for debug purposes
3232 Details : If a checksum failure occurs with a router as part of the
3233 IO path, the NID of the last router that forwarded the bulk data
3234 is printed so it can be identified.
3239 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3240 Details : Due to a race condition, liblustre clients were occasionally
3241 evicted incorrectly.
3243 Severity : enhancement
3245 Description: lfs setstripe use optional parameters instead of postional
3248 Severity : enhancement
3250 Description: Nanosecond timestamp support for ldiskfs
3251 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3252 resolution timestamps. There is not yet support for this at
3253 the Lustre filesystem level.
3256 Frequency : during server recovery
3258 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3259 Details : request with rq_no_resend flag not awake l_wait_event if they get
3263 Frequency : nfs export on patchless client
3265 Description: connectathon hang when test nfs export over patchless client
3266 Details : Disconnected dentry cannot be found with lookup, so we do not need
3267 to unhash it or make it invalid
3270 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3271 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3272 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3277 Description: open req refcounting wrong on reconnect
3278 Details : If reconnect happened between getting open reply from server and
3279 call to mdc_set_replay_data in ll_file_open, we will schedule
3280 replay for unreferenced request that we are about to free.
3281 Subsequent close will crash in variety of ways.
3282 Check that request is still eligible for replay in
3283 mdc_set_replay_data().
3288 Description: disable writes to filesystem when reading health_check file
3289 Details : the default for reading the health_check proc file has changed
3290 to NOT do a journal transaction and write to disk, because this
3291 can cause reads of the /proc file to hang and block HA state
3292 checking on a healthy but otherwise heavily loaded system. It
3293 is possible to return to the previous behaviour during configure
3294 with --enable-health-write.
3296 Severity : enhancement
3298 Description: 64-bit inode version
3299 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3300 to the inode. This will be required for version-based recovery.
3305 Description: MDS fails to start if a duplicate client export is detected
3306 Details : in some rare cases it was possible for a client to connect to
3307 an MDS multiple times. Upon recovery the MDS would detect this
3308 and fail during startup. Handle this more gracefully.
3310 Severity : enhancement
3312 Description: Add -o localflock option to simulate old noflock
3314 Details : This will achieve local-only flock/fcntl locks
3320 Description: log_commit_thread vs filter_destroy race leads to crash
3321 Details : Take import reference before releasing llog record semaphore
3326 Description: Wrong request locking in request set processing
3327 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3328 imp_delayed_list, in this place should be used imp_lock.
3331 Frequency : when reconnection
3333 Description: Grant Leak when osc reconnect to OST
3334 Details : When osc reconnect ost, OST(filter) should check whether it
3335 should grant more space to client by comparing fed_grant and
3336 cl_avail_grant, and return the granted space to client instead
3337 of "new granted" space, because client will call osc_init_grant
3338 to update the client grant space info.
3341 Frequency : when client reconnect to OST
3343 Description: Grant Leak when osc do resend and replay bulk write
3344 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3345 bulk write request, because the grant info while be sync between
3346 OSC and OST when reconnect, and we should ignore the grant info
3347 these of resend/replay write req.
3352 Description: Grant space more than available left space sometimes.
3353 Details : When then OST is about to be full, if two bulk writing from
3354 different clients came to OST. Accord the avaliable space of the
3355 OST, the first req should be permitted, and the second one
3356 should be denied by ENOSPC. But if the seconde arrived before
3357 the first one is committed. The OST might wrongly permit second
3358 writing, which will cause grant space > available space.
3361 Frequency : when client is evicted
3363 Description: Grant might be wrongly erased when osc is evicted by OST
3364 Details : when the import is evicted by server, it will fork another
3365 thread ptlrpc_invalidate_import_thread to invalidate the
3366 import, where the grant will be set to 0. While the original
3367 thread will update the grant it got when connecting. So if
3368 the former happened latter, the grant will be wrongly errased
3369 because of this race.
3374 Description: Checking Stale with correct fid
3375 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3376 to check whether it is stale, because sometimes, we want the
3377 enqueue happened anyway, and op_data.fid2 will not be initialized.
3379 Severity : enhancement
3381 Description: update patchless client
3382 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3385 Frequency : only with 2.4 kernel
3387 Description: random memory corruption
3388 Details : size of struct ll_inode_info is to big for union inode.u and this
3389 can be cause of random memory corruption.
3394 Description: Memory leak in recovery
3395 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3396 It should also check obd_fail before fsfilt_start, otherwise if
3397 fsfilt_start return -EROFS,(failover mds during mds recovery).
3398 then the req will return with repmsg->transno = 0 and rc = EROFS.
3399 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3400 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3401 be freed no matter whether fsfilt_commit success or not.
3404 Frequency : only with huge count clients
3406 Description: Prevents from taking the superblock lock in llap_from_page for
3408 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3409 for llap_from_page call in ll_removepage prevents from taking the
3410 superblock lock for a soon died page.
3415 Description: Not check open intent error before release open handle
3416 Details : in some rare cases, the open intent error is not checked before
3417 release open handle, which may cause
3418 ASSERTION(open_req->rq_transno != 0), because it tries to release
3419 the failed open handle.
3424 Description: Set cat log bitmap only after create log success.
3425 Details : in some rare cases, the cat log bitmap is set too early. and it
3426 should be set only after create log success.
3430 Description: Accessing a block bevice can re-enable I/O when Lustre is
3431 tearing down a device.
3432 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3437 Description: service threads may hog cpus when there are a lot of requests
3438 Details : Insert cond_resched to give other threads a chance to use some CPU
3443 Description: the cat log was not initialized in recovery
3444 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3445 unlink log on mds will not be initialized until mds post
3446 recovery. And also in mds post recovery, the unlink log will
3447 initialization will be done asynchronausly, so there will be race
3448 between add unlink log and unlink log initialization.
3452 Description: brw_stats were being printed incorrectly
3453 Details : brw_stats were being printed as log2 but all of them were not
3454 recorded as log2. Also remove some code duplication arising from
3455 filter_tally_{read,write}.
3459 Frequency : rare, only in recovery.
3460 Description: ASSERTION(req->rq_type != LI_POISON) failed
3461 Details : imp_lock should be held while iterating over imp_sending_list for
3462 prevent destroy request after get timeout in ptlrpc_queue_wait.
3466 Description: replay-single.sh test 52 fails
3467 Details : A lock's skiplist need to be cleanup when it being unlinked
3468 from its resource list.
3472 Description: Short directio read returns full requested size rather than
3474 Details : Direct I/O operations should return actual amount of bytes
3475 transferred rather than requested size.
3477 Severity : enhancement
3479 Description: metadata RPC reduction (e.g. for rm performance)
3480 Details : decrease the amount of synchronous RPC between clients and servers
3481 by canceling conflicing lock before the operation on the client
3482 and packing thier handles into the main operation RPC to server.
3484 Severity : enhancement
3486 Description: Async OSC create to avoid the blocking unnecessarily.
3487 Details : If an OST has no remaining object, system will block on the
3488 creation when it needs to create a new object on this OST. Now,
3489 ways use pre-created objects when available, instead of blocking on
3490 an empty osc while others are not empty. If we must block, we
3491 block for the shortest possible period of time.
3495 Description: improve handling recoverable errors
3496 Details : if request processig with error which can be recoverable on server
3497 request should be resend, otherwise page released from cache and
3500 Severity : enhancement
3502 Description: refine locking for avoid write wrong info into lov_objid file
3503 Details : fix possible races with add new target and write/update data in
3506 --------------------------------------------------------------------------------
3508 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3513 Frequency : on some architectures
3515 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3516 Details : Uninitialized flags sometimes cause configuration commands to
3520 Frequency : patchless clients only
3522 Description: missing __iget() symbol export
3523 Details : The __iget() symbol export is missing. To avoid the need for
3524 this on patchless clients the deathrow inode reaper is turned
3525 off, and we depend on the VM to clean up old inodes. This
3526 dependency was during via the fix for bug 12181.
3528 --------------------------------------------------------------------------------
3530 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3532 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3533 INTEROPERATE with older versions automatically. In many cases a
3534 special upgrade step is needed. Please read the
3535 user documentation before upgrading any part of a live system.
3536 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3537 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3538 * WARNING: Lustre configuration and startup changes are required with
3539 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3543 Severity : enhancement
3545 Description: MountConf
3546 Details : Lustre configuration is now managed via mkfs and mount
3547 commands instead of lmc and lconf. New obd types (MGS, MGC)
3548 are added for dynamic configuration management. See
3549 https://mail.clusterfs.com/wikis/lustre/MountConf for
3552 Severity : enhancement
3554 Description: dynamic OST addition
3555 Details : OSTs can now be added to a live filesystem
3557 Severity : enhancement
3559 Description: startup order invariance
3560 Details : MDTs and OSTs can be started in any order. Clients only
3561 require the MDT to complete startup.
3563 Severity : enhancement
3565 Description: parallel, asynchronous orphan cleanup
3566 Details : orphan cleanup is now performed in separate threads for each
3567 OST, allowing parallel non-blocking operation.
3569 Severity : enhancement
3571 Description: optimized stripe assignment
3572 Details : stripe assignments are now made based on ost space available,
3573 ost previous usage, and OSS previous usage, in order to try
3574 to optimize storage space and networking resources.
3576 Severity : enhancement
3578 Description: Permanently set tunables
3579 Details : All writable /proc/fs/lustre tunables can now be permanently
3580 set on a per-server basis, at mkfs time or on a live system.
3582 Severity : enhancement
3584 Description: Lustre message v2
3585 Details : Add lustre message format v2.
3587 Severity : enhancement
3589 Description: client OST exclusion list
3590 Details : Clients can be started with a list of OSTs that should be
3591 declared "inactive" for known non-responsive OSTs.
3595 Description: SPEC SFS validation failure on NFS v2 over lustre.
3596 Details : Changes the blocksize for regular files to be 2x RPC size,
3597 and not depend on stripe size.
3599 Severity : enhancement
3601 Description: Multiple MD RPCs in flight.
3602 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3603 To avoid overly-overloading MDS, introduce a limit on number of
3604 MDS RPCs in flight for a single client and add /proc controls
3605 to adjust this limit.
3607 Severity : enhancement
3609 Description: client read/write statistics
3610 Details : Add client read/write call usage stats for performance
3611 analysis of user processes.
3612 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3613 file access. extents_stats shows chunk size distribution.
3614 extents_stats_per_process show chunk size distribution per
3617 Severity : enhancement
3619 Description: mds statistics
3620 Details : Add detailed mds operations statistics in
3621 /proc/fs/lustre/mds/*/stats.
3625 Description: Failure of copying files with lustre special EAs.
3626 Details : Client side always return success for setxattr call for lustre
3627 special xattr (currently only "trusted.lov").
3632 Description: Refcount LNET uuids
3633 Details : The global LNET uuid list grew linearly with every startup;
3634 refcount repeated list entries instead of always adding to
3637 Severity : enhancement
3639 Description: Dynamic service threads
3640 Details : Within a small range, start extra service threads
3641 automatically when the request queue builds up.
3644 Frequency : mixed-endian client/server environments
3646 Description: mixed-endian crashes
3647 Details : The new msg_v2 system had some failures in mixed-endian
3650 Severity : enhancement
3652 Description: Easy OST removal
3653 Details : OSTs can be permanently deactivated with e.g. 'lctl
3654 conf_param lustre-OST0001.osc.active=0'
3656 Severity : enhancement
3658 Description: MGS proc entries
3659 Details : Added basic proc entries for the MGS showing what filesystems
3662 Severity : enhancement
3664 Description: provide MGS failover
3665 Details : Added config lock reacquisition after MGS server failover.
3667 Severity : enhancement
3669 Description: add Linux 2.4 support
3670 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3674 Description: a large application tries to do I/O to the same resource and dies
3675 in the middle of it.
3676 Details : Check the req->rq_arrival time after the call to
3677 ost_brw_lock_get(), but before we do anything about
3678 processing it & sending the BULK transfer request. This
3679 should help move old stale pending locks off the queue as
3680 quickly as obd_timeout.
3683 Frequency : when an incorrect nid is specified during startup
3685 Description: ptlrpc connect to non-existant node causes kernel crash
3686 Details : LNET can't be re-entered from an event callback, which
3687 happened when we expire a message after the export has been
3688 cleaned up. Instead, hand the zombie cleanup off to another
3691 Severity : enhancement
3693 Description: plain/inodebits lock performance improvement
3694 Details : Grouping plain/inodebits in granted list by their request modes
3695 and bits policy, thus improving the performance of search through
3699 Frequency : only if OST filesystem is corrupted
3701 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3702 Details : for a short time RPCs with bulk IO are in the replay list,
3703 but replay of bulk IOs is unimplemented. If the OST filesystem
3704 is corrupted due to disk cache incoherency and then replay is
3705 started it is possible to trip an assertion. Avoid putting
3706 committed RPCs into the replay list at all to avoid this issue.
3709 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3711 Description: System hang on startup
3712 Details : This bug allowed the liblustre (e.g. catamount) client to
3713 return to the app before handling all startup RPCs. This
3714 could leave the node unresponsive to lustre network traffic
3715 and manifested as a server ptllnd timeout.
3717 Severity : enhancement
3719 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3720 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3721 boolean to control whether to print peer debug info when a
3722 client's RPC times out.
3725 Frequency : only for kernels with patches from Lustre below 1.4.3
3727 Description: Remove old rdonly API
3728 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3731 Frequency : only for devices with external journals
3733 Description: Set external device read-only also
3734 Details : During a commanded failover stop, we set the disk device
3735 read-only while the server shuts down. We now also set any
3736 external journal device read-only at the same time.
3739 Frequency : when upgrading from 1.4 while trying to change parameters
3741 Description: The wrong (new) MDC name was used when setting parameters for
3742 upgraded MDT's. Also allows changing of OSC (and MDC)
3743 parameters if --writeconf is specified at tunefs upgrade time.
3746 Frequency : when setting specific ost indicies
3748 Description: QOS code breaks on skipped indicies
3749 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3750 created with --index need not be sequential.
3752 Severity : enhancement
3754 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3755 Details : The uninit_groups feature works in conjunction with the kernel
3756 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3757 up the pass1 processing of e2fsck. This is a read-only feature
3758 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3759 mount filesystems that have had this feature enabled.
3761 Severity : enhancement
3763 Description: Improve multi-block allocation algorithm to avoid fragmentation
3764 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3765 allocation locality and avoid filesystem fragmentation.
3767 ------------------------------------------------------------------------------
3769 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3771 * Support for kernels:
3772 2.4.21-47.0.1.EL (RHEL 3)
3773 2.6.5-7.283 (SLES 9)
3774 2.6.9-42.0.10.EL (RHEL 4)
3775 2.6.12.6 vanilla (kernel.org)
3776 2.6.16.27-0.9 (SLES 10)
3777 * Recommended e2fsprogs version: 1.39.cfs5
3779 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3783 Frequency : occasional, depends on client load and configuration
3784 Bugzilla : 12181, 12203
3785 Description: data loss for recently-modified files
3787 Details : In some cases it is possible that recently written or created
3788 files may not be written to disk in a timely manner (this should
3789 normally be within 30s unless client IO load is very high).
3790 The problem appears as zero-length files or files that are a
3791 multiple of 1MB in size after a client crash or client eviction
3792 that are missing data at the end of the file.
3794 This problem is more likely to be hit on clients where files are
3795 repeatedly created and unlinked in the same directory, clients
3796 have a large amount of RAM, have many CPUs, the filesystem has
3797 many OSTs, the clients are rebooted frequently, and/or the files
3798 are not accessed by other nodes after being written.
3800 The presence of the problem can be detected by looking at
3801 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3802 smaller than the second (nr_unused) then dirty files will not
3803 be flushed automatically to disk. "sync; sleep 10" should be
3804 run several times on the node before unmounting it to update
3805 Lustre (this is also safe to run on nodes without this problem).
3807 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3808 that can cause this same problem, so customers using that kernel
3809 also need to update the kernel in addition to Lustre. In order
3810 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3812 It is normal that files written just before a client crash (less
3813 than 30s) may not yet have been flushed to disk, even for local
3817 Frequency : frequent on thin XT3 nodes
3819 Description: UUID collision on thin XT3 Linux nodes
3820 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3821 randomly, since we relied on an insufficiently-seeded PRNG.
3826 Description: OSS hangs after "All ost request buffers busy"
3827 Details : A deadlock between quota and journal operations caused OSS
3828 hangs after printing "All ost request buffers busy."
3831 Frequency : always on liblustre builds
3833 Description: Cleanup compiler warnings on liblustre
3836 Frequency : always on liblustre builds on XT3
3838 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3842 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3843 > 32000 subdirectories
3844 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3845 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3846 subdirectory count crosses 32000. This will aid e2fsck to
3847 correctly handle more than 32000 subdirectories.
3850 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3852 Description: System hang on startup
3853 Details : This bug allowed the liblustre (e.g. catamount) client to
3854 return to the app before handling all startup RPCs. This
3855 could leave the node unresponsive to lustre network traffic
3856 and manifested as a server ptllnd timeout.
3858 Severity : enhancement
3860 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3861 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3862 boolean to control whether to print peer debug info when a
3863 client's RPC times out.
3868 Description: make O_SYNC working on 2.6 kernels
3869 Details : 2.6 kernels use different method for mark pages for write,
3870 so need add a code to lustre for O_SYNC work.
3875 Description: Failure to close file and release space on NFS
3876 Details : Put inode details into lock acquired in ll_intent_file_open.
3877 Use mdc_intent_lock in ll_intent_open to properly
3878 detect all kind of errors unhandled by mdc_enqueue.
3883 Description: proc file read during shutdown sometimes raced obd removal,
3885 Details : Add lock to prevent obd access after proc file removal.
3888 Frequency : Only for files larger than 4GB on 32-bit clients.
3890 Description: improperly doing page alignment of locks
3891 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3892 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3897 Description: RPCs being resent when they shouldn't be
3898 Details : Some RPCs that should not be resent are being resent. This
3899 can cause inconsistencies in the RPC state machine. Do not
3900 resend such requests.
3903 Frequency : rare, only with NFS export
3905 Description: Crash on NFS re-export node
3906 Details : under very unusual load conditions an assertion is hit in
3907 ll_intent_file_open()
3910 Frequency : only if OST filesystem is corrupted
3912 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3913 Details : for a short time RPCs with bulk IO are in the replay list,
3914 but replay of bulk IOs is unimplemented. If the OST filesystem
3915 is corrupted due to disk cache incoherency and then replay is
3916 started it is possible to trip an assertion. Avoid putting
3917 committed RPCs into the replay list at all to avoid this issue.
3922 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3923 Details : Large single O_DIRECT read and write calls can fail to allocate
3924 a sufficiently large buffer to process the request. In case of
3925 allocation failure the allocation is retried with a smaller
3926 buffer and broken into smaller requests.
3928 Severity : enhancement
3930 Description: Add -o localflock option to simulate old noflock behaviour.
3931 Details : This will achieve local-only flock/fcntl locks coherentness.
3936 Description: versioning check is incomplete
3937 Details : Checking the version difference of client vs. server, report
3938 error if the gap is too big.
3943 Description: add support PG_writeback bit
3944 Details : add support for PG_writeback bit for Lustre, for more carefull
3945 work with page cache in 2.6 kernel. This also fix some deadlocks
3946 and remove hack for work O_SYNC with 2.6 kernel.
3948 Severity : enhancement
3950 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3951 Details : The uninit_groups feature works in conjunction with the kernel
3952 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3953 up the pass1 processing of e2fsck. This is a read-only feature
3954 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3955 mount filesystems that have had this feature enabled.
3957 Severity : enhancement
3959 Description: Improve multi-block allocation algorithm to avoid fragmentation
3960 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3961 allocation locality and avoid filesystem fragmentation.
3963 ------------------------------------------------------------------------------
3965 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3967 * Support for kernels:
3968 2.6.9-42.0.3.EL (RHEL 4)
3969 2.6.5-7.276 (SLES 9)
3970 2.4.21-47.0.1.EL (RHEL 3)
3971 2.6.12.6 vanilla (kernel.org)
3972 2.6.16.21-0.8 (SLES10)
3973 * Recommended e2fsprogs version: 1.39.cfs2-0
3975 * The backwards-compatible /proc/sys/portals symlink has been removed
3976 in this release. Before upgrading, please ensure that you change
3977 any configuration scripts or /etc/sysctl.conf files that access
3978 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3979 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3980 in advance of the upgrade on any system running Lustre 1.4.6 or
3981 newer, since /proc/sys/lnet was added in that version.
3982 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3986 Frequency : only when quota is used
3988 Description: avoid scanning export list for quota master
3989 Details : Change the algorithms to avoid scanning export list in order
3990 to improve the efficiency.
3993 Frequency : MDS failover only, very rarely
3995 Description: "went back in time" messages on mds failover
3996 Details : The greatest transno may be lost when the current operation
3997 finishes with an error (transno==0) and the client's last_rcvd
3998 record is over-written. Save the greatest transno in the
3999 mds_last_transno for this case.
4002 Frequency : always for specific kernels and striping counts
4004 Description: client may get "Matching packet too big" without ACL support
4005 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4006 when trying to access files in certain configurations. The
4007 clients should in fact be denied when mounting because they do
4008 not understand ACLs.
4011 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4013 Description: many clients connecting with IO in progress causes connect timeouts
4014 Details : Avoid synchronous journal commits to avoid delays caused by many
4015 clients connecting/disconnecting when bulk IO is in progress.
4016 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4017 OST_IO_PORTAL to avoid delays behind potentially many pending
4021 Frequency : occasionally with multiple writers to a single file
4023 Description: shared writes to file may result in wrong size reported by stat()
4024 Details : Allow growing of kms when extent lock is cancelled
4027 Frequency : always with random mmap IO to multi-striped file
4029 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4030 Details : Only if the hole is at the end of OST object so that kms is too
4031 small. Fix is to increase kms accordingly in ll_nopage.
4034 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4036 Description: writes to a missing object would leak memory on the OST
4037 Details : If there is an inconsistency between the MDS and OST filesystems,
4038 such that the MDS references an object that doesn't exist, writes
4039 to that object will leak memory due to incorrect cleanup in the
4040 error handling path, eventually running out of memory on the OST.
4045 Description: Creating too long symlink causes lustre errors
4046 Details : Check symlink and name lengths before sending requests to MDS.
4049 Frequency : only if flock is enabled (not on by default)
4051 Description: posix locks not released on fd closure on 2.6.9+
4052 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4053 kernels, this caused such locks not to be released on fd close and
4054 then assertions on fs unmount about still used locks.
4057 Frequency : MDS failover only, very rarely
4059 Description: clients may get ASSERTION(granted_lock != NULL)
4060 Details : When request was taking a long time, and a client was resending
4061 a getattr by name lock request. The were multiple lock requests
4062 with the same client lock handle and
4063 mds_getattr_name->fixup_handle_for_resent_request found one of the
4064 lock handles but later failed with ASSERTION(granted_lock != NULL).
4069 Description: handle->h_buffer_credits > 0, assertion failure
4070 Details : h_buffer_credits is zero after truncate, causing assertion
4071 failure. This patch extends the transaction or creates a new
4075 Frequency : NFS re-export or patchless client
4076 Bugzilla : 11179, 10796
4077 Description: Crash on NFS re-export node (__d_move)
4078 Details : We do not want to hash the dentry if we don't have a lock.
4079 But if this dentry is later used in d_move, we'd hit uninitialised
4080 list head d_hash, so we just do this to init d_hash field but
4081 leave dentry unhashed.
4084 Frequency : NFS re-export or patchless client
4086 Description: NFS exports has problem with symbolic link
4087 Details : lustre client didn't properly install dentry when re-exported
4088 to NFS or running patchless client.
4091 Frequency : NFS re-export or patchless client
4093 Description: Various nfs/patchless fixes.
4094 Details : fixes reuse disconected alias for lookup process - this fixes
4095 warning "find_exported_dentry: npd != pd",
4096 fix permission error with open files at nfs.
4097 fix apply umask when do revalidate.
4100 Frequency : occasional
4102 Description: Crash on NFS re-export node
4103 Details : calling clear_page() on the wrong pointer triggered oops in
4104 generic_mapping_read().
4107 Frequency : rarely, using O_DIRECT IO
4109 Description: unaligned directio crashes client with LASSERT
4110 Details : check for unaligned buffers before trying any requests.
4113 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4115 Description: stale data returned from RAID cache
4116 Details : If only a small amount of IO is done to the RAID device before
4117 reading it again it is possible to get stale data from the RAID
4118 cache instead of reading it from disk.
4121 Frequency : always for sles10 kernel
4123 Description: sles10 support
4124 Details : ll_follow_link: compile fixes and using of nd_set_link
4125 under newer kernels.
4128 Frequency : depends on arch, kernel and compiler version, always on sles10
4131 Description: recursive or deep enough symlinks cause stack overflow
4132 Details : getting rid of large stack-allocated variable in
4136 Frequency : depends on hardware
4138 Description: lustre write performance loss in the SLES10 kernel
4139 Details : the performance loss is caused by using of write barriers in the
4140 ext3 code. The SLES10 kernel turns barrier support on by
4141 default. The fix is to undo that change for ldiskfs.
4143 ------------------------------------------------------------------------------
4145 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4147 * Support for kernels:
4148 2.6.9-42.0.3EL (RHEL 4)
4149 2.6.5-7.276 (SLES 9)
4150 2.4.21-47.0.1.EL (RHEL 3)
4151 2.6.12.6 vanilla (kernel.org)
4155 Frequency : quota enabled and large files being deleted
4157 Description: releasing more than 4GB of quota at once hangs OST
4158 Details : If a user deletes more than 4GB of files on a single OST it
4159 will cause the OST to spin in an infinite loop. Release
4160 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4165 Description: statfs data retrieved from /proc may be stale or zero
4166 Details : When reading per-device statfs data from /proc, in the
4167 {kbytes,files}_{total,free,avail} files, it may appear
4168 as zero or be out of date.
4171 Frequency : systems with MD RAID1 external journal devices
4173 Description: lconf's call to blkid is confused by RAID1 journal devices
4174 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4175 of returning all block devices that match the journal UUID.
4178 Frequency : always, for aggregate stripe size over 4GB
4180 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4181 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4182 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4185 Frequency : always if "lfs find" used on a local file/directory
4187 Description: "lfs find" segfaults if used on a local file/directory
4188 Details : The case where a directory component was not specified wasn't
4189 handled correctly. Handle this properly.
4192 Frequency : always on ppc64
4194 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4195 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4196 perform endianness conversion.
4199 Frequency : rarely (truncate to non-zero file size after write under load)
4200 Bugzilla : 10730, 10687
4201 Description: Files padded with zeros to next 4K multiple
4202 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4203 it is possible that files that are truncated to a non-zero size
4204 immediately after being written are filled with zero bytes beyond
4205 the truncated size. No file data is lost.
4207 Severity : enhancement
4209 Description: Allow recovery/failover for liblustre clients.
4210 Details : liblustre clients were unaware of failover configurations until
4213 Severity : enhancement
4215 Description: user file locks should fail when not mounting with flock option
4216 Details : Set up an error-returning stub in ll_file_operations.lock field
4217 to prevent incorrect behaviour when client is mounted without
4218 flock option. Also, set up properly f_op->flock field for
4222 Frequency : always on ia64
4224 Description: "lfs df" loops on printing out MDS statfs information
4225 Details : The obd_ioctl_data was not initialized and in some systems
4226 this caused a failure during the ioctl that did not return
4227 an error. Initialize the struct and return an error on failure.
4230 Frequency : SLES 9 only
4232 Description: Error of copying files with lustre special EAs as root
4233 Details : Client side always return success for setxattr call for lustre
4234 special xattr (currently only "trusted.lov").
4237 Frequency : rarely on clusters with both ia64+i386 clients
4239 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4240 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4241 clients are doing concurrent writes to the same file it is
4242 possible that the ia64 clients may LASSERT because the OST
4243 extent locks are not PAGE_SIZE aligned. Ensure that grown
4244 locks are always aligned on the request boundary.
4247 Frequency : specific use, occasional
4249 Description: Overwriting in use executable truncates on-disk binary image
4250 Details : If one node attempts to overwrite an executable in use by
4251 another node, we now correctly return ETXTBSY instead of
4252 truncating the file.
4254 Severity : enhancement
4256 Description: Async OSC create to avoid the blocking unnecessarily.
4257 Details : If an OST has no remaining object, system will block on the creating
4258 when need to create a new object on this OST. Now, ways use
4259 pre-created objects when available, instead of blocking on an
4260 empty osc while others are not empty. If we must block, we block
4261 for the shortest possible period of time.
4266 Description: chmod on Lustre root is propagated to other clients
4267 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4268 invalid by the follow_mount time.
4271 Frequency : liblustre clients only
4273 Description: Race in 'instant cancel' lock handling could lead to such locks
4274 never to be granted in case of SMP MDS
4275 Details : Do not destroy not yet granted but cbpending locks in
4279 Frequency : replay/resend of open
4281 Description: non null lock assetion failure in mds_intent_policy
4282 Details : Trying to replay/resend lockless open requests resulted in
4283 mds_open() returning 0 with no lock. Now it sets a flag if
4284 a lock is going to be returned.
4286 Severity : enhancement
4288 Description: Checksum enhancements
4289 Details : New checksum enhancements allow for resending RPCs that failed
4292 Severity : enhancement
4294 Description: Tunables on number of dirty pages in cacche
4295 Details : Allow to set limit on number of dirty pages cached.
4300 Description: client crash on unmount - lock still has references
4301 Details : In some error handling cases it was possible to leak a lock
4302 reference on a client while accessing a file. This was not
4303 harmful to the client during operation, but would cause the
4304 client to crash when the filesystem is unmounted.
4307 Frequency : specific case, rare
4309 Description: ETXTBSY on mds though file not in use by client
4310 Details : ETXTBSY is no longer incorrectly returned when attempting to
4311 chmod or chown a directory that the user previously tried to
4312 execute or a currently-executing binary.
4315 Frequency : extremely rare except on liblustre-based clients
4317 Description: Lustre space not freed when files are deleted
4318 Details : Clean up open-unlinked files after client eviction. Previously
4319 the unlink was skipped and the files remained as orphans.
4324 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4325 Details : In some cases it was possible to send a blocking callback to
4326 a client doing a glimpse, even though that client didn't get
4327 a lock granted. When the glimpse lock is cancelled on the OST
4328 the freed lock is left on the waiting list and corrupted the list.
4331 Frequency : all core dumps
4333 Description: Broke core dumps to lustre
4334 Details : Negative dentry may be unhashed if parent does not have UPDATE
4335 lock, but some callers, e.g. do_coredump, expect dentry to be
4336 hashed after successful create, hash it in ll_create_it.
4338 ------------------------------------------------------------------------------
4340 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4342 * Support for kernels:
4343 2.6.9-42.0.2.EL (RHEL 4)
4344 2.6.5-7.276 (SLES 9)
4345 2.4.21-40.EL (RHEL 3)
4346 2.6.12.6 vanilla (kernel.org)
4350 Frequency : always on RHEL 3
4352 Description: Number of open files grows over time
4353 Details : The number of open files grows over time, whether or not
4354 Lustre is started. This was due to a filp leak introduced
4355 by one of our kernel patches.
4357 ------------------------------------------------------------------------------
4359 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4361 * Support for kernels:
4362 2.6.9-42.EL (RHEL 4)
4363 2.6.5-7.276 (SLES 9)
4364 2.4.21-40.EL (RHEL 3)
4365 2.6.12.6 vanilla (kernel.org)
4370 Bugzilla : 5719, 9635, 9792, 9684
4371 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4372 Details : If a server is under heavy load and cannot reply to new
4373 connection requests before the client resends the (re)connect,
4374 the connection handling code can behave badly if two service
4375 threads are concurrently handing separate (re)connections from
4376 the same client. Add better locking to the connection handling
4377 code, and ensure that only a single connection will be processed
4378 for a given client UUID, even if the lock is dropped.
4380 Severity : enhancement
4382 Description: add TCP zero-copy support to kernel
4383 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4384 sends if the hardware supports scatter-gather and checksumming.
4385 This allows socklnd to do client-write and server-read more
4386 efficiently and reduce CPU utilization from skbuf copying.
4389 Frequency : only if NFS exporting from client
4391 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4392 Details : If setting attributes on a file created under NFS that had
4393 never been opened it would be possible to oops the client
4394 if the file had no objects.
4397 Frequency : always for liblustre
4399 Description: liblustre client does MDS+OSTs setattr RPC for each write
4400 Details : When doing a write from a liblustre client, the client
4401 incorrectly issued an RPC to the MDS and each OST the file was
4402 striped over in order to update the timestamps. When writing
4403 with small chunks and many clients this could overwhelm the MDS
4404 with RPCs. In all cases it would slow down the write because
4405 these RPCs are unnecessary.
4407 Severity : enhancement
4409 Description: allow number of MDS service threads to be changed at module load
4410 Details : It is now possible to change the number of MDS service threads
4411 running. Adding "options mds mds_num_threads={N}" to the MDS's
4412 /etc/modprobe.conf will set the number of threads for the next
4413 time Lustre is restarted (assuming the "mds" module is also
4414 reloaded at that time). The default number of threads will
4415 stay the same, 32 for most systems.
4420 Description: OST crash if filesystem is unformatted or corrupt
4421 Details : If an OST is started on a device that has never been formatted
4422 or if the filesystem is corrupt and cannot even mount then the
4423 error handling cleanup routines would dereference a NULL pointer.
4428 Description: NULL pointer deref in llap_from_page.
4429 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4430 incorrect) mapping if the page was truncated/reclaimed while it was
4431 searched for. Check for this condition and skip such pages when
4432 doing readahead. Introduce extra check to llap_from_page() to
4433 verify page->mapping->host is non-NULL (so page is not anonymous).
4436 Frequency : Sometimes when using sys_sendfile
4438 Description: "page not covered by a lock" warnings from ll_readpage
4439 Details : sendfile called ll_readpage without right page locks present.
4440 Now we introduced ll_file_sendfile that does necessary locking
4441 around call to generic_file_sendfile() much like we do in
4445 Frequency : with certain MDS communication failures at client mount time
4447 Description: NULL pointer deref after failed client mount
4448 Details : a client connection request may delayed by the network layer
4449 and not be sent until after the PTLRPC layer has timed out the
4450 request. If the client fails the mount immediately it will try
4451 to clean up before the network times out the request. Add a
4452 reference from the request import to the obd device and delay
4453 the cleanup until the network drops the request.
4456 Frequency : occasionally during client (re)connect
4458 Description: assertion failure during client (re)connect
4459 Details : processing a client connection request may be delayed by the
4460 client or server longer than the client connect timeout. This
4461 causes the client to resend the connection request. If the
4462 original connection request is replied in this interval, the
4463 client may trip an assertion failure in ptlrpc_connect_interpret()
4464 which thought it would be the only running connect process.
4467 Frequency : only with obd_echo servers and clients that are rebooted
4469 Description: kernel BUG accessing uninitialized data structure
4470 Details : When running an obd_echo server it did not start the ping_evictor
4471 thread, and when a client was evicted an uninitialized data
4472 structure was accessed. Start the ping_evictor in the RPC
4473 service startup instead of the OBD startup.
4475 Severity : enhancement
4476 Bugzilla : 10193 (patchless)
4477 Description: Remove dependency on various unexported kernel interfaces.
4478 Details : No longer need reparent_to_init, exit_mm, exit_files,
4479 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4482 Frequency : rare (only users of deprecated and unsupported LDAP config)
4484 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4485 Details : LDAP apparently contains 'lustreName' attributes instead of
4486 'name'. A simple remapping of the name is sufficient.
4489 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4491 Description: waiting_locks_callback trips kernel BUG if client is evicted
4492 Details : Running with the dump_on_timeout debug flag turned on makes
4493 it possible that the waiting_locks_callback() can try to dump
4494 the Lustre kernel debug logs from an interrupt handler. Defer
4495 this log dumping to the expired_lock_main() thread.
4497 Severity : enhancement
4499 Description: Support NFS exporting on 2.6 kernels.
4500 Details : Implement non-rawops metadata methods for NFS server to use without
4501 changing NFS server code.
4504 Frequency : very rare (synthetic metadata workload only)
4506 Description: two racing renames might cause an MDS thread to deadlock
4507 Details : Running the "racer" program may cause one MDS thread to rename
4508 a file from being the source of a rename to being the target of
4509 a rename at exactly the same time that another thread is doing
4510 so, and the second thread has already enqueued these locks after
4511 doing a lookup of the target and is trying to relock them in
4512 order. Ensure that we don't try to re-lock the same resource.
4515 Frequency : only very large systems with liblustre clients
4517 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4518 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4519 all OSTs in parallel. This allows the eviction of stale liblustre
4520 clients to proceed much faster than if they were done in series,
4521 and also offers similar improvements for other set_info RPCs.
4526 Description: excessive CPU usage during initial read phase on client
4527 Details : During the initial read phase on a client, it would agressively
4528 retry readahead on the file, consuming too much CPU and impacting
4529 performance (since 1.4.5.8). Improve the readahead algorithm
4530 to avoid this, and also improve some other common cases (read
4531 of small files in particular, where "small" is files smaller than
4532 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4537 Description: MDS crash when receiving packet with unknown intent.
4538 Details : Do not LBUG in unknown intent case, just return -EFAULT
4540 Severity : enhancement
4541 Bugzilla : 9293, 9385
4542 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4543 Details : Do not serialize getattr (non-intent version) and statfs.
4546 Frequency : occasional, when OST network is overloaded/intermittent
4548 Description: client evicted by OST after bulk IO timeout
4549 Details : If a client sends a bulk IO request (read or write) the OST
4550 may evict the client if it is unresposive to its data GET/PUT
4551 request. This is incorrect if the network is overloaded (takes
4552 too long to transfer the RPC data) or dropped the OST GET/PUT
4553 request. There is no need to evict the client at all, since
4554 the pinger and/or lock callbacks will handle this, and the
4555 client can restart the bulk request.
4558 Frequency : Always when mmapping file with no objects
4560 Description: client crashes when mmapping file with no objects
4561 Details : Check that we actually have objects in a file before doing any
4562 operations on objects in ll_vm_open, ll_vm_close and
4568 Description: Request leak when working with deleted CWD
4569 Details : Introduce advanced request refcount tracking for requests
4570 referenced from lustre intent.
4572 Severity : Enhancement
4574 Description: Cache open file handles on client.
4575 Details : MDS now will return special lock along with openhandle, if
4576 requested and client is allowed to hold openhandle, even if unused,
4577 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4578 closing files for every read/write openration.
4580 Severity : Enhancement
4582 Description: Cache open negative dentries on client when possible.
4583 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4584 negative dentries on lock revocation.
4589 Description: Remounting a client read-only wasn't possible with a zconf mount
4590 Details : It wasn't possible to remount a client read-only with llmount.
4592 Severity : enhancement
4593 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4594 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4595 MPICH is included to add Lustre-specific ADIO interfaces.
4596 This is based closely on the UFS ADIO layer and only differs
4597 in file creation, in order to allow the OST striping to be set.
4598 This is user-contributed code and not supported by CFS.
4603 Description: extended inode attributes (immutable, append-only) work improperly
4604 when 2.4 and 2.6 kernels are used on client/server or vice versa
4605 Details : Introduce kernel-independent values for these flags.
4607 Severity : enhancement
4610 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4611 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4612 granularity. Now, Fractional MB granularity is be supported,
4613 this is very useful for low memory system.
4615 Severity : enhancement
4617 Description: Getattr by fid
4618 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4619 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4620 have locks and avoids getattr rpc requests for it on every stat.
4623 Frequency : Always, for filesystems larger than 2TB
4625 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4626 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4627 filesystems larger than 2TB to be mounted with mballoc enabled.
4630 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4632 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4633 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4634 filesystems larger than 2TB is mounted then the kernel will
4635 silently corrupt the start of the filesystem. CONFIG_LBD is
4636 enabled for all CFS-supported kernels, but the possibility of
4637 this happening with a modified kernel config exists.
4639 Severity : enhancement
4641 Description: add client O_DIRECT support for 2.6 kernels
4642 Details : It is now possible to do O_DIRECT reads and writes to files
4643 in the Lustre client mountpoint on 2.6 kernel clients.
4645 Severity : enhancement
4647 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4648 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4649 parallel, not waiting for response from every OST before sending
4650 a rpc to the next OST.
4655 Description: setattr vs write race when updating file timestamps
4656 Details : Client processes that update a file timestamp into the past
4657 right after writing to the file (e.g. tar) it is possible that
4658 the updated file modification time can be reset to the current
4659 time due to a race between processing the setattr and write RPC.
4661 Severity : enhancement
4663 Description: Bring 'lfs find' closer in line with regular Linux find.
4664 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4665 -print0 options and obtains all the needed info through the lustre
4668 Severity : enhancement
4670 Description: support up to 1024 configured devices on one node
4671 Details : change obd_dev array from statically allocated to dynamically
4672 allocated structs as they are first used to reduce memory usage
4677 Description: Flush dirty partially truncated pages during truncate
4678 Details : Immediatelly flush partially truncated pages in filter_setattr,
4679 this way we completely avoid having any pages in page cache on OST
4680 and can retire ugly workarounds during writes to flush such pages.
4685 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4686 Details : protect inode from truncation within vfs_unlink() context
4687 just take a reference before calling vfs_unlink() and release it
4688 when parent's i_sem is free.
4693 Description: last_id value checked outside lock on OST caused LASSERT failure
4694 Details : If there were multiple MDS->OST object precreate requests in
4695 flight, it was possible that the OST's last object id was checked
4696 outside a lock and incorrectly tripped an assertion. Move checks
4697 inside locks, and discard old precreate requests.
4700 Frequency : always, if extents are used on OSTs
4702 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4703 Details : OSTs using the extents format would not zero the high 16 bits of
4704 the index physical block number. This is not a problem for any
4705 OST filesystems smaller than 16TB, and no kernels support ext3
4706 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4707 new/modified files) and can be fixed for existing filesystems
4708 with e2fsprogs-1.39-cfs1.
4713 Description: import connection selection may be incorrect if timer wraps
4714 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4715 import connections to be ignored if the 32-bit jiffies counter
4716 wraps. Use a 64-bit jiffies counter.
4719 Frequency : very large clusters immediately after boot
4721 Description: LNET request buffers exhausted under heavy short-term load
4722 Details : If a large number of client requests are generated on a service
4723 that has previously never seen so many requests it is possible
4724 that the request buffer growth cannot keep up with the spike in
4725 demand. Instead of dropping incoming requests, they are held in
4726 the LND until the RPC service can accept more requests.
4729 Frequency : Sometimes during replay
4731 Description: Assertion failure in ll_local_open after replay.
4732 Details : If replay happened on an open request reply before we were able
4733 to set replay handler, reply will become not swabbed tripping the
4734 assertion in ll_local_open. Now we set the handler right after
4735 recognising of open request
4738 Frequency : very rare
4740 Description: kernel reports "badness in vsnprintf"
4741 Details : Reading from the "recovery_status" /proc file in small chunks
4742 may cause a negative length in lprocfs_obd_rd_recovery_status()
4743 call to vsnprintf() (which is otherwise harmless). Exit early
4744 if there is no more space in the output buffer.
4746 Severity : enhancement
4748 Description: clear OBD RPC statistics by writing to them
4749 Details : It is now possible to clear the OBD RPC statistics by writing
4750 to the "stats" file.
4755 Description: Client mtime is not the same on different clients after utimes
4756 Details : In some cases, the client was using the utimes() syscall on
4757 a file cached on another node. The clients now validate the
4758 ctime from the MDS + OSTs to determine which one is right.
4763 Description: Inability to activate failout mode
4764 Details : lconf script incorrectly assumed that in pythong string's numeric
4765 value is used in comparisons.
4768 Frequency : always with multiple stripes per file
4770 Description: Inefficient object allocation for mutli-stripe files
4771 Details : When selecting which OSTs to stripe files over, for files with
4772 a stripe count that divides evenly into the number of OSTs,
4773 the MDS is always picking the same starting OST for each file.
4774 Return the OST selection heuristic to the original design.
4779 Description: mount failures may take full timeout to return an error
4780 Details : Under some heavy load conditions it is possible that a
4781 failed mount can wait for the full obd_timeout interval,
4782 possibly several minutes, before reporting an error.
4783 Instead return an error as soon as the status is known.
4785 Frequency : quota enabled and large files being deleted
4787 Description: releasing more than 4GB of quota at once hangs OST
4788 Details : If a user deletes more than 4GB of files on a single OST it
4789 will cause the OST to spin in an infinite loop. Release
4790 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4795 Description: statfs data retrieved from /proc may be stale or zero
4796 Details : When reading per-device statfs data from /proc, in the
4797 {kbytes,files}_{total,free,avail} files, it may appear
4798 as zero or be out of date.
4801 Frequency : systems with MD RAID1 external journal devices
4803 Description: lconf's call to blkid is confused by RAID1 journal devices
4804 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4805 of returning all block devices that match the journal UUID.
4808 Frequency : always, for aggregate stripe size over 4GB
4810 Description: assertion fails when trying to use 4GB stripe size
4811 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4812 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4815 Frequency : always on ppc64
4817 Description: the first write on an ext3 filesystem with mballoc got stuck
4818 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4819 perform endianness conversion.
4821 ------------------------------------------------------------------------------
4823 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4825 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4826 INTEROPERATE with older versions automatically. Please read the
4827 user documentation before upgrading any part of a live system.
4828 * WARNING: Lustre networking configuration changes are required with
4829 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4832 * Support for kernels:
4833 2.6.9-22.0.2.EL (RHEL 4)
4834 2.6.5-7.244 (SLES 9)
4835 2.6.12.6 vanilla (kernel.org)
4838 Severity : enhancement
4839 Bugzilla : 7981/8208
4840 Description: Introduced Lustre Networking (LNET)
4841 Details : LNET is new networking infrastructure for Lustre, it includes
4842 a reorganized network configuration mode (see the user
4843 documentation for full details) as well as support for routing
4844 between different network fabrics. Lustre Networking Devices
4845 (LNDs) for the supported network fabrics have also been
4846 created for this new infrastructure.
4848 Severity : enhancement
4849 Description: Introduced Access control lists
4850 Details : clients can set ACLs on files and directories in order to have
4851 more fine-grained permissions than the standard Unix UGO+RWX.
4852 The MDS must be started with the "-o acl" mount option.
4854 Severity : enhancement
4855 Description: Introduced filesystem quotas
4856 Details : Administrators may now establish per-user quotas on the
4859 Severity : enhancement
4861 Description: Configuration change for the XT3
4862 The PTLLND is now used to run Lustre over Portals on the XT3
4863 The configure option(s) --with-cray-portals are no longer used.
4864 Rather --with-portals=<path-to-portals-includes> is used to
4865 enable building on the XT3. In addition to enable XT3 specific
4866 features the option --enable-cray-xt3 must be used.
4871 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4872 Details : A race between allocating a new llcd and re-getting the llcd_lock
4873 allowed another thread to grab newly-allocated llcd.
4875 Severity : enhancement
4877 Description: 2.6 OST async journal commit and locking fix to improve performance
4878 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4879 2.6 kernels are now async as they already were in 2.4 kernels so
4880 that they can commit concurrently with the network bulk transfer.
4881 For block-allocated files the filter allocation semaphore is held
4882 to avoid filesystem fragmentation during allocation. BKL lock
4883 removed for 2.6 xattr operations where it is no longer needed.
4888 Description: lconf incorrectly determined whether two IP networks could talk
4889 Details : In some more complicated routing and multiple-network
4890 configurations, lconf will avoid trying to make a network
4891 connection to a disjoint part of the IP space. It was doing the
4892 math incorrectly for one set of cases.
4897 Description: Fix for potential infinite loop processing records in an llog.
4898 Details : If an llog record is corrupted/zeroed, it is possible to loop
4899 forever in llog_process(). Validate the llog record length
4900 and skip the remainder of the block on error.
4903 Frequency : occasional (liblustre only)
4905 Description: liblustre could not open files whose last component is a symlink
4906 Details : sysio_path_walk() would incorrectly pass the open intent to
4907 intermediate path components.
4910 Frequency : rare (liblustre only with non-standard tuning)
4911 Bugzilla : 7201 (7350)
4912 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4913 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4914 after it has been released, only do this for referenced locks
4916 Severity : enhancement
4918 Description: specifying an (invalid) directory default stripe_size of -1
4919 would reset the directory default striping
4920 Details : stripe_size -1 was used internally to signal directory stripe
4921 removal, now use "all default" to signal dir stripe removal
4922 as a directory striping of "all default" is not useful
4925 Frequency : common for large clusters running liblustre clients
4927 Description: doing an ls when liblustre clients are running is slow
4928 Details : sending a glimpse AST to a liblustre client waits for every AST
4929 to time out, as liblustre clients will not respond. Since they
4930 cannot cache data we refresh the OST lock LVB from disk instead.
4932 Severity : enhancement
4934 Description: doing an ls at the same time as file IO can be slow
4935 Details : enqueue and other "small" requests can be blocked behind many
4936 large IO requests. Create a new OST IO portal for non-IO
4937 requests so they can be processed faster.
4940 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4942 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4943 created file with mode 000
4944 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4945 allow an ftruncate() as a non-root user to a file with mode 000.
4946 Lustre now respects this flag to disable mode checks when
4947 truncating a file owned by the user
4950 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4953 Description: Revoking locks from clients that went dead or catatonic might take
4955 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4956 cancellation of such locks instant on servers without waiting for
4957 any reply from clients. Clients drops these locks when cancel
4958 notification from server is received without replying.
4961 Frequency : liblustre-only, when liblustre client dies or becomes busy
4963 Description: Doing ls on Linux clients can take a long time with active
4965 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4966 granting such locks to it in the first place if possible. Locks
4967 are taken by proxy on the OST during the read or write and
4968 dropped immediately afterward. Add connect flags handling, do
4969 not grant locks to liblustre clients for glimpse ASTs.
4971 Severity : enhancement
4973 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4974 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4975 sequential reads and behaves badly with applications doing random
4976 reads. Improve it by reading ahead at least read region, and
4977 avoiding excessive large RPC for small reads.
4979 Severity : enhancement
4981 Description: Creating more than 1000 files for a single job may cause a load
4982 imbalance on the OSTs if there are also a large number of OSTs.
4983 Details : qos_prep_create() uses an OST index reseed value that is an
4984 even multiple of the number of available OSTs so that if the
4985 reseed happens in the middle of the object allocation it will
4986 still utilize the OSTs as uniformly as possible.
4991 Description: OST or MDS may oops in ping_evictor_main()
4992 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4993 but doesn't restart at beginning of obd_exports_timed list
4996 Severity : enhancement
4998 Description: improve by-nid export eviction on the MDS and OST
4999 Details : allow multiple exports with the same NID to be evicted at one
5000 time without re-searching the exports list.
5003 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5005 Description: MDS may oops in groups_free()
5006 Details : in rare race conditions a newly allocated group_info struct is
5007 freed again, and this can be NULL. The 2.4 compatibility code
5008 for groups_free() checked for a NULL pointer, but 2.6 did not.
5011 Frequency : common for liblustre clients doing little filesystem IO
5012 Bugzilla : 9352, 7313
5013 Description: server may evict liblustre clients accessing contended locks
5014 Details : if a client is granted a lock or receives a completion AST
5015 with a blocking AST already set it would not reply to the AST
5016 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5019 Frequency : lfs setstripe, only systems with more than 160 OSTs
5021 Description: unable to set striping with a starting offset beyond OST 160
5022 Details : llapi_create_file() incorrectly limited the starting stripe
5023 index to the maximum single-file stripe count.
5026 Frequency : LDAP users only
5028 Description: lconf did not handle in-kernel recovery with LDAP properly
5029 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5031 Severity : enhancement
5033 Description: bind OST threads to NUMA nodes to improve performance
5034 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5035 node and do their allocations there to localize memory access
5037 Severity : enhancement
5039 Description: llmount can determine client NID directly from Myrinet (GM)
5040 Details : the client NID code from gmnalnid was moved directly into
5041 llmount, removing the need to use this or specifying the
5042 client NID explicitly when mounting GM clients with zeroconf
5045 Frequency : if client is started with down MDS
5047 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5048 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5049 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5050 ptlrpc_import_delayed_req() because we didn't check if the
5051 request was interrupted, and we also didn't break out of the
5052 event loop if there was no timeout
5057 Description: data loss during non-page-aligned writes to a single file from
5058 both multiple nodes and multiple threads on one node at same time
5059 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5060 inconsistency led to false short-reads, that were cached and later
5061 used by ->prepare_write() to fill in partially written page,
5062 leading to data loss.
5065 Frequency : always, if lconf --abort_recovery used
5067 Description: lconf --abort_recovery fails with 'Operation not supported'
5068 Details : lconf was attempting to abort recovery on the MDT device and not
5071 Severity : enhancement
5073 Description: remove cleanup logs
5074 Details : replace lconf-generated cleanup logs with lustre internal
5075 cleanup routines. Eliminates the need for client-cleanup and
5078 Severity : enhancement
5080 Description: add support for EAs (user and system) on lustre filesystems
5081 Details : it is now possible to store extended attributes in the Lustre
5082 client filesystem, and with the user_xattr mount option it
5083 is possible to allow users to store EAs on their files also
5085 Severity : enhancement
5087 Description: Add possibility (config option) to show minimal available OST free
5089 Details : When compiled with --enable-mindf configure option, statfs(2)
5090 (and so, df) will return least minimal free space available from
5091 all OSTs as amount of free space on FS, instead of summary of
5092 free spaces of all OSTs.
5094 Severity : enhancement
5096 Description: do not expand extent locks acquired on OST-side
5097 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5098 by server: they are not cached anyway.
5101 Frequency : when mmap is used/binaries executed from Lustre
5103 Description: Unmmap pages before throwing them away from read cache.
5104 Details : llap_shrink cache now attempts to unmap pages before discarding
5105 them (if unmapping failed - do not discard). SLES9 kernel has
5106 extra checks that trigger if this unmapping is not done first.
5111 Description: lconf didn't resolve symlinks before checking to see whether a
5112 given mountpoint was already in use
5115 Frequency : when migrating failover services
5116 Bugzilla : 6395, 9514
5117 Description: When migrating a subset of services from a node (e.g. failback
5118 from a failover service node) the remaining services would
5119 time out and evict clients.
5120 Details : lconf --force (implied by --failover) sets the global obd_timeout
5121 to 5 seconds in order to quickly disconnect, but this caused
5122 other RPCs to time out too quickly. Do not change the global
5123 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5125 Severity : enhancement
5126 Frequency : if MDS is started with down OST
5127 Bugzilla : 9439,5706
5128 Description: Allow startup/shutdown of an MDS without depending on the
5129 availability of the OSTs.
5130 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5131 Add appropriate locking and lov-osc refcounts for safe
5132 cleaning. Add osc abort_inflight calls in case the
5133 synchronize never started.
5136 Frequency : occasional (Cray XT3 only)
5138 Description: root not authorized to access files in CRAY_PORTALS environment
5139 Details : The client process capabilities were not honoured on the MDS in
5140 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5141 been accessed by an authorized user then root was able to access
5142 the file on the local client also. The root user capabilities
5143 are now allowed on the MDS, as this environment has secure UID.
5146 Frequency : occasional
5148 Description: ldiskfs "too long searching" message happens too often
5149 Details : A debugging message (otherwise harmless) prints too often on
5150 the OST console. This has been reduced to only happen when
5151 there are fragmentation problems on the filesystem.
5156 Description: Division by zero in statfs when all OSCs are inactive
5157 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5158 lov_statfs divides by value returned by lov_get_stripecnt().
5162 Bugzilla : 9489, 3273
5163 Description: First write from each client to each OST was only 4kB in size,
5164 to initialize client writeback cache, which caused sub-optimal
5165 RPCs and poor layout on disk for the first written file.
5166 Details : Clients now request an initial cache grant at (re)connect time
5167 and so that they can start streaming writes to the cache right
5168 away and always do full-sized RPCs if there is enough data.
5169 If the OST is rebooted the client also re-establishes its grant
5170 so that client cached writes will be honoured under the grant.
5175 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5176 Details : Now I/O RPCs go to different portal number and (presumably) fast
5177 lock requests (and glimses) and other RPCs get their own service
5178 threads pool that should be able to service those RPCs
5181 Severity : enhancement
5183 Description: Ability to exchange lustre version between client and servers and
5184 issue warnings at client side if client is too old. Also for
5185 liblustre clients there is ability to refuse connection of too old
5187 Details : New 'version' field is added to connect data structure that is
5188 filled with version info. That info is later checked by server and
5192 Frequency : rare, liblustre only.
5193 Bugzilla : 9296, 9581
5194 Description: Two simultaneous writes from liblustre at offset within same page
5195 might proceed at the same time overwriting eachother with stale
5197 Details : I/O lock withing llu_file_prwv was released too early, before data
5198 actually was hitting the wire. Extended lock-holding time until
5199 server acknowledges receiving data.
5202 Frequency : extremely rare. Never observed in practice.
5204 Description: avoid generating lustre_handle cookie of 0.
5205 Details : class_handle_hash() generates handle cookies by incrementing
5206 global counter, and can hit 0 occasionaly (this is unlikely, but
5207 not impossible, because initial value of cookie counter is
5208 selected randonly). Value of 0 is used as a sentinel meaning
5209 "unassigned handle" --- avoid it. Also coalesce two critical
5210 sections in this function into one.
5212 Severity : enhancement
5214 Description: allow liblustre clients to delegate truncate locking to OST
5215 Details : To avoid overhead of locking, liblustre client instructs OST to
5216 take extent lock in ost_punch() on client's behalf. New connection
5217 flag is added to handle backward compatibility.
5219 Severity : enhancement
5220 Bugzilla : 4928, 7341, 9758
5221 Description: allow number of OST service threads to be specified
5222 Details : a module parameter allows the number of OST service threads
5223 to be specified via "options ost ost_num_threads={N}" in the
5224 OSS's /etc/modules.conf or /etc/modprobe.conf.
5228 Bugzilla : 6146, 9635, 9895
5229 Description: servers crash with bad pointer in target_handle_connect()
5230 Details : In rare cases when a client is reconnecting it was possible that
5231 the connection request was the last reference for that export.
5232 We would temporarily drop the export reference and get a new
5233 one, but this may have been the last reference and the export
5234 was just destroyed. Get new reference before dropping old one.
5236 Severity : enhancement
5237 Frequency : if client is started with failover MDS
5239 Description: Allow multiple MDS hostnames in the mount command
5240 Details : Try to read the configuration from all specified MDS
5241 hostnames during a client mount in case the "primary"
5244 Severity : enhancement
5246 Description: Stop sending data to evicted clients as soon as possible.
5247 Details : Check if the client we are about to send or are sending data to
5248 was evicted already. (Check is done every second of waiting,
5249 for which l_wait_event interface was extended to allow checking
5250 of exit condition at specified intervals).
5253 Frequency : rare, normally only when NFS exporting is done from client
5255 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5257 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5260 Frequency : always, if srand() is called before liblustre initialization
5262 Description: Liblustre uses system PRNG disturbing its usage by user application
5263 Details : Introduce internal to lustre fast and high-quality PRNG for
5264 lustre usage and make liblustre and some other places in generic
5265 lustre code to use it.
5267 Severity : enhancement
5268 Bugzilla : 9477, 9557, 9870
5269 Description: Verify that the MDS configuration logs are updated when xml is
5270 Details : Check if the .xml configuration logs are newer than the config
5271 logs stored on the MDS and report an error if this is the case.
5272 Request --write-conf, or allow starting with --old_conf.
5274 Severity : enhancement
5276 Description: Handle symlinks in the path when checking if Lustre is mounted.
5277 Details : Resolve intermediate symlinks when checking if a client has
5278 mounted a filesystem to avoid duplicate client mounts.
5283 Description: lconf can hit an error exception but still return success.
5284 Details : The lconf command catches the Command error exception at the top
5285 level script context and will exit with the associated exit
5286 status, but doesn't ensure that this exit status is non-zero.
5291 Description: failure of ptlrpc thread startup can cause oops
5292 Details : Starting a ptlrpc service thread can fail if there are a large
5293 number of threads or the server memory is very fragmented.
5294 Handle this without oopsing.
5297 Frequency : always, only if liblustre and non-default acceptor port was used
5299 Description: liblustre cannot connect to servers with non-default acceptor port
5300 Details : tcpnal_set_default_params() was not called and was therefore
5301 ignoring the environment varaible TCPNAL_PORT, as well as other
5302 TCPNAL_ environment variables
5307 Description: two objects could be created on the same OST for a single file
5308 Details : If an OST is down, in some cases it was possible to create two
5309 objects on a single OST for a single file. No problems other
5310 than potential performance impact and spurious error messages.
5314 Bugzilla : 5681, 9562
5315 Description: Client may oops in ll_unhash_aliases
5316 Details : Client dcache may become inconsistent in race condition.
5317 In some cases "getcwd" can fail if the current directory is
5323 Description: Inode refcounting problems in NFS export code
5324 Details : link_raw functions used to call d_instantiate without obtaining
5325 extra inode reference first.
5329 Bugzilla : 9942, 9903
5330 Description: Referencing freed requests leading to crash, memleaks with NFS.
5331 Details : We used to require that call to ll_revalidate_it was always
5332 followed by ll_lookup_it. Also with revalidate_special() it is
5333 possible to call ll_revalidate_it() twice for the same dentry
5334 even if first occurence returned success. This fix changes semantic
5335 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5336 reference on a request referred from the intent.
5337 ll_intent_release() then releases such a request.
5340 Frequency : rare, normally benchmark loads only
5342 Description: unlinked inodes were kept in memory on the client
5343 Details : If a client is repeatedly creating and unlinking files it
5344 can accumulate a lot of stale inodes in the inode slab cache.
5345 If there is no other client load running this can cause the
5346 client node to run out of memory. Instead flush old inodes
5347 from client cache that have the same inode number as a new inode.
5350 Frequency : SLES9 2.6.5 kernel and long filenames only
5351 Bugzilla : 9969, 10379
5352 Description: utime reports stale NFS file handle
5353 Details : SLES9 uses out-of-dentry names in some cases, which confused
5354 the lustre dentry revalidation. Change it to always use the
5358 Frequency : rare, unless heavy write-truncate concurrency is continuous
5359 Bugzilla : 4180, 6984, 7171, 9963, 9331
5360 Description: OST becomes very slow and/or deadlocked during object unlink
5361 Details : filter_destroy() was holding onto the parent directory lock
5362 while truncating+unlinking objects. For very large objects this
5363 may block other threads for a long time and slow overall OST
5364 responsiveness. It may also be possible to get a lock ordering
5365 deadlock in this case, or run out of journal credits because of
5366 the combined truncate+unlink. Solution is to do object truncate
5367 first in one transaction without parent lock, and then do the
5368 final unlink in a new transaction with the parent lock. This
5369 reduces the lock hold time dramatically.
5372 Frequency : rare, 2.4 kernels only
5374 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5375 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5376 before unmounting their filesystem in order to avoid deadlock.
5377 The kernel_locked() function in 2.4 kernels only checks whether
5378 the kernel lock is held, not whether it is this process that is
5379 holding it as 2.6 kernels do.
5384 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5385 Details : The client ptlrpc code may be trying to reconnect to a down
5386 server before a previous connection attempt has timed out.
5387 Increase the reconnect interval to be longer than the connection
5388 timeout interval to avoid sending duplicate connections to
5392 Frequency : echo_client brw_test command
5394 Description: fix echo_client to work with OST preallocated code
5395 Details : OST preallocation code (5137) didn't take echo_client IO path
5396 into account: echo_client calls filter methods outside of any
5397 OST thread and, hence, there is no per-thread preallocated
5398 pages and buffers to use. Solution: hijack pga pages for IO. As
5399 a byproduct, this avoids unnecessary data copying.
5403 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5404 Description: Client can oops in mdc_commit_close() after open replay
5405 Details : It was possible for the MDS to return an open request with no
5406 transaction number in mds_finish_transno() if the client was
5407 evicted, but without actually returning an error. Clients
5408 would later try to replay that open and may trip an assertion
5409 Simplify the client close codepath, and always return an error
5410 from the MDS in case the open is not successful.
5413 Frequency : rare, 2.6 OSTs only
5415 Description: OST may deadlock under high load on fragmented files
5416 Details : If there was a heavy load and highly-fragmented OST filesystems
5417 it was possible to have all the OST threads deadlock waiting on
5418 allocation of biovecs, because the biovecs were not released
5419 until the entire RPC IO was completed. Instead, release biovecs
5420 as soon as they are complete to ensure forward IO progress.
5422 Severity : enhancement
5424 Description: Support for specifying external journal device at mount
5425 Details : If an OST or MDS device is formatted with an external journal
5426 device, this device major/minor is stored in the ext3 superblock
5427 and may not be valid for failover. Allow detecting and
5428 specifying the external journal at mount time.
5433 Description: Mounting an MDS with pending unlinked files may cause oops
5434 Details : target_finish_recovery() calls mds_postrecov() which returned
5435 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5436 considers this an error and immediately begins cleaning up the
5437 lov, just after starting the mds_lov process
5439 Severity : enhancement
5441 Description: Implement 'lfs df' to report actual free space on per-OST basis
5442 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5443 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5444 usage of objects; '-h' to report in human readable format.
5446 ------------------------------------------------------------------------------
5448 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5455 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5456 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5457 at startup during orphan unlinking. Instead, skip buddy update
5458 before setup, as it will be regenerated after recovery is complete.
5463 Description: If an OST is inactive, its locks might reference stale inodes.
5464 Details : lov_change_cbdata() must iterate over all namespaces, even if
5465 they are inactive to clear inode references from the lock.
5467 Severity : enhancement
5468 Frequency : occasional, if non-standard max_dirty_mb used
5470 Description: Client will block write RPCs if not enough grant
5471 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5472 then the client will block writes while waiting for another RPC
5473 to complete instead of consuming its dirty limit. With change
5474 we get improved performance when max_dirty_mb is small.
5476 Severity : enhancement
5477 Bugzilla : 3389, 6253
5478 Description: Add support for supplementary groups on the MDS.
5479 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5480 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5481 lookups for user supplementary groups into a cache.
5485 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5486 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5487 for liblustre O_EXCL is enforced by MDS.
5491 Description: readdir never returns NULL in liblustre.
5492 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5493 dentry in struct dirent.
5497 Frequency : liblustre only, depends on application IO pattern
5498 Description: liblustre clients evicted if not contacting servers
5499 Details : Don't put liblustre clients into the ping_evictor list, so
5500 they will not be evicted by the pinger ever.
5502 Severity : enhancement
5504 Description: Add ability to evict clients by NID from MDS.
5505 Details : By echoing "nid:$NID" string into
5506 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5507 $NID would be instantly evicted from this MDS and from all active
5508 OSTs connected to it.
5512 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5513 Details : lookup_it_finish() used to query file size from OSTs that was not
5518 Description: service threads change working directory to that of init
5519 Details : Starting lustre service threads may pin the working directory
5520 of the parent thread, making that filesystem busy. Threads
5521 now change to the working directory of init to avoid this.
5525 Frequency : during shutdown only
5526 Description: shutdown with a failed MDS or OST can cause unmount to hang
5527 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5532 Frequency : 2.6 only
5533 Description: chmod/chown may include an extra supplementary group
5534 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5535 supplementary group and if none is specified this is used.
5538 Bugzilla : 5479 (6816)
5540 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5541 Details : If lookup is in progress on a file that is unlinked we might try
5542 to revalidate the inode and fail in revalidate after lookup is
5543 complete and ll_file_open() enqueues the open again but
5544 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5547 Frequency : always, if lconf --abort_recovery used
5549 Description: lconf --abort_recovery fails with 'Operation not supported'
5550 Details : lconf was attempting to abort recovery on the MDT device and not
5553 ------------------------------------------------------------------------------
5555 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5560 Frequency : rare (only unsupported configurations with a node running as an
5562 Bugzilla : 6514, 5137
5563 Description: Mounting a Lustre file system on a node running as an OST could
5565 Details : OSTs now preallocates memory needed to write out data at
5566 startup, instead of when needed, to avoid having to
5567 allocate memory in possibly low memory situations.
5568 Specifically, if the file system is mounted on on OST,
5569 memory pressure could force it to try to write out data,
5570 which it needed to allocate memory to do. Due to the low
5571 memory, it would be unable to do so and the node would
5572 become unresponsive.
5574 Severity : enhancement
5576 Description: Addition of lconf --service command line option
5577 Details : lconf now accepts a '--service <arg>' option, which is
5578 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5580 Severity : enhancement
5582 Description: Failover mode is now the default for OSTs.
5583 Details : By default, OSTs will now run in failover mode. To return to
5584 the old behaviour, add '--failout' to the lmc line for OSTs.
5586 Severity : enhancement
5588 Description: Health checks are now provided for MDS and OSTs
5589 Details : Additional detailed health check information on MSD and OSTs
5590 is now provided through the procfs health_check value.
5593 Frequency : occasional, depends on IO load
5595 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5596 after numerous create/delete cycles
5597 Details : The ext3 inode allocation policy would not allocate new inodes
5598 very well on the OSTs because there are no new directories
5599 being created. Instead we look for groups with free space if
5600 the parent directories are nearly full.
5605 Description: Network or server problems during mount may cause partially
5606 mounted clients instead of returning an error.
5607 Details : The config llog parsing code may overwrite the error return
5608 code during mount error handling, returning success instead
5614 Description: MDS can fail to allocate large reply buffers
5615 Details : After long uptimes the MDS can fail to allocate large reply
5616 buffers (e.g. zconf client mount config records) due to memory
5617 fragmentation or consumption by the buffer cache. Preallocate
5618 some large reply buffers so that these replies can be sent even
5619 under memory pressure.
5623 Frequency : rare (liblustre)
5624 Description: fsx running with liblustre complained that using truncate() to
5625 extend the file doesn't work. This patch corrects that issue.
5626 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5627 ATTR_SIZE and lsm use in llu_setattr_raw.
5631 Frequency : rare, only 2.6 kernels
5632 Description: Unusual file access patterns on the MDS may result in inode
5633 data being lost in very rare circumstances.
5634 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5635 code in the 2.6 kernel caused the inode and/or EA data not to
5636 be read from disk, causing single-file corruption.
5640 Frequency : rare, only 2.6 filesystems using extents
5641 Description: Heavy concurrent write and delete load may cause data corruption.
5642 Details : It was possible under high-load situations to have an extent
5643 metadata block in the block device cache from a just-unlinked
5644 file overwrite a newly-allocated data block. We now unmap any
5645 metadata buffers that alias just-allocated data blocks.
5649 Frequency : filesystems with default stripe_count larger than 77
5650 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5651 Details : lconf specifies an inode size of 4096 bytes when the default
5652 stripe_count is larger than 77. This conflicts with the default
5653 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5654 this case to avoid pinning too much memory for large EAs.
5656 ------------------------------------------------------------------------------
5658 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5663 Frequency : rare (extremely heavy IO load with hundreds of clients)
5665 Description: Client is evicted, gets IO error writing to file
5666 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5667 caused clients to be evicted for AST timeouts. The fixes in
5668 bug 5192 mean we no longer need to have such short AST timeouts
5669 so ldlm_timeout has been increased.
5672 Frequency : occasional during --force or --failover shutdown under load
5673 Bugzilla : 5949, 4834
5674 Description: Server oops/LBUG if stopped with --force or --failover under load
5675 Details : a collection of import/export refcount and cleanup ordering
5676 issues fixed for safer force cleanup
5679 Frequency : only filesystems larger than 120 OSTs
5680 Bugzilla : 5990, 6223
5681 Description: lfs getstripe would oops on a very large filesystem
5682 Details : lov_getconfig used kfree on vmalloc'd memory
5685 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5686 Bugzilla : 6242, 6243
5687 Description: reading from files that had been truncated to a non-zero size
5688 but never opened returned no data
5689 Details : ll_file_read() reads zeros from no-object files to EOF
5694 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5695 Details : There's little harm in aborting MDS/OSS recovery and letting it
5696 try again, so I removed the LASSERT and return an error instead.
5698 Severity : enhancement
5700 Description: New debugging infrastructure for tracking down data corruption
5701 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5702 (b) cover more of the client-side code path, and (c) try to narrow
5703 down where problems occurred
5707 Bugzilla : 3819, 4364, 4397, 6313
5708 Description: Racing close and eviction MDS could cause assertion in mds_close
5709 Details : It was possible to get multiple mfd references during close and
5710 client eviction, leading to one thread referencing a freed mfd.
5712 Severity: : enhancement
5713 Bugzilla : 3262, 6359
5714 Description: Attempts to reconnect to servers are now more aggressive.
5715 Details : This builds on the enhanced upcall-less recovery that was added
5716 in 1.4.2. When trying to reconnect to servers, clients will
5717 now try each server in the failover group every 10 seconds. By
5718 default, clients would previously try one server every 25 seconds.
5723 Description: After recovery, certain operations trigger a failed
5724 assertion on a client.
5725 Details : Failing over an mds, using lconf -d --failover, while a
5726 client was doing a readdir() call would cause the client to
5727 LBUG after recovery completed and the readdir() was resent.
5729 Severity : enhancement
5731 Description: Default groups are now added by lconf
5732 Details : You can now run lconf --group <servicename> without having to
5733 manually add groups with lmc.
5736 Frequency : occasional
5738 Description: Nodes with an elan id of 0 trigger a failed assertion
5741 Frequency : always when accessing e.g. tty/console device nodes
5743 Description: tty and some other devices nodes cannot be used on lustre
5744 Details : file's private_data field is used by device data and lustre
5745 values in there got lost. New field was added to struct file to
5746 store fs-specific private data.
5749 Frequency : when exporting Lustre via NFS
5751 Description: NFSD failed occasionally when looking up a path component
5752 Details : NFSD is looking up ".." which was broken in ext3 directories
5753 that had grown large enough to become hashed.
5756 Frequency : Clusters with multiple interfaces not on the same subnet
5758 Description: Nodes will repeatedly try to reconnect to an interface which it
5759 cannot reach and report an error to the log.
5760 Details : Extra peer list entries will be created by lconf with some peers
5761 unreachable. lconf now validates the peer before adding it.
5764 Frequency : Only if a default stripe is set on the filesystem root.
5766 Description: Setting a default stripe on the filesystem root prevented the
5767 filesystem from being remounted.
5768 Details : The client was sending extra request flags in the root getattr
5769 request and did not allocate a reply buffer for the dir EA.
5772 Frequency : occasional, higher if lots of files are accessed by one client
5773 Bugzilla : 6159, 6097
5774 Description: Client trips assertion regarding lsm mismatch/magic
5775 Details : While revalidating inodes the VFS looks up inodes with ifind()
5776 and in rare cases can find an inode that is being freed.
5777 The ll_test_inode() code will free the lsm during ifind()
5778 when it finds an existing inode and then the VFS later attaches
5779 this free lsm to a new inode.
5783 Bugzilla : 6422, 7030
5784 Description: MDS deadlock between mkdir and client eviction
5785 Details : Creating a new file via mkdir or mknod (starting a transaction
5786 and getting the ns lock) can deadlock with client eviction
5787 (gets ns lock and trying to finish a synchronous transaction).
5790 Frequency : occasional
5791 Description: While starting a server, the fsfilt_ext3 module could not be
5793 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5794 kernels. Previously, lconf would still use the ext3 name
5795 when trying to load modules. Now, it will correctly use
5796 ext3 on 2.4 and ldiskfs on 2.6.
5798 Severity : enhancement
5799 Description: The default stripe count has been changed to 1
5800 Details : The interpretation of the default stripe count (0, to lfs
5801 or lmc) has been changed to mean striping across a single
5802 OST, rather than all available. For general usage we have
5803 found a stripe count of 1 or 2 works best.
5805 Severity : enhancement
5806 Description: Add support for compiling against Cray portals.
5807 Details : Conditional compiling for some areas that are different
5811 Frequency : occasional
5812 Bugzilla : 6409, 6834
5813 Description: Creating files with an explicit stripe count may lead to
5814 a failed assertion on the MDS
5815 Details : If some OSTs are full or unavailable, creating files may
5816 trigger a failed assertion on the MDS. Now, Lustre will
5817 try to use other servers or return an error to the
5821 Frequency : occasional
5823 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5824 caused lock timeout during readahead (since 1.4.2).
5825 Details : Processes doing readahead might match a lock that hasn't been
5826 granted yet if there are overlapping and conflicting lock
5827 requests. The readahead process waits on ungranted lock
5828 (original lock is CBPENDING), while OST waits for that process
5829 to cancel CBPENDING read lock and eventually evicts client.
5831 Severity : enhancement
5833 Description: Initial enabling of flock support for clients
5834 Details : Implements fcntl advisory locking and file status functions.
5835 This feature is provided as an optional mount flag (default
5836 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5837 locking are implemented yet, and those that are are not guaranteed
5838 to be completely correct in production environments.
5839 mount -t lustre -o [flock|noflock] ...
5842 Frequency : occasional
5844 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5845 assertion in the ext3 JBD (journaling) layer.
5846 Details : The b_committed_data struct is protected by the big kernel lock
5847 in 2.4 kernels, serializing journal_commit_transaction() and
5848 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5849 access to this struct. In 2.6 kernels there is finer grained
5850 locking to improve SMP performance of the JBD layer.
5854 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5859 Description: merge and cleanup kernel patches.
5862 -----------------------------------------------------------------------------
5864 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5866 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5867 versions of Lustre. Please update all servers and clients to
5868 version 1.4.2 or later at the same time. You must also run
5869 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5870 to update the configuration logs.
5872 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5873 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5874 - swab llog records of type '0' so we get proper header size/idx (5861)
5875 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5876 - fix rename of one directory over another leaking an inode (5953)
5877 - avoid SetPageDirty on 2.6 (5981)
5878 - don't re-add just-being-destroyed locks to the waiting list (5653)
5879 - when creating new directories, inherit the parent's custom
5880 striping settings if present parent (3048)
5881 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5882 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5883 - don't hold client locks on temporary worklist from l_lru (5666)
5884 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5885 - automatically evict dead clients (5921)
5886 - Update file size properly in create+truncate+fstat case (6196)
5887 - Do not unhash mountpoint dentries, do not allow removal of
5889 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5890 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5891 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5892 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5893 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5894 - Correctly handle reads of files with no objects (6243)
5895 - lctl recover will also mark a device active if deactivate used (5933)
5897 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5898 - allow --write-conf on an MDS with different nettype than client (5619)
5899 - don't write config llogs to MDS for mounts not from that MDS (5617)
5900 - lconf should create multiple TCP connections from a client (5201)
5901 - init scripts are now turned off by default; run chkconfig --on
5902 lustre and chkconfig --on lustrefs to use them
5903 - upcalls are no longer needed for clients to recover to failover
5905 - add --abort-recovery option to lconf to abort recovery on device
5907 - add support for an arbitrary number of OSTs (3026)
5908 - Quota support protocol changes.
5909 - forward compatibility changes to wire structs (6007)
5910 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5911 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5912 - improved "lustre status" script
5913 - initialize blocksize for non-regular files (6062)
5914 - added --disable-server and --disable-client configure options (5782)
5915 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5916 - Vanilla 2.4.29 support
5917 - increase maximum number of obd devices to 520 (6242)
5918 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5919 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5920 SLES 9 2.6.5 kernels
5921 - limit stripes per file to 160 (the maximum EA size) (6093)
5923 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5926 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5927 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5928 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5929 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5930 - small ext3 extents cleanups and fixes (5733)
5931 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5932 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5933 - avoid needless client->OST connect, fix handle mismatch (5317)
5934 - fix DLM error path that led to out-of-sync client, long delays (5779)
5935 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5936 - fix several locking issues related to i_size (5492,5624,5654,5672)
5937 - don't move pending lock onto export if it is already evicted (5683)
5938 - fix kernel oops when creating .foo in unlinked directory (5548)
5939 - fix deadlock in obdfilter statistics vs. object create (5811)
5940 - use time_{before,after} to avoid timer jiffies wrap (5882)
5941 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5942 - Do not leak request if server was not able to process it (5154)
5943 - If mds_open unable to find parent dir, make that negative lookup(5154)
5944 - don't create new directories with extent-mapping (5909, 5936)
5946 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5947 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5948 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5949 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5950 - support for RHEL4 kernel on i686 (5773)
5951 - provide error messages when incompatible logs are encountered (5898)
5953 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5954 * version 1.4.0.10 (1.4.1 release candidate 1)
5956 - don't keep a lock reference when lock is not granted (4238)
5957 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5958 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5959 - drop import inflight refcount on signal_completed_replay error (5255)
5960 - unlock page after async write error during send (3677)
5961 - handle missing objects in filter_preprw_read properly (5265)
5962 - no transno return for symlink open, don't save no-trasno open (3440)
5963 - don't try to complete elan receive that already failed (4012)
5964 - free RPC server reply state on error (5406)
5965 - clean up thread from ptlrpc_start_thread() on error (5160)
5966 - readahead could read extra page into cache that wasn't ejected (5388)
5967 - prevent races in class_attach/setup/cleanup/detach (5260)
5968 - don't dereference de->d_inode after l_dput of de (5458)
5969 - use "int" for stripe value returned from lock_to_stripe (5544)
5970 - mballoc allocation and error-checking fixes in 2.6 (5504)
5971 - block device patches to fix I/O request sizes in 2.6 (5482)
5972 - look up hostnames for IB nals (5602)
5973 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5974 - don't start multiple acceptors for the same port (5277)
5975 - fix incorrect LASSERT in mds_getattr_name (5635)
5976 - export a proc file for general "ping" checking (5628)
5977 - fix "lfs check" to not block when the MDS is down (5628)
5979 - service request history (4965)
5980 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5981 - create an "evict_client" /proc entry on OSTs, like the MDS has
5982 - fix mount usage message, return errors per mount(8) (5168)
5983 - change grep [] to grep "[]" in tests so they work in more UMLs
5984 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5985 - remove ancient LOV_MAGIC_V0 EA support (5047)
5986 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5987 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5988 - allow readahead allocations to fail when low on memory (5383)
5989 - mmap locking landed again, after considerable improvement (2828)
5990 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5992 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5995 - send OST transaction number in read/write reply to free req (4966)
5996 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5997 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5998 - the watchdog thread now runs as interruptible (5246)
5999 - flock/lockf fixes (but it's still disabled, pending 5135)
6000 - don't use EXT3 constants in llite code (5094)
6001 - memory shortage at startup could cause assertion (5176)
6003 - reorganization of lov code
6004 - single portals codebase
6006 - add extents/mballoc support (5025)
6007 - direct I/O reads in the obdfilter (4048)
6008 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6010 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6013 - send OST transaction number in read/write reply to free req (4966)
6014 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6015 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6016 - the watchdog thread now runs as interruptible (5246)
6017 - handle missing objects in filter_preprw_read properly (5265)
6018 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6019 - drop import inflight refcount on signal_completed_replay error (5255)
6020 - unlock page after async write error during send (3677)
6021 - return original error code on reconstructed replies (3761)
6022 - no transno return for symlink open, don't save no-trasno open (3440)
6024 - add pid to ldlm debugging output (4922)
6025 - bump the watchdog timeouts -- we can't handle 30sec yet
6026 - extra debugging for orphan dentry/inode bug (5259)
6028 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6031 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6032 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6033 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6034 - fix NULL dereference in /proc/sys/portals/routes (4827)
6035 - allow failed mdc_close() operations to be interrupted (4561)
6036 - stop precreate on OST before MDS would time out on it (4778)
6037 - don't send partial-page writes before EOF from client (4410)
6038 - discard client grant for sub-page writes on large-page clients (4520)
6039 - don't free dentries not owned by NFS code, check generation (4806)
6040 - fix lsm leak if mds_create_objects() fails (4801)
6041 - limit debug_daemon file size, always print CERROR messages (4789)
6042 - use transno after validating reply (3892)
6043 - process timed out requests if import state changes (3754)
6044 - update mtime on OST during writes, return in glimpse (4829)
6045 - add mkfsoptions to LDAP (4679)
6046 - use ->max_readahead method instead of zapping global ra (5039)
6047 - don't interrupt __l_wait_event() during strace
6049 - add software watchdogs to catch hung threads quickly (4941)
6050 - make lustrefs init script start after nfs is mounted
6051 - fix CWARN/ERROR duplication (4930)
6052 - return async write errors to application if possible (2248)
6053 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6054 - print NAL number in %x format (4645)
6055 - update barely-supported suse-2.4.21-171 series (4842)
6056 - support for sles 9 %post scripts
6057 - support for building 2.6 kernel-source packages
6058 - support for sles km_* packages
6060 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6063 - ignore -ENOENT errors in osc_destroy (3639)
6064 - notify osc create thread that OSC is being cleaned up (4600)
6065 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6066 - reconstruct ost_handle() like mds_handle() (4657)
6067 - create a new thread to do import eviction to avoid deadlock (3969)
6068 - let lconf resolve symlinked-to devices (4629)
6069 - don't unlink "objects" from directory with default EA (4554)
6070 - hold socknal file ref over connect in case target is down (4394)
6071 - allow more than 32000 subdirectories in a single directory (3244)
6072 - fix blocks count for O_DIRECT writes (3751)
6073 - OST returns ENOSPC from object create when no space left (4539)
6074 - don't send truncate RPC if file size isn't changing (4410)
6075 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6076 - bind to privileged port in socknal and tcpnal (3689)
6078 - rate limit CERROR/CWARN console message to avoid overload (4519)
6079 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6080 - basic mmap support (3918)
6081 - kernel patch series update from b1_4 (4711)
6083 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6086 - avoid crash during MDS cleanup with OST shut down (2775)
6087 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6088 - don't use bad inodes on the MDS (3744)
6089 - dynamic object preallocation to improve recovery speed (4236)
6090 - don't hold spinlock over lock dumping or change debug flags (4401)
6091 - don't zero obd_dev when it is force cleaned (3651)
6092 - print grants to console if they go negative (4431)
6093 - "lctl deactivate" will stop automatic recovery attempts (3406)
6094 - look for existing locks in ldlm_handle_enqueue() (3764)
6095 - don't resolve lock handle twice in recovery avoiding race (4401)
6096 - revalidate should check working dir is a directory (4134)
6098 - don't always mark "slow" obdfilter messages as errors (4418)
6100 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6103 - don't close LustreDB during write_conf until it is done (3860)
6104 - fix typo in lconf for_each_profile (3821)
6105 - allow dumping logs from multiple threads at one time (3820)
6106 - don't allow multiple threads in OSC recovery (3812)
6107 - fix debug_size parameters (3864)
6108 - fix mds_postrecov to initialize import for llog ctxt (3121)
6109 - replace config semaphore with spinlock (3306)
6110 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6111 - don't allow enqueue to complete on a destroyed export (3822)
6112 - down write_lock before checking llog header bitmap (3825)
6113 - recover from lock replay timeout (3764)
6114 - up llog sem before sending rpc (3652)
6115 - reduce ns lock hold times when setting kms (3267)
6116 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6117 - fix NULL deref and obd_dev leak on setup error (3312)
6118 - replace some LBUG about llog ops with error handling (3841)
6119 - don't match INVALID dentries from d_lookup and spin (3784)
6120 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6121 - fix invalid assertion in ptlrpc_set_wait (3880)
6123 - add libwrap support for the TCP acceptor (3996)
6124 - add /proc/sys/portals/routes for non-root route listing (3994)
6125 - allow setting MDS UUID in .xml (2580)
6126 - print the stack of a process that LBUGs (4228)
6128 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6131 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6132 - make sure to unset replay flag from failed open requests (3440)
6133 - if default stripe count is 0, use OST count for inode size (3636)
6134 - update parent mtime/ctime on client for create/unlink (2611)
6135 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6136 - free recovery state on server during a forced cleanup (3571)
6137 - unregister_reply for resent reqs (3063)
6138 - loop back devices mounting and status check on 2.6 (3563)
6139 - fix resource-creation race that can provoke i_size == 0 (3513)
6140 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6141 - more debugging for page-accounting assertion (3746)
6142 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6143 - avoid deadlock after precreation failure (3758)
6144 - fix race and lock order deadlock in orphan handling (3450, 3750)
6145 - add validity checks when grabbing inodes from l_ast_data (3599)
6147 - add /proc/.../recovery_status to obdfilter (3428)
6148 - lightweight CDEBUG infrastructure, debug daemon (3668)
6149 - change default OSC RPC parameters to be better on small clusters
6150 - turn off OST read cache for files smaller than 32MB
6151 - install man pages and include them in rpms (3100)
6152 - add new init script for (un)mounting lustre filesystems (2593)
6153 - run chkconfig in %post for init scripts (3701)
6154 - drop scimac NAL (unmaintained)
6156 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6159 - clean kiobufs before and after use (3485)
6160 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6161 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6162 - change init script to fail more clearly if not run as root (1528)
6163 - allow clients to reconnect during replay (1742)
6164 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6165 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6166 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6167 - fix cond_resched (3554)
6168 - only evict unfinished clients after recovery (3515)
6169 - allow bulk resend, prevent data loss (3570)
6170 - dynamic ptlrpc request buffer allocation (2102)
6171 - don't allow unlinking open directory if it isn't empty (2904)
6172 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6173 - remove extraneous obd dereference causing LASSERT failure (3334)
6174 - don't use get_cycles() when creating temp. files on the mds (3156)
6175 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6176 - handle EEXIST for set-stripe, set proper directory name (3336)
6178 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6179 - fix ksocknal_fmb_callback() error messages (2918)
6181 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6184 - don't copy lvb into (possibly NULL) reply on error (2983)
6185 - don't deref dentry after dput, don't free lvb on error (2922)
6186 - use the kms to determine writeback rpc length (2947)
6187 - increment oti_logcookies when osc is inactive (2948)
6188 - update client's i_blocks count via lvb messages (2543)
6189 - handle intent open/close of special files properly (1557)
6190 - mount MDS with errors=remount-ro, like obdfilter (2009)
6191 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6192 - don't use cancelling-locks' kms values (2947)
6193 - use highest lock extent for kms, not last one (2925)
6194 - don't dereference ERR_PTR() dentry in error handling path (3107)
6195 - fix thread race in portals_debug_dumplog() (3122)
6196 - create lprocfs device entries at setup instead of at attach (1519)
6197 - common AST error handler, don't evict client on completion race (3145)
6198 - zero nameidata in detach_mnt in 2.6 (3118)
6199 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6200 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6201 - initialize RPC timeout timer earlier for 2.6 (3219)
6202 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6203 - print nal/nid for unknown nid (3258)
6204 - additional checks for oscc recovery before doing precreate (3284)
6205 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6206 - don't crash in mdc_close for bad permissions on open (3285)
6207 - zero i_rdev for non-device files (3147)
6208 - clear page->private before handing to FS, better assertion (3119)
6209 - tune the read pipeline (3236)
6210 - fix incorrect decref of invalidated dentry (2350)
6211 - provide read-ahead stats and refine rpc in flight stats (3328)
6212 - don't hold journal transaction open across create RPC (3313)
6213 - update atime on MDS at close time (3265)
6214 - close LDAP connection when recovering to avoid server load (3315)
6215 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6216 - don't leak open file on MDS after open resend (3325)
6217 - serialize filter_precreate and filter_destroy_precreated (3329)
6218 - loop device shouldn't call sync_dev() for nul device (3092)
6219 - clear page cache after eviction (2766)
6220 - resynchronize MDS->OST in background (2824)
6221 - refuse to mount the same filesystem twice on same mountpoint (3394)
6222 - allow llmount to create routes for mounting behind routers (3320)
6223 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6224 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6225 - fix rare problem with rename on htree directories (3417)
6227 - allow default OST striping configuration per directory (1414)
6228 - fix compilation for qswnal for 2.6 kernels (3125)
6229 - increase maximum number of MDS request buffers for large systems
6230 - change liblustreapi to be useful for external progs like lfsck (3098)
6231 - increase local configuration timeout for slow disks (3353)
6232 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6234 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6237 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6238 - don't overwrite extent policy data in reply if lock was blocked (2901)
6239 - drop filter export grants atomically with removal from device (2663)
6240 - del obd_self_export from work_list in class_disconnect_exports (2908)
6241 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6242 - swab reply message in mdc_close, other PPC fixes (2464)
6243 - fix destroying of named logs (2325)
6244 - overwrite old logs when running lconf --write_conf (2264)
6245 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6246 - fix race in target_handle_connect (2898)
6247 - mds_reint_create() should take same inode create lock (2926)
6248 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6249 - don't close files for self_export to avoid uninitialized obd (2936)
6250 - allow MDS with the same name as client node (2939)
6251 - hold dentry reference for closed log files for unlink (2325)
6252 - reserve space for all logs during transactions (2059)
6253 - don't evict page beyond end of stripe extent (2925)
6254 - don't oops on a deleted current working directory (2399)
6255 - handle hard links to targets without a parent properly (2517)
6256 - don't dereference NULL lock when racing during eviction (2867)
6257 - don't grow lock extents when lots of conflicting locks (2919)
6259 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6262 - account for cache space usage on clients to avoid data loss (974)
6263 - lfsck support in lustre kernel code (2349)
6264 - reduce journal credits needed for BRW writes (2370)
6265 - orphan handling to avoid losing space on client/server crashes
6266 - ptlrpcd can be blocked, stopping ALL progress (2477)
6267 - use lock value blocks to assist in proper KMS, faster stat (1021)
6268 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6269 - recovery for initial connections (2355)
6270 - fixes for mds_cleanup_orphans (1934)
6271 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6272 - block all file creations until orphan recovery completes (1901)
6273 - client remove rq_connection from request struct (2423)
6274 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6275 - recovery timer race (2670)
6276 - mdc_close recovey bug (2532)
6277 - ptlrpc cleanup bug (2710)
6278 - mds timeout on local locks (2588)
6279 - namespace lock held during RPCs (2431)
6280 - handle interrupted sync write properly (2503)
6281 - don't try to handle a message that hasn't been replied to (2699)
6282 - client assert failure during cleanup after abort recovery (2701)
6283 - leak mdc device after failed mount (2712)
6284 - ptlrpc_check_set allows timedout requests to complete (2714)
6285 - wait for inflight reqs when ptlrpcd finishes (2710)
6286 - make sure unregistered services are removed from the srv_list
6287 - reset bulk XID's when resending them (caught by 1138 test)
6288 - unregister_bulk after timeout
6289 - fix lconf error (2694)
6290 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6291 - readahead locks pages, leaves pending causing memory pressure (2673)
6292 - increase OST request buffers to 4096 on large machines (2729)
6293 - fix up permission of existing directories in simple_mkdir (2661)
6294 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6295 - don't assign transno to errored transactions (2742)
6296 - don't delete objects on OST if given a bogus objid from MDS (2751)
6297 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6298 - if rq_no_resend, then timeout request after recovery (2432)
6299 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6300 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6301 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6302 - MDS recovery shouldn't delete objects that it has given out (2730)
6303 - if enqueue arrives after completion, don't clobber LVB (2819)
6304 - don't unlock pages twice when trigger_group_io returns error (2814)
6305 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6306 - don't write pages to disk if there was an error (1450)
6307 - don't ping imports that have recovery disabled (2676)
6308 - take buffered bytes into account when balancing socknal conn (2817)
6309 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6310 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6311 - remove little-endian swabbing of llog records (1987)
6312 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6313 - retry reposting request buffers if they fail (1191)
6314 - grow extent at grant time to avoid granting a revoked lock (2809)
6315 - lock revoke doesn't evict page if covered by a second lock (2765)
6316 - disable VM readahead to avoid reading outside lock extents (2805)
6318 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6319 - updated kernel patches for hp-2.4.20 kernel (2681)
6321 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6324 - fix truncated write corruption (2366)
6325 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6327 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6328 - ensure len(uuid) < 37 in lmc (1171)
6329 - fix ia64 OOPS in llog_test (2255)
6330 - zero end of page at obdfilter for partial page writes (2648)
6331 - don't leave stale dentries around after renames (bug 2428)
6332 - fix timeouts when evicting a client with a single lock held (2642)
6333 - set deadline for the initial HELLO message to drain (2634)
6334 - print out dotted-quad IP addresses in the socknal (2302)
6336 - additional debugging for MDS client eviction problem (2443)
6337 - fix mkfsoptions support for osts (2603, 2604)
6339 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6342 - add series for the vanilla 2.6.0 kernel
6343 - add series for the vanilla 2.4.24 kernel
6344 - add series for a cray x86/64 UL kernel drop
6345 - fix xattr patches for the vanilla 2.4.19 series
6347 - generate true UUIDs in lmc (1171)
6348 - have portals stack dumping break in UML (2466)
6349 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6350 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6351 - wake callback waiting threads on client eviction (2460)
6352 - Add --ptldebug and --subsystem to lmc (1719)
6353 - update assertion to allow safe interrupt allocation
6354 - set rq_no_resend for cancel requests (2432)
6355 - recalculate ptlrpcd timeout after resend (2494)
6356 - call vfs_rmdir when removing pending directories (2368)
6357 - fix renaming a file to itself (2429)
6358 - lmc creates a default one-stripe lov (2454)
6359 - expand procfs space to handle large clusters (2326)
6360 - increase UML stack to avoid overflow
6361 - update lconf's list of debug and subsystem masks
6362 - fix lfs find --obd (2510)
6363 - /proc tunable for disabling filter read caching (2591)
6364 - stop rpm packages from altering slapd.conf (2301)
6365 - disable nagle in the socknal under 0conf (2578)
6366 - choose mds inode size based on stripe count (2572)
6367 - fix kernel-source rpm problems (2516)
6369 - add --disable-doc to avoid pdf generation (2421)
6370 - update documentation, tests, type-os, comments
6371 - avoid format warnings on ia64
6372 - remove the TOE NAL
6373 - tiny code cleanups by removing unused fields
6375 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6378 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6379 - remove the most blatant lies from BUILDING file (2371)
6380 - change default debug level to reasonable production setting
6381 - reduce client side cache size to reduce cache flush time
6382 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6383 - make TCP zerocopy and pinger support enabled by default (2476)
6384 - sync writes completed after process exits caused crashes (2319)
6385 - maintain correct mount count on the MDS (2356)
6386 - backout 1557, because 2316 wasn't really fixed
6387 - better file I/O statistics gathering in /proc
6388 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6389 - another kernel patch to fix zero-copy TCP function export
6390 - don't take duplicate lock when processing re-sent getattr (2420)
6391 - lctl uses obd_self_export instead of creating new conn (2353)
6392 - MDS/OST recovery case which requires object creation asserted (2425)
6393 - move lfs from /usr/sbin to /usr/bin in packages
6394 - fix race between mds_client_add and mds_client_free (2417)
6395 - use kmalloc instead of slabs in portals (2430)
6396 - don't create duplicate records when a failover MDS is present (2442)
6397 - remove unnecessary mount age check (2332)
6398 - don't remove directory inodes from locks prematurely (2451)
6399 - don't break if MDS service name is the same as hostname (2103)
6400 - fix races in client write RPC generation when cache full (2482)
6402 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6405 - remove now-unused request->rq_obd (278)
6406 - if an allocation fails, print out how much memory we've used (1933)
6407 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6408 - add the "configurable stack size" patch to most series files (1256)
6409 - ability to write large log records, for 100+ OST configs (2306)
6410 - fix NULL deref when filter_prep fails (2314)
6411 - fix operator precedence error in filter_sync
6412 - dynamic allocation of socknal TX descriptors (2315)
6413 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6414 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6415 - ensure that CWARN messages in Portals always get to the syslog
6416 - __init/__exit are not for prototype decls (ldlm_init/exit)
6417 - x86-64 compile warning fixes
6418 - fix gateway LMC keyword conflict (2318)
6419 - fix MDS lock inversions in getattr/reint paths (1844)
6420 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6421 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6422 - DLM race condition prevented some lock evictions (2328)
6423 - ENOMEM detection and retry on socknal sends (2230)
6424 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6425 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6426 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6427 - better error messages when a client is rejected during recovery (1505)
6428 - avoid cancelling locks which were never granted, after failure (2330)
6429 - fix i_sem/journal inversion in mds_client_add (2333)
6430 - fix truncate/getattr lock cycle deadlock (2334)
6431 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6432 - fix two rare exit paths which could leak an l_lock() ref (2321)
6433 - fencepost error in MDS/OST orphan recovery (2226)
6434 - make log record alignment 8 bytes (1988)
6435 - lstripe now fails when requested offset > ost_count (2237)
6436 - ensure that all kernel series have a complete list.h (1607)
6437 - fix crashes in special-file operations (2316)
6438 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6440 - allow configurable automake binary, for testing new versions
6441 - small update to the lfs documentation
6443 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6445 * fix negative export reference count in fsfilt_sync (2312)
6447 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6448 * release candidate 0.9.1
6450 - orphans are moved into the PENDING directory for possible recovery
6451 - replayed opens now open by fid for orphan/rename safety (1042)
6452 - last close of an orphan inode generates a transno (683)
6453 - chdir() and mount() now pin the directory entry (1020)
6454 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6455 - discard very old requests without processing them (1502)
6456 - remove some common, well-understood CERRORs (1505)
6457 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6458 - clear "grant" flags in OST replies until OST grant code lands (1644)
6459 - fix read performance by not clobbering i_blksize on client (1598)
6460 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6461 - make LRU size a /proc tunable, clears locks when reduced (707)
6462 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6463 - server locks take references on exports now (1558)
6464 - build fixes for 2.4.20-rh trees (1663)
6465 - return an error from lov_create if all OSCs are inactive (1751)
6466 - fix import levels when a reconnect happens without a timeout (1597)
6467 - exit early from mds_open if we get a lookup error (1749)
6468 - partial page read at EOF wouldn't wait for disk before sending (1642)
6469 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6470 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6471 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6472 - write/truncate lock inversion (1639)
6473 - Don't auto-load obdclass, portals modules during cleanup (1495)
6474 - fix timestamps from jumping to "now" (1763)
6475 - extra journal assertions (1648)
6476 - add an extra multiunlink test (1771)
6477 - fix read_record/write_record API (1776)
6478 - fix leak of offset_extent, possible incorrect i_size later (1772)
6479 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6480 - Debugging for the kqswnal_get_idle_tx problems (1820)
6481 - Allow recovery to be attempted multiple times (1536)
6482 - Write out MDS last_rcvd file after it is first created (1600)
6483 - Fix tx_descriptor leak in failed transmit situations (1827)
6484 - ext3 journaling fixes for assertion failure after IO error (1871)
6485 - class_export_put() on freed export after completion AST error (1896)
6486 - Fix revalidate looping in VFS (1322)
6487 - Don't access a freed export during MDS_REINT timeout (1521)
6488 - Add open-unlink recovery support on the MDS (1673,1764)
6489 - Return an error if no MDS data was read from last_rcvd (1946)
6490 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6491 - Don't setup a disk device that doesn't match exported UUID (317)
6492 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6493 - avoid committing NULL handle in force close
6494 - local.sh is now a one-stripe LOV configuration
6495 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6496 - fix ext3 htree duplicate directory entry corruption (1516)
6497 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6498 - update to new LOV EA format (2097)
6499 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6501 2003-06-15 Phil Schwan <phil@clusterfs.com>
6504 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6505 - per-import/export recovery handling (958, 931, 959)
6506 - multiple last-rcvd slots, for serving multiple FSes (949)
6507 - connections are again shared between multiple imp/exports (963, 964)
6508 - "umount -f" would hang if any requests needed to be sent (393, 978)
6509 - avoid pinning large req buffer by copying for queued messages (989)
6510 - add "uuid" to "lctl device" command to help upcalls (991)
6511 - "open" RPCs with transnos would confuse recovery counters (1037)
6512 - do proper endian conversion of all wire messages (288, 340, 891)
6513 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6514 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6515 - fix ptlrpc_connection leak in target_handle_connect (1174)
6516 - fix import refcounting bug in OST and MDS cleanup (1134)
6517 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6518 - fix very unlikely obd_types race condition (501)
6519 - remove osc_open hack for echo_client (1187)
6520 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6521 - a failure in read_inode2 leads to deadlock (1139)
6522 - cancel ack-locks as soon as transaction is committed (1072)
6523 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6524 - make sure to commitrw after any preprw to avoid deadlock (1162)
6525 - failing to execute a file in a lustre FS would lock inode (1203)
6526 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6527 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6528 - obd_uuid2tgt would walk off the end of the list (1255)
6529 - on IA64 the osc would give portals incorrect bulk size (1258)
6530 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6531 - fix lock inversion caused by new llite matching code (1282)
6532 - limit the number of dirty pages on a client to 10MB (1286)
6533 - timed out locks were not being corrected cancelled (1289)
6534 - fix O_DIRECT above 4GB on IA-32 (1292)
6535 * major user-visible changes
6536 - fail out/fail over policy now controlled by the upcall (993)
6538 - add OBD_PING to check server availability and failure (954)
6539 - lustre messages are now sent in sending host order (288, 340, 891)
6540 - add eadatalen to MDS getattr reply (340)
6541 - OST read replies may contain second buffer, with per-page status (593)
6543 2003-03-11 Phil Schwan <phil@clusterfs.com>
6546 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6547 - failed open()s could cause deadlock; fixed (867, 869)
6548 - stop cancelling OST locks when files are closed (481)
6549 - overlapping XID spaces caused network corruption (851, 853)
6550 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6551 - setattr_raw added, to do single-RPC, server-side setattrs
6552 - lmc/lconf syntax change for OST UUIDs
6553 - fix crashy race condition between ptlrpc_free_req and osc_close
6554 - don't use request in mdc_enqueue if we hit a timeout (889)
6555 - don't set the inode i_size for regular files from the MDS (896)
6556 - handle out of order completion AST (842)
6557 - don't LBUG if a lock request times out after receiving AST (913)
6558 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6559 - if a bad lock AST arrives, send an error instead of dropping entirely
6560 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6561 - fix leak in bulk IO when only partially completed (899, 900, 926)
6562 - fix O_DIRECT for ia64 (55)
6563 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6564 - C-z after timeout could hang a process forever; fixed (977)
6566 - client-side I/O cache (678, 924, 929, 941, 970)
6568 - READPAGE and SETATTRs which don't take server-side locks get
6571 2003-02-11 Phil Schwan <phil@clusterfs.com>
6574 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6575 - Fix fsfilt_extN_readpage() to read a full page of directory
6576 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6577 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6578 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6579 - Most intent operations were removed, in favour of a new RPC mode
6580 that does a single RPC to the server and bypasses most of the VFS
6581 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6582 - Aggressively cancel local locks on DLM servers
6583 - mds_reint_unlink sends EA to the client if it's the last nlink.
6584 client uses that EA to unlink OST objects.
6585 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6586 - recursive symlinks were fixed (439)
6587 - fixed NULL deref in DEBUG_REQ
6588 - filter_update_lastobjid no longer calls sync, which annoyed extN
6589 - fixed multi-client small-writes to a single file problem (445)
6590 - fixed mtime updates during file writes (607)
6591 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6592 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6593 - fixed Linux OST size reporting problem (444, 656)
6594 - OST now updates object mtime with writes or setattr (607, 619)
6595 - client verifies file size before zeroing page past EOF (445)
6596 - OST now writes last allocated objid to disk with allocation (108)
6597 - LOV on echo now works (409)
6599 - mds_reint_unlink sends a new buffer, with the EA included. this
6600 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6601 set if a regular file was being unlinked, and it was the last link
6602 - use PtlGet from the target for bulk writes (315)
6603 - OST now updates object mtime with writes or setattr (607, 619)
6604 - LDLM now has a grant-time callback to revalidate locked items, if
6606 - Many MDS operations were reorganized to combat race conditions
6608 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6609 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6611 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6614 - Fully reactivate OST imports after reconnection (512, others)
6615 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6616 - More graceful error handling for truncating on dead OST (515)
6617 - Don't error out unless we're actually accessing dead stripes (474)
6618 - Fix garbage sizes when stripes are missing (410)
6619 - LRU counters were broken, causing constant lock purge (433, 432)
6620 - garbage on read from stripes with failed OSTs (441)
6621 - mark OSCs as active before reconnecting during recovery (438)
6622 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6623 - lfind did not preserve OST order in output (443)
6624 - symlinks cause hung clients, incorrect data (439)
6625 - stop dereferencing request after dropping refcount (457)
6626 - don't LASSERT(spin_is_locked) on non-SMP (455)
6627 - fixes for many rename() bugs
6628 - fstat didn't correctly synchronize attributes (399)
6629 - server must handle lock cancellation during blocking AST prep (487)
6630 - bulk descriptors were free()d too soon (511)
6631 - fix paths in lconf, which would load incorrect modules (451, 507)
6632 - fix confusing lconf 'host not found' error message (386)
6633 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6634 - fix race condition in mdc_blocking_ast() for inode access (526)
6635 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6636 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6637 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6638 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6639 - fix lstripe in several ways (kernel side) (527)
6640 - fix request leak in ldlm_cli_enqueue (262)
6641 - incorrect OSC was marked inactive after OST failure
6642 - call mds_fs_cleanup before unmounting filesystem (524)
6643 - fix races between taking ns_lock and ldlm_lock_change_resource
6644 - fix races updating LOV export open file list
6645 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6646 - fix recovery NULL deref in ldlm_cli_cancel_unused
6647 - fix some DLM races by using new hash table for lock handles (419)
6648 - permit the client to specify desired inodes, at replay
6649 - duplicate requests when we queue them for replay reintegration
6650 - fix last_rcvd offset calculation
6651 - sync after each recovered transaction, so we always make progress
6652 - never, not always, ERESTART requests without transnos
6653 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6655 - skip replay if the MDS says that the client is already connected
6656 - don't check for a recovery-enabled export to match lctl's UUID
6657 - don't INC_USE_COUNT for phantom exports
6658 - don't crash when cleaning up phantom exports (567)
6659 - don't double-finish or set replay data for errored mdc_open requests
6660 - abort requests when they time out, so we don't get old replies
6661 - send/receive replies for AST messages again
6662 - if the client says that it doesn't have the lock, cancel it on the
6664 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6665 mark it as destroyed, it will all work out when decref is called
6666 - fix module use counts (22, 581)
6668 - ASTs now expect a reply (server cancels lock on error reply)
6670 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6673 - fix many simultaneous client startup (392)
6674 - fix dentry->d_it clobbering
6675 - credentials weren't being shipped for readdir/getattr operations
6676 - remove invalid assertions triggered during some concurrent MD
6678 - proper Lustre versions added (336, 389)
6679 - fix memory leak for create error case (398)
6680 - fix LOV locking bug that would get cli/srv out of sync
6681 - fix echo client over LOV (409)
6682 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6683 - fix double-O_EXCL intent crash (424)
6684 - avoid sending multiple lock CANCELs (352)
6686 - MDS can do multi-client recovery (modulo bugs in new code)
6688 - many updates, edits, cleanups
6690 2002-11-18 Phil Schwan <phil@clusterfs.com>
6693 - fix null d_it dereference (346)
6694 - fix full OST/dbench hang (333)
6695 - fix permission problem with file removal (286)
6696 - fix removal of OSCs from LOV when they fail
6697 - fix NULL deref during bulk timeout (214)
6698 - fix problems related to multiple filesystems on one MDS (241)
6699 - fixed serious subtle metadata locking bugs
6700 - free locks on clients when inodes are removed due to memory
6702 - fix inode pointer in lock data (285)
6703 - partial support for multiple MDS on a single host (241)
6704 - data locks weren't cancelled at clear_inode time (290, 311)
6705 - intent locks could lead to unbounded lock growth (205)
6706 - added a maximum lock count, an LRU list, and a flusher
6707 - fix multiple rename (365)
6708 - properly abstracted the echo client
6709 - OSC locked 1 byte too many; fixed
6710 - rewrote brw callback code:
6711 - fixed recovery bugs related to LOVs (306)
6712 - fixed too-many-pages-in-one-write crash (191)
6713 - fixed (again) crash in sync_io_timeout (214)
6714 - probably fixed callback-related race (385)
6716 - Add capability to MDS protocol
6717 - LDLM cancellations and callbacks on different portals
6719 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6722 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6723 - defer open object destruction to close time (601981, 138)
6724 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6725 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6726 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6727 - don't call obd_setattr in ll_file_release for destroyed objects
6728 * protocol change to lustre_msg: move |version| and add |flags|
6729 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6730 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6731 * added replay of create, unlink, link and rename operations during
6732 MDS failover; recovery should be much more robust now
6733 * remove failed OSCs from LOVs (only lov_create uses this so far)
6734 * the lustre-HOWTO was brought (more) up to date (582544)
6736 2002-10-23 Phil Schwan <phil@clusterfs.com>
6739 - in-use dentries weren't being reused properly (617851)
6740 - prevent multiple LDLM setup (599178)
6741 - fix LOV size calculations for truncate (617853)
6742 - fix client handling of MDS intent errors (POSIX)
6743 - fix permission bug in lovstripe.c test (624321)
6744 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6745 - truncate past end of file could corrupt data
6746 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6747 - a race in recovery could return ETIMEDOUT to apps (623947)
6748 - building outside the source directory was fixed
6749 * the lustre-HOWTO was brought (more) up to date (582544)
6750 * major progress was made on recovery functionality
6752 2002-10-10 Phil Schwan <phil@clusterfs.com>
6755 - recovery deadlock fix
6756 - rm -rf causes LBUG fix (617817)
6757 - file open by multiple tasks fix (618962)
6758 - directory permissions bugs (602707 and 620007)
6759 - journal_stop fixed with locking (611313)
6760 - O_APPEND failures resolved (618273, perhaps 614459)
6761 - lconf PATH fix (619770)
6762 - IA64 build fix (621450)
6763 - RPC buffer sizes scale with amount of memory
6765 2002-10-01 Phil Schwan <phil@clusterfs.com>
6768 - locks would be cancelled without throwing away data pages,
6769 resulting in inconsistent data (605627)
6770 - inode attributes were not always being refreshed (605627, 612449)
6771 - lconf now continues to cleanup after lctl reports an error
6772 - MDS now enforces user permissions (602707)
6773 - lprocfs cleanup fixed, but not yet enabled (614157)
6774 - fixed infinite server hang, should a client not respond to an AST
6775 - avoid going into recovery if user calls readlink() with a buffer
6776 that's too small (613941)
6777 - AST RPCs no longer require replies (614867) -- this may be changed
6778 - don't crash server if client sends an IOV that's too big (611336)
6779 - fixed lock conversion deadlock (611892)
6780 - fixed the following of symlinks (614622)
6781 * recovery: the server can remove locks from a client that dies, other
6782 clients can make progress
6783 * more extN patch fixes
6784 * compile-time configurable ptlrpc buffer allocations
6786 - collaborative read cache document
6787 - Lustre Lite Performance CDR document-in-progress
6789 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6792 - fix typo in patch-2.4.18
6794 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6797 - clear ptlrpc request each time in handle_incoming_request()
6798 - unlink of files now destroys the object on the OST
6800 2002-09-19 Peter Braam <braam@clusterfs.com>
6802 * add hard link support
6803 * change obdfile creation method
6804 * kernel patch changed
6806 2002-09-19 Peter Braam <braam@clusterfs.com>
6809 - stack overflow bug in extN fixed
6811 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6813 * documentation updates
6814 - add man pages for config tools
6815 - update tests/README to describe testing with new config tools
6816 - finish metadata API descriptions
6817 * bug fixes and cleanups
6818 - statfs workaround for 16TB limit
6819 - LOV stripe allocation improved, can stripe on subset of OSTs
6820 - LOV file size/IO offset was wrong for files > 4GB in size
6821 - object EA data was being dropped, caused files to be unreadable
6822 - memory overflow with non-LOV OST caused memory corruption
6823 - fixed regression tests to work with new config tools, obdfilter
6824 - fixed bug when directory size became larger than 1 block
6825 - fixed bug (for single client case) when PWD was deleted
6826 - invalidate local directory pages when doing intent-based ops
6827 - avoid LDLM oops when lock callback contained bad data
6829 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6831 * documentation updates
6832 * bug fixes and cleanups
6833 - configuration tools
6836 - 64-bit compile warnings
6837 - 64-bit internal statfs data
6839 * test_brw on persistent OST devices
6841 * lprocfs (disabled)
6843 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6845 * documentation updates
6846 * bug fixes and cleanups
6847 * configuration tools
6849 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6852 * many small fixes to 0_5_4
6853 * io/network handling
6854 * thinkos in MDS operations
6856 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6859 * crucial basic fixes to 0.5.3
6860 * IOR, Iozone work over Elan
6863 2002-08-07 Phil Schwan <phil@clusterfs.com>
6864 * version 0_5_3, our first alpha
6865 * we use the new Portals iovs
6866 * documentation updates
6867 * bug fixes and cleanups
6868 * small changes in the DLM wire protocol
6870 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6871 * version 0_5_1 with some initial stability,
6872 * locking on MD and file I/O.
6873 * documentation updates
6874 * several bug fixes since 0.5.0
6875 * small changes in wire protocol
6877 2002-07-18 Phil Schwan <phil@clusterfs.com>
6879 * delivered as Lustre Light Alpha
6880 * fixed a crash after handling invalid MDS requests
6881 * fixed directory pages for architectures with non-4k pages sizes
6883 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6884 * release version v0_4_4
6885 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6887 * Fixes a number of interruption problems with OST operations.
6888 * Update documentation for portals header changes
6889 * Move all wire protocol structs/defines to lustre_idl.h
6890 * Fixes symlink length bug.
6891 * Add tcpdump to repository.
6893 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6894 * release version v0_4_3
6895 * Fixes statfs for inodes on extN.
6896 * Fixes bug in runtests which would delete /etc/hosts.
6897 * Use 64-bit object IDs wherever possible (not into VFS though)
6898 Remove ost_get_info, which is unused by lustre, and out of date.
6900 2002-07-03 Peter Braam <braam@clusterfs.com>
6901 * release version v0_4_2 Fixes a lookup error (type not passed)
6902 * move forward to head of Portals
6903 * move forward to latest Lustre kernel
6905 2002-06-25 Peter Braam <braam@clusterfs.com>
6906 * release version v0_4_1. Hopefully stable on single node use.