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 5.14.0-284.18.1.el9 (RHEL9.2)
7 5.14.0-162.23.1.el9 (RHEL9.1)
8 4.18.0-477.15.1.el8 (RHEL8.8)
9 * Other server kernels known to build and work at some point (others may also work):
10 3.10.0-862.14.4.el7 (RHEL7.5)
11 3.10.0-957.27.2.el7 (RHEL7.6)
12 3.10.0-1062.18.1.el7 (RHEL7.7)
13 3.10.0-1127.19.1.el7 (RHEL7.8)
14 3.10.0-1160.88.1.el7 (RHEL7.9)
15 4.18.0-80.11.2.el8 (RHEL8.0)
16 4.18.0-147.8.1.el8 (RHEL8.1)
17 4.18.0-193.28.1.el8 (RHEL8.2)
18 4.18.0-240.22.1.el8 (RHEL8.3)
19 4.18.0-305.25.1.el8 (RHEL8.4)
20 4.18.0-348.23.1.el8 (RHEL8.5)
21 4.18.0-372.32.1.el8 (RHEL8.6)
22 4.18.0-425.10.1.el8 (RHEL8.7)
23 4.4.120-92.70 (SLES12 SP2)
24 4.4.180-94.100 (SLES12 SP3)
25 4.4.0-131 (Ubuntu 16.04)
26 4.15.0-32 (Ubuntu 18.04)
27 5.4.0-48 (Ubuntu 20.04)
28 vanilla linux 5.4.0 (ZFS + ldiskfs)
29 vanilla linux 5.4.21 (ZFS + ldiskfs)
30 vanilla linux 5.4.136 (ZFS + ldiskfs)
31 5.10.0-60.79.0.103.oe2203 (openEuler 22.03 LTS)
32 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
33 * Client primary kernels built and tested during release cycle:
34 5.14.0-284.18.1.el9 (RHEL9.2)
35 5.14.0-162.23.1.el9 (RHEL9.1)
36 4.18.0-477.15.1.el8 (RHEL8.8)
37 5.4.0-37 (Ubuntu 20.04)
38 5.14.21-150400.24.46 (SLES15 SP4)
39 * Other clients known to build on these kernels at some point (others may also work):
40 3.10.0-862.14.4.el7 (RHEL7.5)
41 3.10.0-957.27.2.el7 (RHEL7.6)
42 4.14.0-49.13.1.el7a (RHEL7.5)
43 4.14.0-115.2.2.el7a (RHEL7.6)
44 3.10.0-1062.18.1.el7 (RHEL7.7)
45 3.10.0-1127.19.1.el7 (RHEL7.8)
46 3.10.0-1160.88.1.el7 (RHEL7.9)
47 4.18.0-80.11.2.el8 (RHEL8.0)
48 4.18.0-147.8.1.el8 (RHEL8.1)
49 4.18.0-193.28.1.el8 (RHEL8.2)
50 4.18.0-240.22.1.el8 (RHEL8.3)
51 4.18.0-305.25.1.el8 (RHEL8.4)
52 4.18.0-348.23.1.el8 (RHEL8.5)
53 4.18.0-372.32.1.el8 (RHEL8.6)
54 4.18.0-425.10.1.el8 (RHEL8.7)
55 5.14.0-70.30.1.el9 (RHEL9.0)
56 4.4.120-92.70 (SLES12 SP2)
57 4.4.180-94.100 (SLES12 SP3)
58 4.12.14-95.48 (SLES12 SP4)
59 4.12.14-122.133 (SLES12 SP5)
60 4.12.14-197.75 (SLES15 SP1)
61 5.3.18-24.96 (SLES15 SP2)
62 5.3.18-150300.59.93 (SLES15 SP3)
63 4.4.0-131 (Ubuntu 16.04)
64 4.15.0-48 (Ubuntu 18.04)
65 5.8.0-53 (Ubuntu 20.04.2 HWE)
66 5.11.0-31 (Ubuntu 20.04.3 HWE)
67 5.11.0 (vanilla kernel.org)
68 5.10.0 (openEuler 22.03 LTS)
69 * Recommended e2fsprogs version: 1.47.0-wc1 or newer
70 * Recommended ZFS version: 2.1.11
71 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
72 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
73 stack overflow. For more information, please refer to bugzilla 17630.
74 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
75 the re-exporting nodes to be mounted with "32bitapi" mount option
77 --------------------------------------------------------------------------------
82 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
83 for currently supported client and server kernel versions.
84 * Server primary kernels built and tested during release cycle:
85 4.18.0-348.2.1.el8 (RHEL8.5)
86 * Other server kernels known to build and work at some point (others may also work):
87 3.10.0-862.14.4.el7 (RHEL7.5)
88 3.10.0-957.27.2.el7 (RHEL7.6)
89 3.10.0-1062.18.1.el7 (RHEL7.7)
90 3.10.0-1127.19.1.el7 (RHEL7.8)
91 3.10.0-1160.49.1.el7 (RHEL7.9)
92 4.18.0-80.11.2.el8 (RHEL8.0)
93 4.18.0-147.8.1.el8 (RHEL8.1)
94 4.18.0-193.28.1.el8 (RHEL8.2)
95 4.18.0-240.22.1.el8 (RHEL8.3)
96 4.18.0-305.25.1.el8 (RHEL8.4)
97 4.4.120-92.70 (SLES12 SP2)
98 4.4.180-94.100 (SLES12 SP3)
99 4.4.0-131 (Ubuntu 16.04)
100 4.15.0-32 (Ubuntu 18.04)
101 5.4.0-48 (Ubuntu 20.04)
102 vanilla linux 5.4.0 (ZFS + ldiskfs)
103 vanilla linux 5.4.21 (ZFS + ldiskfs)
104 vanilla linux 5.4.136 (ZFS + ldiskfs)
105 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
106 * Client primary kernels built and tested during release cycle:
107 4.18.0-348.2.1.el8 (RHEL8.5)
108 5.4.0-37 (Ubuntu 20.04)
109 5.3.18-59.27 (SLES15 SP3)
110 * Other clients known to build on these kernels at some point (others may also work):
111 3.10.0-862.14.4.el7 (RHEL7.5)
112 3.10.0-957.27.2.el7 (RHEL7.6)
113 4.14.0-49.13.1.el7a (RHEL7.5)
114 4.14.0-115.2.2.el7a (RHEL7.6)
115 3.10.0-1062.18.1.el7 (RHEL7.7)
116 3.10.0-1127.19.1.el7 (RHEL7.8)
117 3.10.0-1160.49.1.el7 (RHEL7.9)
118 4.18.0-80.11.2.el8 (RHEL8.0)
119 4.18.0-147.8.1.el8 (RHEL8.1)
120 4.18.0-193.28.1.el8 (RHEL8.2)
121 4.18.0-240.22.1.el8 (RHEL8.3)
122 4.18.0-305.25.1.el8 (RHEL8.4)
123 4.4.120-92.70 (SLES12 SP2)
124 4.4.180-94.100 (SLES12 SP3)
125 4.12.14-95.48 (SLES12 SP4)
126 4.12.14-122.91 (SLES12 SP5)
127 4.12.14-197.75 (SLES15 SP1)
128 5.3.18-24.96 (SLES15 SP2)
129 4.4.0-131 (Ubuntu 16.04)
130 4.15.0-48 (Ubuntu 18.04)
131 5.8.0-53 (Ubuntu 20.04.2 HWE)
132 5.11.0-31 (Ubuntu 20.04.3 HWE)
133 5.11.0 (vanilla kernel.org)
134 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
135 * Recommended ZFS version: 2.0.7
136 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
137 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
138 stack overflow. For more information, please refer to bugzilla 17630.
139 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
140 the re-exporting nodes to be mounted with "32bitapi" mount option
142 --------------------------------------------------------------------------------
146 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
147 for currently supported client and server kernel versions.
148 * Server primary kernels built and tested during release cycle:
149 3.10.0-1062.18.1.el7 (RHEL7.7)
150 3.10.0-1127.19.1.el7 (RHEL7.8)
151 3.10.0-1160.6.1.el7 (RHEL7.9)
152 * Other server kernels known to build and work at some point (others may also work):
153 3.10.0-862.14.4.el7 (RHEL7.5)
154 3.10.0-957.27.2.el7 (RHEL7.6)
155 4.18.0-80.11.2.el8 (RHEL8.0)
156 4.18.0-147.8.1.el8 (RHEL8.1)
157 4.18.0-193.28.1.el8 (RHEL8.2)
158 4.18.0-240.1.1.el8 (RHEL8.3)
159 4.4.120-92.70 (SLES12 SP2)
160 4.4.180-94.100 (SLES12 SP3)
161 4.4.0-131 (Ubuntu 16.04)
162 4.15.0-32 (Ubuntu 18.04)
163 5.4.0-48 (Ubuntu 20.04)
164 vanilla linux 5.4.0 (ZFS + ldiskfs)
165 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
166 * Client primary kernels built and tested during release cycle:
167 3.10.0-1062.18.1.el7 (RHEL7.7)
168 3.10.0-1127.19.1.el7 (RHEL7.8)
169 3.10.0-1160.6.1.el7 (RHEL7.9)
170 4.12.14-95.48 (SLES12 SP4)
171 4.12.14-122.20 (SLES12 SP5)
172 4.12.14-197.75 (SLES15 SP1)
173 5.3.18-24.24 (SLES15 SP2)
174 4.15.0-48 (Ubuntu 18.04)
175 * Other clients known to build on these kernels at some point (others may also work):
176 3.10.0-862.14.4.el7 (RHEL7.5)
177 3.10.0-957.27.2.el7 (RHEL7.6)
178 4.14.0-49.13.1.el7a (RHEL7.5)
179 4.14.0-115.2.2.el7a (RHEL7.6)
180 4.18.0-80.11.2.el8 (RHEL8.0)
181 4.18.0-147.8.1.el8 (RHEL8.1)
182 4.18.0-193.28.1.el8 (RHEL8.2)
183 4.18.0-240.1.1.el8 (RHEL8.3)
184 4.4.120-92.70 (SLES12 SP2)
185 4.4.180-94.100 (SLES12 SP3)
186 4.4.0-131 (Ubuntu 16.04)
187 5.4.0-37 (Ubuntu 20.04)
188 5.4.0 (vanilla kernel.org)
189 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
190 * Recommended ZFS version: 2.0.0
191 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
192 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
193 stack overflow. For more information, please refer to bugzilla 17630.
194 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
195 the re-exporting nodes to be mounted with "32bitapi" mount option
197 --------------------------------------------------------------------------------
201 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
202 for currently supported client and server kernel versions.
203 * Server primary kernels built and tested during release cycle:
204 3.10.0-1062.1.1.el7 (RHEL7.7)
205 * Other server kernels known to build and work at some point (others may also work):
206 3.10.0-862.14.4.el7 (RHEL7.5)
207 3.10.0-957.27.2.el7 (RHEL7.6)
208 4.18.0-80.7.1.el8 (RHEL8.0)
209 4.4.120-92.70 (SLES12 SP2)
210 4.4.180-94.100 (SLES12 SP3)
211 4.4.0-131 (Ubuntu 16.04)
212 4.15.0-32 (Ubuntu 18.04)
213 vanilla linux 4.6.7 (ZFS only)
214 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
215 * Client primary unpatched kernels built and tested during release cycle:
216 3.10.0-1062.1.1.el7 (RHEL7.7)
217 4.12.14-95.29 (SLES12 SP4)
218 4.15.0-48 (Ubuntu 18.04)
219 * Other clients known to build on unpatched kernels at some point (others may also work):
220 3.10.0-862.14.4.el7 (RHEL7.5)
221 3.10.0-957.27.2.el7 (RHEL7.6)
222 4.14.0-49.13.1.el7a (RHEL7.5)
223 4.14.0-115.2.2.el7a (RHEL7.6)
224 4.18.0-80.7.1.el8 (RHEL8.0)
225 4.4.120-92.70 (SLES12 SP2)
226 4.4.180-94.100 (SLES12 SP3)
227 4.4.0-131 (Ubuntu 16.04)
228 vanilla linux 4.15.0 (ZFS only)
229 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
230 * Recommended ZFS / SPL version: 0.7.13
231 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
232 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
233 stack overflow. For more information, please refer to bugzilla 17630.
234 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
235 the re-exporting nodes to be mounted with "32bitapi" mount option
236 * Note for Ubuntu 14.04.5:
237 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
238 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
240 --------------------------------------------------------------------------------
244 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
245 for currently supported client and server kernel versions.
246 * Server known to build on the following kernels (others may also work):
247 2.6.32-696.18.7.el6 (RHEL6.9)
248 3.10.0-862.14.4.el7 (RHEL7.5)
249 3.10.0-957.el7 (RHEL7.6)
250 4.4.120-92.70 (SLES12 SP2)
251 4.4.162-94.69 (SLES12 SP3)
252 4.4.0-131 (Ubuntu 16.04)
253 4.15.0-32 (Ubuntu 18.04)
254 vanilla linux 4.6.7 (ZFS only)
255 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
256 * Client known to build on unpatched kernels (others may also work):
257 2.6.32-696.18.7.el6 (RHEL6.9)
258 2.6.32-754.6.3.el6 (RHEL6.10)
259 3.10.0-862.14.4.el7 (RHEL7.5)
260 3.10.0-957.el7 (RHEL7.6)
261 4.14.0-49.13.1.el7a (RHEL7.5)
262 4.14.0-115.2.2.el7a (RHEL7.6)
263 4.4.120-92.70 (SLES12 SP2)
264 4.4.162-94.69 (SLES12 SP3)
265 4.4.0-131 (Ubuntu 16.04)
266 4.15.0-32 (Ubuntu 18.04)
267 vanilla linux 4.15.0 (ZFS only)
268 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
269 * Recommended ZFS / SPL version: 0.7.9
270 * Tested with ZFS / SPL version: 0.7.9
271 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
272 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
273 stack overflow. For more information, please refer to bugzilla 17630.
274 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
275 the re-exporting nodes to be mounted with "32bitapi" mount option
276 * Note for Ubuntu 14.04.5:
277 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
278 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
280 --------------------------------------------------------------------------------
282 2018-04-03 Intel Corporation
284 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
285 for currently supported client and server kernel versions.
286 * Server known to build on patched kernels:
287 2.6.32-431.29.2.el6 (RHEL6.5)
288 2.6.32-504.30.3.el6 (RHEL6.6)
289 2.6.32-573.26.1.el6 (RHEL6.7)
290 2.6.32-642.15.1.el6 (RHEL6.8)
291 2.6.32-696.18.7.el6 (RHEL6.9)
292 3.10.0-693.21.1.el7 (RHEL7.4)
293 3.0.101-0.47.71 (SLES11 SP3)
294 3.0.101-107 (SLES11 SP4)
295 3.12.74-60.64.40 (SLES12 SP1)
296 4.4.114-92.64 (SLES12 SP2)
297 4.4.114-94.11 (SLES12 SP3)
298 3.13.0-101 (Ubuntu 14.04, ZFS only)
299 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
300 4.4.0-116 (Ubuntu 16.04)
301 vanilla linux 4.6.7 (ZFS only)
302 * Client known to build on unpatched kernels:
303 2.6.32-431.29.2.el6 (RHEL6.5)
304 2.6.32-504.30.3.el6 (RHEL6.6)
305 2.6.32-573.26.1.el6 (RHEL6.7)
306 2.6.32-642.15.1.el6 (RHEL6.8)
307 2.6.32-696.18.7.el6 (RHEL6.9)
308 3.10.0-693.21.1.el7 (RHEL7.4)
309 3.0.101-0.47.71 (SLES11 SP3)
310 3.0.101-107 (SLES11 SP4)
311 3.12.74-60.64.40 (SLES12 SP1)
312 4.4.114-92.64 (SLES12 SP2)
313 4.4.114-94.11 (SLES12 SP3)
314 3.13.0-101 (Ubuntu 14.04)
315 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
316 4.4.0-116 (Ubuntu 16.04)
318 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
319 * Recommended ZFS / SPL version: 0.7.6
320 * Tested with ZFS / SPL version: 0.7.6
321 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
322 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
323 stack overflow. For more information, please refer to bugzilla 17630.
324 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
325 the re-exporting nodes to be mounted with "32bitapi" mount option
326 * Note for Ubuntu 14.04.5:
327 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
328 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
330 --------------------------------------------------------------------------------
332 2017-07-05 Intel Corporation
334 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
335 for currently supported client and server kernel versions.
336 * Server known to build on patched kernels:
337 2.6.32-431.29.2.el6 (RHEL6.5)
338 2.6.32-504.30.3.el6 (RHEL6.6)
339 2.6.32-573.26.1.el6 (RHEL6.7)
340 2.6.32-642.15.1.el6 (RHEL6.8)
341 2.6.32-696.3.1.el6 (RHEL6.9)
342 3.10.0-514.21.1.el7 (RHEL7.3)
343 3.0.101-0.47.71 (SLES11 SP3)
344 3.0.101-100 (SLES11 SP4)
345 3.12.74-60.64.40 (SLES12 SP1)
346 4.4.59-92.17 (SLES12 SP2)
347 vanilla linux 4.6.7 (ZFS only)
348 * Client known to build on unpatched kernels:
349 2.6.32-431.29.2.el6 (RHEL6.5)
350 2.6.32-504.30.3.el6 (RHEL6.6)
351 2.6.32-573.26.1.el6 (RHEL6.7)
352 2.6.32-642.15.1.el6 (RHEL6.8)
353 2.6.32-696.3.1.el6 (RHEL6.9)
354 3.10.0-514.21.1.el7 (RHEL7.3)
355 3.0.101-0.47.71 (SLES11 SP3)
356 3.0.101-100 (SLES11 SP4)
357 3.12.74-60.64.40 (SLES12 SP1)
358 4.4.59-92.17 (SLES12 SP2)
360 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
361 * Recommended ZFS / SPL version: 0.6.5.9
362 * Tested with ZFS / SPL version: 0.6.5.9
363 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
364 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
365 stack overflow. For more information, please refer to bugzilla 17630.
366 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
367 the re-exporting nodes to be mounted with "32bitapi" mount option
369 --------------------------------------------------------------------------------
371 12-07-2016 Intel Corporation
373 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
374 for currently supported client and server kernel versions.
375 * Server known to build on patched kernels:
376 2.6.32-431.29.2.el6 (RHEL6.5)
377 2.6.32-504.30.3.el6 (RHEL6.6)
378 2.6.32-573.26.1.el6 (RHEL6.7)
379 2.6.32-642.11.1.el6 (RHEL6.8)
380 3.10.0-514.el7 (RHEL7.3)
381 3.0.101-0.47.71 (SLES11 SP3)
382 3.0.101-80 (SLES11 SP4)
383 3.12.62-60.64.8 (SLES12 SP1)
384 vanilla linux 4.5.7 (ZFS only)
385 * Client known to build on unpatched kernels:
386 2.6.32-431.29.2.el6 (RHEL6.5)
387 2.6.32-504.30.3.el6 (RHEL6.6)
388 2.6.32-573.26.1.el6 (RHEL6.7)
389 2.6.32-642.11.1.el6 (RHEL6.8)
390 3.10.0-514.el7 (RHEL7.3)
391 3.0.101-0.47.71 (SLES11 SP3)
392 3.0.101-80 (SLES11 SP4)
393 3.12.62-60.64.8 (SLES12 SP1)
395 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
396 * Recommended ZFS / SPL version: 0.6.5.8
397 * Tested with ZFS / SPL version: 0.6.5.8
398 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
399 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
400 stack overflow. For more information, please refer to bugzilla 17630.
401 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
402 the re-exporting nodes to be mounted with "32bitapi" mount option
404 --------------------------------------------------------------------------------
406 02-29-2016 Intel Corporation
408 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
409 for currently supported client and server kernel versions.
410 * Server known to build on patched kernels:
411 2.6.32-431.29.2.el6 (RHEL6.5)
412 2.6.32-504.30.3.el6 (RHEL6.6)
413 2.6.32-573.12.1.el6 (RHEL6.7)
414 3.10.0-327.3.1.el7 (RHEL7.2)
415 3.0.101-0.47.71 (SLES11 SP3)
416 3.0.101-68 (SLES11 SP4)
417 vanilla linux 4.2.1 (ZFS only)
418 * Client known to build on unpatched kernels:
419 2.6.32-431.29.2.el6 (RHEL6.5)
420 2.6.32-504.30.3.el6 (RHEL6.6)
421 2.6.32-573.12.1.el6 (RHEL6.7)
422 3.10.0-327.3.1.el7 (RHEL7.2)
423 3.0.101-0.47.71 (SLES11 SP3)
424 3.0.101-68 (SLES11 SP4)
427 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
428 * Recommended ZFS / SPL version: 0.6.4.2
429 * Tested with ZFS / SPL version: 0.6.4.2
430 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
431 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
432 stack overflow. For more information, please refer to bugzilla 17630.
433 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
434 the re-exporting nodes to be mounted with "32bitapi" mount option
436 --------------------------------------------------------------------------------
438 03-10-2015 Intel Corporation
440 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
441 for currently supported client and server kernel versions.
442 * Server known to build on patched kernels:
443 2.6.32-431.29.2.el6 (RHEL6.5)
444 2.6.32-504.8.1.el6 (RHEL6.6)
445 3.0.101-0.46 (SLES11 SP3)
446 * Client known to build on unpatched kernels:
447 2.6.32-431.29.2.el6 (RHEL6.5)
448 2.6.32-504.8.1.el6 (RHEL6.6)
449 3.10.0-123.20.1.el7 (RHEL7)
450 3.0.101-0.46 (SLES11 SP3)
451 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
452 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
453 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
454 stack overflow. For more information, please refer to bugzilla 17630.
455 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
456 the re-exporting nodes to be mounted with "32bitapi" mount option
458 Severity : enhancement
460 Description: control OST-index in IDIF via ROCOMPAT flag.
461 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
462 last_rcvd file. For new formatted OST device, it will be auto set;
463 for the case of upgrading from old OST device, you can enable it
464 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
465 enabled, for new created OST-object, its IDIF-in-LMA will contain
466 the OST-index; for the existing OST-object, the OSD will convert
467 old format IDIF as new format IDIF with OST-index stored in the
468 LMA EA when accessing such OST-object or via OI scrub. Once such
469 flag is enabled, it cannot be reverted back, so the system cannot
470 be downgraded to the orignal incompatible version.
472 --------------------------------------------------------------------------------
474 07-30-2014 Intel Corporation
476 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
477 for currently supported client and server kernel versions.
478 * Server known to build on patched kernels:
479 2.6.32-431.20.3.el6 (RHEL6.5)
480 3.0.101-0.31 (SLES11 SP3)
481 * Client known to build on unpatched kernels:
482 2.6.32-431.20.3.el6 (RHEL6.5)
483 3.0.101-0.31 (SLES11 SP3)
486 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
487 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
488 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
489 stack overflow. For more information, please refer to bugzilla 17630.
490 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
491 the re-exporting nodes to be mounted with "32bitapi" mount option
493 --------------------------------------------------------------------------------
495 10-31-2013 Intel Corporation
497 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
498 * Server builds for kernels:
499 2.6.32-358.18.1.el6 (RHEL6.4)
500 * Client builds for unpatched kernels:
501 2.6.32-358.18.1.el6 (RHEL6.4)
502 3.0.92-0.8 (SLES11 SP3)
504 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
505 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
506 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
507 stack overflow. For more information, please refer to bugzilla 17630.
508 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
509 the re-exporting nodes to be mounted with "32bitapi" mount option
511 --------------------------------------------------------------------------------
513 05-30-2013 Intel Corporation
515 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
516 * Server builds for kernels:
517 2.6.32-358.6.2.el6 (RHEL6.4)
518 * Client builds for unpatched kernels:
519 2.6.32-358.6.2.el6 (RHEL6.4)
520 3.0.74-0.6.6 (SLES11 SP2)
522 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
523 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
524 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
525 stack overflow. For more information, please refer to bugzilla 17630.
526 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
527 the re-exporting nodes to be mounted with "32bitapi" mount option
529 --------------------------------------------------------------------------------
531 9-30-2012 Whamcloud, Inc.
533 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
534 * Server support for kernels:
535 2.6.32-279.5.1.el6 (RHEL6)
536 * Client support for unpatched kernels:
537 2.6.32-279.5.1.el6 (RHEL6)
539 2.6.18-238.19.1.el5 (RHEL5)
540 2.6.16.60-0.69.1 (SLES10)
541 * Recommended e2fsprogs version: 1.41.90.wc4
542 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
543 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
544 stack overflow. For more information, please refer to bugzilla 17630.
545 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
546 the re-exporting nodes to be mounted with "32bitapi" mount option
548 Severity : enhancement
550 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
552 Severity : enhancement
554 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
556 Severity : enhancement
558 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
560 Severity : enhancement
562 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
564 Severity : enhancement
566 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
568 --------------------------------------------------------------------------------
570 04-30-2012 Whamcloud, Inc.
572 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
573 * Server support for kernels:
574 2.6.32-220.4.2.el6 (RHEL6)
575 * Client support for unpatched kernels:
576 2.6.32-220.4.2.el6 (RHEL6)
578 2.6.18-238.19.1.el5 (RHEL5)
579 2.6.16.60-0.69.1 (SLES10)
580 * Recommended e2fsprogs version: 1.41.90.wc4
581 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
582 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
583 stack overflow. For more information, please refer to bugzilla 17630.
584 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
585 the re-exporting nodes to be mounted with "32bitapi" mount option
587 Severity : enhancement
589 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
591 Severity : enhancement
593 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
595 Severity : enhancement
597 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
599 Severity : enhancement
601 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
603 --------------------------------------------------------------------------------
605 09-30-2011 Whamcloud, Inc.
607 * Server support for kernels:
608 2.6.18-238.19.1.el5 (RHEL5)
609 2.6.32-131.6.1.el6 (RHEL6)
610 * Client support for unpatched kernels:
611 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
612 2.6.18-238.19.1.el5 (RHEL5)
613 2.6.32-131.6.1.el6 (RHEL6)
614 2.6.29.4-167.fc11 (FC11)
615 2.6.16.60-0.69.1 (SLES10)
617 * Recommended e2fsprogs version: 1.41.90.wc3
618 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
619 Lustre filesystem with 4K stack may cause a stack overflow. For more
620 information, please refer to bugzilla 17630.
621 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
622 mounted with "32bitapi" mount option
624 Severity : enhancement
625 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
627 Severity : enhancement
628 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
630 Severity : enhancement
631 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
634 Frequency : only with SLES10
636 Description: Use OFED "KMP" provided by Novell
637 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
638 Lustre is now built against this package. That means you need to
639 install the ofed-kmp package from Novell for the patchless client.
640 Note that the ofed-kmp that Novell ships may not exactly match the
641 kernel version but should still be compatible.
643 Severity : enhancement
645 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
646 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
648 --------------------------------------------------------------------------------
650 2010-07-15 Oracle, Inc.
652 * Server support for kernels:
653 2.6.18-164.11.1.el5 (RHEL 5)
654 2.6.18-164.11.1.0.1.el5 (OEL 5)
655 * Client support for unpatched kernels:
656 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
657 2.6.18-164.11.1.el5 (RHEL 5),
658 2.6.18-164.11.1.0.1.el5 (OEL 5)
659 2.6.16.60-0.42.8 (SLES 10),
661 2.6.29.4-167.fc11 (FC11)
662 * Recommended e2fsprogs version: 1.41.10-sun2
664 Severity : enhancement
666 Description: Distribute lfs_migrate script with lustre
667 Details : The lfs_migrate mentioned in the manual is now included in the
668 lustre tree in lustre/scripts and will distributed in lustre RPMs.
670 Severity : enhancement
672 Description: Failover nid list added to import proc file.
674 Severity : enhancement
676 Description: "som_preview" mount option required to use SOM
677 Details : The clients will need to use som_preview mount option
681 Description: Lustre does not do 1MB IOs to HW RAID
682 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
684 Severity : enhancement
686 Description: lfs quota output is not very convenient for awk/sed-parsing
687 Details : Some positions in lfs quota output table could be empty or
688 non-empty which made it hard to parse it with scripts, now
689 a dash is put instead of space where there is not supposed
694 Description: Update to OFED 1.5.1
695 Details : OFED 1.5.1 will be used in environments that don't use
696 vendor supplied OFED.
700 Description: Remove lnet. prefix from lctl params display
701 Details : Also renamed lnet.memused to lnet_memused to
706 Description: Use conf_param -d to remove permanant settings.
710 Description: Add list_param -R
711 Details : lctl list_param now supports recursive mode.
714 Bugzilla : 21829, 22312
715 Description: Fix llobdstat.
716 Details : 1. Now updating the write_bytes counter
717 2. Port script fixes from 1.8.
721 Description: decrease the usage of memory on clients.
722 Details : 1. On clients, recycle dentries and inodes unused.
723 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
728 Description: llobdstat fix and enhancement
729 Details : add a counter to set a limit to how many samples will be returned
730 fix a wildcard in the path to limit to obdfilter stats only
734 Description: lfs check only allowed for root
735 Details : Code cleanups and added a regresion test of non-root use of
740 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
741 Details : Fix hr->hr_index overload.
745 Description: Fix FIEMAP functionality.
749 Description: FC11 patchless client support added
753 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
754 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
758 Description: New acc-sm script and various test-framework enhancments.
762 Description: Previously landed patch was reverted.
766 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
770 Description: Add ldiskfs maxdirsize mount option.
774 Description: Fix mkfs.lustre support for 16TB LUNs.
778 Description: Print mmp_check_interval
779 Details : print mmp_check_interval and make it possible to abort
780 mount operation in case it takes too long.
784 Description: Add cascading_rw to lustre mpi tests.
788 Description: Store SOM data in EA.
789 Details : Also several SOM related cleanups.
793 Description: Add refcount to md_open_data
797 Description: Use refcount on lov_request_set to prevent memory corruption.
801 Description: Statahead: use parent dir's i_mutex for synchronization.
805 Description: loadgen improvements.
809 Description: Enable peer health detection in ksocklnd and ko2iblnd.
813 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
817 Description: Fix race on obd_devcie:md_stats
821 Description: Fix SLES 11 client builds.
822 Details : Resolved uninitalized pointer gcc warnings.
826 Description: Allow non-root access for "lfs check".
827 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
831 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
833 Severity : enhancement
835 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
836 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
840 Description: Sanity test_27p fails.
841 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
842 Sanity tests 27 were cleaned up and improved.
846 Description: Unify req format on client/servers
847 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
848 buffers and swabbers approach..
853 Description: drop join file stuff.
855 Severity : enhancement
857 Description: Add 2.6.27 and SLES11 (patchless client) support
861 Description: Fix in ptlrpc_expire_one_request() to print the signed time
864 Severity : enhancement
866 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
869 Severity : enhancement
871 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
873 Severity : enhancement
875 Description: correctly handle big reply message.
876 Details : send LNet event if reply is bigger than buffer and adjust this buffer
881 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
883 Severity : enhancement
885 Description: Add support for OEL5.
889 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
891 Severity : enhancement
893 Description: "weak-modules" support
894 Details : Implement "weak-modules" support which enables kernel modules
895 to be used with any kernel that implements the same kABI. In
896 order to achieve this modules are now installed in
897 /lib/modules/$(uname -r)/updates/kernel on all distributions.
901 Description: client could not reconnect to OST because of an active request.
902 Details : The client now retries to reconnect to the same server, if a
903 connect request failed with EBUSY or -EAGAIN. Test case 35b
904 conf-sanity.sh has been created.
908 Description: Don't allow make backward step on assignin osc next id.
909 Details : race between allocation next id and ll_sync thread can be cause
910 of set wrong osc next id and can be kill valid ost objects.
914 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
918 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
922 Description: Error handling in osc_statfs_interpret() has been improved.
923 Details : Check in osc_statfs_interpret() for EBADR.
927 Description: Do not send statfs() requests to OSTs disabled by administrator.
928 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
930 Severity : enhancement
932 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
934 Severity : enhancement
936 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
938 Severity : enhancement
940 Description: Enhancement for lfs(1) command to use numeric uid/gid.
944 Description: Improved cleanup procedure for conf-sanity test_46a.
948 Description: Do not start bulk transfer, if client tried to reconnect.
949 Details : Do not start bulk transfer, if client tried to reconnect,
950 but failed due to active rpcs. Check for exp_abort_active_req
951 has been implemented.
955 Description: Prevent inconsistences between linux and lustre mount structures.
956 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
957 Make the sleep interruptible.
959 Severity : enhancement
961 Description: provide server to client comms path
962 Details : server to client communications path via new LDLM_SET_INFO rpc
966 Description: truncate starts GFP_FS allocation under transaction causing deadlock
967 Details : ldiskfs_truncate calls grab_cache_page which may start page
968 allocation under an open transaction. This may lead to
969 calling prune_icache with consequent lustre reentrance.
973 Description: send correctly lsm on open replay
974 Details : MDS is trust to LSM size on replay open, but client can set wrong size
979 Description: enable client interface failover
980 Details : When a child reconnects from another NID, properly update export
981 nid hash position and ldlm reverse import.
983 Severity : enhancement
985 Description: Communicate OST degraded/readonly state via statfs to MDS
986 Details : Flags in the statfs returned from OSTs indicate whether the
987 OST is in a degraded RAID state, or if the filesystem has
988 turned read-only after a filesystem error is detected.
993 Description: don't panic if EPROTO was hit when reading symlink
994 Details : correctly handling request reference in error cases.
996 Severity : enhancement
998 Description: Add LustreNetLink, a kernel-userspace communcation path.
1000 Severity : enhancement
1002 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1007 Description: access to llog context before init.
1008 Details : move handling CATALOGS file at osc layer and forbid access to llog
1009 context before init.
1012 Frequency : rare, only when client is writing to OST on local node
1014 Description: Avoid deadlock for local client writes
1015 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1016 memory freeing context. This allows OST threads to set the
1017 PF_MEMALLOC flag on task structures in order to allocate memory
1018 from reserved pools and complete IO.
1019 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1020 so that the OST threads generate memory pressure and allow
1021 inactive pages to be reclaimed.
1023 Severity : enhancement
1025 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1030 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1031 Details : move ldlm namespace creation in setup phase to avoid grab
1032 _lprocfs_lock with cli_sem held.
1036 Description: Temporarily disable grant shrink.
1037 Details : Disable the feature for debugging.
1040 Frequency : only when formatting test filesystems
1042 Description: Unable to run several mkfs.lustre on loop devices at the same time
1043 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1044 formatting. The solution is to proper handle the error.
1046 Severity : enhancement
1048 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1050 Severity : enhancement
1052 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1057 Description: resolve race between obd_disconnect and class_disconnect_exports
1058 Details : if obd_disconnect will be called to already disconnected export he
1059 forget release one reference and osc module can't unloaded.
1061 Severity : enhancement
1063 Description: Allow tuning service thread via /proc
1064 Details : For each service a new
1065 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1066 created that can be used to set min/max thread counts, and get the
1067 current number of running threads.
1072 Descriptoin: don't return error if have partially created objects for file.
1073 Details : lov_update_create_set uses set->set_success as index for created
1074 objects, so if some requests will be failed, they will have hole
1075 at end of array and qos_shrink_lsm for allocate correct lsm.
1077 Severity : enhancement
1079 Description: Update OFED support to 1.4.1
1084 Description: Reduce small size read RPC
1085 Details : Set read-ahead limit for every file and only do read-ahead when
1086 available read-ahead pages are bigger than 1M to avoid small size
1089 Severity : enhancement
1091 Description: move AT tunable parameters for more consistent usage
1092 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1094 Severity : enhancement
1096 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1097 Details : allow skip disconnected ost for send statfs request and hide error
1103 Description: fix racy locking of mballoc block bitmaps causing BUG
1104 Details : The locking of the mballoc buddy bitmap and the in-memory
1105 block bitmap was using two different spin locks in some
1106 cases. This made it possible to incorrectly access the
1107 mballoc bitmap while another process was modifying it,
1108 causing a sanity assertion to fail. While no on-disk corruption
1109 was reported, there was some risk of this happening.
1112 Frequency : rare, on llog test 6
1114 Descriptoin: don't allow connect to already connected import
1115 Details : allowing connect to already connected import is hide connecting problem.
1120 Descriptoin: allow kill process which wait statahead result
1121 Details : for some reasons 'ls' can stick in waiting result from statahead,
1122 in this case need way for kill this process.
1124 Severity : enhancement
1126 Description: Add state history info file, enhance import info file
1127 Details : Track import connection state changes in a new osc/mdc proc file;
1128 add overview-type data to the osc/mdc import proc file.
1130 Severity : enhancement
1132 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1134 Severity : enhancement
1136 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1141 Descriptoin: OST grant too much space to client even there are not enough space.
1142 Details : Client will shrink its grant cache to OST if there are no write
1143 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1144 this grant cache if there are already not enough avaible space
1145 (left_space < total_clients * 32M).
1148 Frequency : start MDS on uncleanly shutdowned MDS device
1150 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1151 Details : stay in waiting mds<>ost recovery finished produce random bugs
1152 due race between two ll_sync thread for one lov target. send
1153 ACTIVATE event only if connect realy finished and import have
1157 Frequency : rare, connect and disconnect target at same time
1159 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1160 Details : don't call obd_disconnect under lov_lock. this long time
1161 operation and can block ptlrpcd which answer to connect request.
1166 Descriptoin: don't lose wakeup for imp_recovery_waitq
1167 Details : recover_import_no_retry or invalidate_import and import_close can
1168 both sleep on imp_recovery_waitq, but we was send only one wakeup
1172 Frequency : always with long access acl
1174 Descriptoin: mds can't pack reply with long acl.
1175 Details : mds don't control size of acl but they limited by reint/getattr
1178 Severity : enhancement
1180 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1182 Severity : enhancement
1184 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1187 Frequency : start MDS on uncleanly shutdowned MDS device
1189 Descriptoin: aborting recovery hang on MDS
1190 Details : don't throttle destroy RPCs for the MDT.
1193 Frequency : on remount
1195 Description: external journal device not working after the remount
1196 Details : clear dev_rdonly flag for external journal devices in
1202 Description: shutdown vs evict race
1203 Details : client_disconnect_export vs connect request race.
1204 if client will evicted at this time - we start invalidate
1205 thread without referece to import and import can be freed
1209 Frequency : rare, need acl's on inode.
1211 Description: client can't handle ost additional correctly
1212 Details : if ost was added after client connected to mds client can have
1213 hit lnet_try_match_md ... to big messages to wide striped files.
1214 in this case need teach client to handle config events about add
1215 lov target and update client max ea size at that event.
1217 Severity : enhancement
1219 Description: Changelogs
1220 Details : Changelogs are a lightweight mechanism to track filesystem
1221 metadata and namespace changes. The changelog is recorded
1222 permanently on the MDTs, and is periodically "consumed" / purged
1223 when records are no longer needed.
1225 Severity : enhancement
1227 Description: compact fld format with extents
1228 Details : Store range of seq rather than every seq in FLD. Seq
1229 controller update FLD rather than clients. In Case of CMD, mdt0
1230 has FLD, all other metadata server act as non persistent proxy
1231 for FLD queries and cache fld entries in fld cache.
1236 Description: don't skip ost target if they assigned to file
1237 Details : Drop slow OSCs if we can, but not for requested start idx.
1238 This means "if OSC is slow and it is not the requested
1239 start OST, then it can be skipped, otherwise skip it only
1240 if it is inactive/recovering/out-of-space.
1244 Description: more cleanup in mds_lov
1245 Details : not send LOV EA under replay, we can't know about they size at this
1246 time. Don't allow client connect to mds before any ost connected,
1247 for avoid problems with LOV EA size and returning EIO to client.
1249 Severity : enhancement
1251 Description: Interoperability at server side (Disk interoperability)
1253 Severity : enhancement
1255 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1257 Severity : enhancement
1259 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1261 Severity : enhancement
1263 Description: New client IO stack (CLIO).
1265 Severity : enhancement
1267 Description: Commit on sharing. Eliminate inter-client dependencies between
1268 uncommitted transactions by doing transaction commits.
1269 Thereby clients may recovery independently.
1272 Frequency : Create a symlink file with a very long name
1274 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1275 Details : If there is no extra space in the request for early cancels,
1276 ldlm_req_handles_avail() returns 0 instead of a negative value.
1278 Severity : enhancement
1280 Description: Add /proc entry for import status
1281 Details : The mdc, osc, and mgc import directories now have
1282 an import directory that contains useful import data for debugging
1283 connection problems.
1285 Severity : enhancement
1287 Description: Re-disable certain /proc logging
1288 Details : Enable and disable client's offset_stats, extents_stats and
1289 extents_stats_per_process stats logging on the fly.
1292 Frequency : Only on FC kernels 2.6.22+
1294 Description: oops in statahead
1295 Details : Do not drop reference count for the dentry from VFS when lookup,
1296 VFS will do that by itself.
1298 Severity : enhancement
1300 Description: Generic /proc file permissions
1301 Details : Set /Proc file permissions in a more generic way to enable non-
1302 root users operate on some /proc files.
1306 Description: Hitting mdc_commit_close() ASSERTION
1307 Details : Properly handle request reference release in
1308 ll_release_openhandle().
1312 Description: quota recovery deadlock during mds failover
1313 Details : This patch includes att18982, att18236, att18237 in bz14840.
1315 1. fix osts hang when mds does failover with quotaon
1316 2. prevent watchdog storm when osts threads wait for the
1321 Frequency : only patchless client
1322 Description: add workaround for race between add/remove dentry from hash
1324 Severity : enhancement
1326 Description: Allow OST glimpses to return PW locks
1330 Description: LBUG when llog conf file is full
1331 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1335 Description: Prevent import from entering FULL state when server in recovery
1339 Description: service mount cannot take device name with ":"
1340 Details : Only when device name contains ":/" will mount treat it as
1346 Description: replace ptlrpcd with the statahead thread to interpret the async
1347 statahead RPC callback
1351 Frequency : on recovery
1352 Description: I/O failures after umount during fail back
1353 Details : if client reconnected to restarted server we need join to recovery
1354 instead of find server handler is changed and process self
1355 eviction with cancel all locks.
1357 Severity : enhancement
1359 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1361 Severity : enhancement
1363 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1365 Severity : enhancement
1367 Description: Upcall on Lustre log has been dumped
1368 Details : Allow for a user mode script to be called once a Lustre log has
1369 been dumped. It passes the filename of the dumped log to the
1370 script, the location of the script can be specified via
1371 /proc/sys/lnet/debug_log_upcall.
1376 Description: avoid idr_remove called for id which is not allocated.
1377 Details : Move assigment s_dev for clustered nfs to end of initialization,
1378 to avoid problem with error handling.
1383 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1384 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1385 into CONN_USED_HASH and this prodice warning when put connection
1386 again in unused hash.
1389 Severity : enhancement
1391 Description: File striping can now be set to use an arbitrary pool of OSTs.
1393 Severity : enhancement
1395 Description: Export bytes_read/bytes_write count on OSC/OST.
1399 Description: Early reply size mismatch, MGC loses connection
1400 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1401 the connect flags are properly negotiated.
1406 Description: quotas are not honored with O_DIRECT
1407 Details : all writes with the flag O_DIRECT will use grants which leads to
1408 this problem. Now using OBD_BRW_SYNC to guard this.
1412 Description: add quota statistics
1413 Details : 1. sort out quota proc entries and proc code.
1414 2. add quota statistics
1416 Severity : enhancement
1418 Description: enable quota support for HEAD.
1422 Description: Properly propagate oinfo flags from lov to osc for statfs
1423 Details : restore missing copy oi_flags to lov requests.
1425 Severity : enhancement
1427 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1428 : l_getgroups(8), lst(8), routerstat(8)
1429 Details : included man pages for llobdstat(8), llstat(8),
1430 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1432 Severity : enhancement
1434 Description: configure's --enable-quota should check the
1435 : kernel .config for CONFIG_QUOTA
1436 Details : configure is terminated if --enable-quota is passed but
1437 : no quota support is in kernel
1441 Description: Remove portals compatibility
1442 Details : Remove portals compatibility, not interoperable with releases
1447 Description: Resolve device initialization race
1448 Details : Prevent proc handler from accessing devices added to the
1449 obd_devs array but yet be intialized.
1451 Severity : enhancement
1453 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1455 Severity : enhancement
1457 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1462 Description: Using wrong pointer in osc_brw_prep_request
1463 Details : Access to array[-1] can produce panic if kernel compiled with
1464 CONFIG_PAGE_ALLOC enabled
1468 Description: Client runs out of low memory
1469 Details : Consider only lowmem when counting initial number of llap pages
1473 Description: *optional* service tags registration
1474 Details : if the "service tags" package is installed on a Lustre node
1475 When the filesystem is mounted, a local-node service tag will
1476 be created. See http://inventory.sun.com/ for more information
1477 about the Service Tags asset management system.
1481 Description: Kernel BUG tries to release flock
1482 Details : Lustre does not destroy flock lock before last reference goes
1483 away. So always drop flock locks when client is evicted and
1484 perform unlock regardless of successfulness of speaking to MDS.
1488 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1492 Description: sanity test 65a fails if stripecount of -1 is set
1493 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1498 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1499 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1502 Severity : enhancement
1504 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1508 Description: do not process already freed flock
1509 Details : flock can possibly be freed by another thread before it reaches
1510 to ldlm_flock_completion_ast.
1514 Description: LBUG during stress test
1515 Details : Need properly lock accesses the flock deadlock detection list.
1519 Description: oops in page fault handler
1520 Details : kernel page fault handler can return two special 'pages' in error
1521 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1525 Description: timeout with invalidate import.
1526 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1527 should be handled by ptlrpcd. This produce long age waiting and
1528 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1530 Severity : enhancement
1532 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1536 Description: enable MGS and MDT services start separately
1537 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1538 a co-located MGS without starting the MGS, which is a complement
1539 to 'nosvc' mount option.
1543 Frequency : after recovery
1544 Description: precreate to many object's after del orphan.
1545 Details : del orphan st in oscc last_id == next_id and this triger growing
1546 count of precreated objects. Set flag LOW to skip increase count
1547 of precreated objects.
1551 Frequency : rare, on clear nid stats
1552 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1553 Details : when clean nid stats sometimes try destroy live entry,
1554 and this produce panic in free.
1558 Description: Stack overflow during MDS log replay
1559 ease stack pressure by using a thread dealing llog_process.
1563 Description: wait until IO finished before start new when do lock cancel.
1564 Details : VM protocol want old IO finished before start new, in this case
1565 need wait until PG_writeback is cleared until check dirty flag and
1566 call writepages in lock cancel callback.
1568 Severity : enhancement
1570 Description: using special macro for print time and cleanup in includes.
1574 Description: mds_mfd_close() ASSERTION(rc == 0)
1575 Details : In mds_mfd_close(), we need protect inode's writecount change
1576 within its orphan write semaphore to prevent possible races.
1580 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1584 Frequency : rare, on shutdown ost
1585 Description: don't hit live lock with umount ost.
1586 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1587 use shrink_dcache_sb instead.
1591 Description: don't panic with use echo client
1592 Details : echo client pass NULL as client nid pointer and this produce null
1593 pointer dereference.
1597 Description: fix build on ppc32
1598 Details : compile code with -m64 flag produce wrong object file for ppc32.
1602 Description: add message levels for liblustreapi
1606 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1607 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1608 will result in an -EAGAIN and in the exit call path we call
1609 alloc_rr() which will with increasing aggressiveness attempt to
1610 aquire precreated objects on the minimum number of required OSCs.
1614 Description: Use old size assignment to avoid deadlock
1615 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1616 the scheduling while holding a spinlock. We do not need locking
1617 for size in ll_update_inode() because size is only updated from
1618 the MDS for directories or files without objects, so there is no
1619 other place to do the update, and concurrent access to such inodes
1620 are protected by the inode lock.
1624 Description: resolve "_IOWR redefined" build error on SLES10
1628 Description: dump the memory debugging after all modules are unloaded to
1629 suppress false negative in conf_sanity test 39
1631 Severity : enhancement
1633 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1638 Description: fix SLES kernel versioning
1639 Details : the kernel version for our SLES 10 kernel did not include a "-"
1640 before the "smp" at the end. while this was not a problem in
1641 general, it did mean that software trying to use the kernel
1642 version to try to detect a vendor specific kernel would fail.
1643 this was most evident by the OFED build scripts.
1647 Description: Don't update lov_desc members until making sure they are valid
1648 Details : When updating lov_desc members via proc fs, need fix their
1649 validities before doing the real update.
1653 Description: don't put request into delay list while invalidate in flight.
1654 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1655 import in flight. this produce timeout for invalidate and sometimes
1656 can cause stale data.
1660 Frequency : on ppc only
1661 Description: not convert ost objects for directory because it's not exist.
1662 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1666 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1668 Severity : enhancement
1670 Description: Update to support 2.6.22.14 vanilla kernel.
1674 Frequency : rare, on recovery
1675 Description: read procfs can produce deadlock in some situation
1676 Details : Holding lprocfs lock which send rpc can produce block for destroy
1677 obd objects and this also block reconnect with -EALREADY. This
1678 isn't fix all lprocfs bugs - but make it rare.
1680 Severity : enhancement
1682 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1685 Frequency : frequent on X2 node
1687 Description: mdc_set_open_replay_data LBUG
1688 Details : Set replay data for requests that are eligible for replay.
1692 Description: lustre_mgs: operation 101 on unconnected MGS
1693 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1694 MGC, and late on MGC cannot successfully connect to MGS and a lot
1695 of the error messages complaining that MGS is not connected.
1699 Frequency : on network error
1700 Description: panic with double free request if network error
1701 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1702 but it's true only for synchronus enqueue, for async enqueue
1703 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1706 Severity : enhancement
1708 Description: client-side metadata stat-ahead during readdir(directory readahead)
1709 Details : perform client-side metadata stat-ahead when the client detects
1710 readdir and sequential stat of dir entries therein
1713 Frequency : on start mds
1715 Description: Implement get_info(last_id) in obdfilter.
1718 Frequency : occasional
1720 Description: Correctly check stale fid, not start epoch if ost not support SOM
1721 Details : open with flag O_CREATE need set old fid in op_fid3 because
1722 op_fid2 was overwritten with new generated fid, but mds can answer
1723 with one of these two fids and both is not stale. Setattr
1724 incorrectly started an epoch and assume will be called
1725 done_writing, but without SOM done_writing ever being called.
1728 Frequency : rare, depends on device drivers and load
1730 Description: MDS or OSS nodes crash due to stack overflow
1731 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1732 In some cases, in conjunction with device drivers that use a lot
1733 of stack the MDS (or possibly OSS) service threads could overflow
1734 the stack. One change which was identified to consume additional
1735 stack has been reworked to avoid the extra stack usage.
1738 Frequency : occasional
1740 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1741 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1742 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1743 oscc_has_objects and see OSC has no objests and call
1744 oscc_internal_create to resend create request.
1746 Severity : enhancement
1748 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1750 Severity : enhancement
1752 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1755 Frequency : very rare
1757 Description: Fix replay if there is an un-replied request and open
1758 Details : In some cases, older replay request will revert the
1759 mcd->mcd_last_xid on MDS which is used to record the client's
1760 latest sent request.
1762 Severity : enhancement
1764 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1766 Severity : enhancement
1768 Description: Add rhel5 support to HEAD.
1770 Serverity : enhancement
1772 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1777 Description: Don't allow skipping OSTs if index has been specified.
1778 Details : Don't allow skipping OSTs if index has been specified, make
1779 locking in internal create lots better.
1783 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1784 Details : access to bitfield in structure is always rounded to long
1785 and this produce problem with not atomic change any bit.
1789 Description: Lustre make rpms failed.
1790 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1791 builds Lustre rpms from tarball.
1794 Frequency : rare on shutdown ost
1796 Description: If llog cancel was not send before clean_exports phase, this can
1797 produce deadlock in llog code.
1798 Details : If llog thread has last reference to obd and call class_import_put
1799 this produce deadlock because llog_cleanup_commit_master wait when
1800 last llog_commit_thread exited, but this never success because was
1801 called from llog_commit_thread.
1805 Description: allow userland application know is lost one of stripes.
1806 Details : fill lvb_blocks with error code on ost and return it to
1807 application if error flag found.
1811 Description: NULL lov_tgts causing MDS oops
1812 Details : more safe checks for NULL lov_tgts for avoid oops.
1814 Severity : enhancement
1816 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1820 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1821 Details : If an OST is down the MDS will hang indefinitely in
1822 obd_statfs_rqset() waiting for the statfs data. While for
1823 MDS QOS usage of statfs, it should not stuck in waiting.
1825 Severity : enhancement
1827 Description: remote_acl support
1828 Details : Support ACL-based permission check for remote user.
1829 Support setfacl/getfacl for remote user with the utils
1830 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1831 the system "{s,g}etfacl" utils.
1833 Severity : enhancement
1835 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1837 Severity : enhancement
1839 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1843 Description: make mgs_setparam() handle fsname containing dash
1844 Details : fsname containing a dash does not work with lctl conf_param
1846 Severity : enhancement
1848 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1850 Severity : enhancement
1852 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1856 Frequency : rare, at shutdown
1857 Description: access already free / zero obd_namespace.
1858 Details : if client_disconnect_export was called without force flag set,
1859 and exist connect request in flight, this can produce access to
1860 NULL pointer (or already free pointer) when connect_interpret
1861 store ocd flags in obd_namespace.
1865 Frequency : only at startup
1866 Description: not alloc memory with spinlock held.
1867 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1868 if any spinlock held.
1870 Severity : enhancement
1872 Description: make lustre randomly fail allocating memory
1873 Details : Make lustre randomly failed allocating memory for testing purpose.
1875 Severity : enhancement
1877 Description: lost problems with lov objid file
1878 Details : Fixes some scability and access to not inited memory problems
1879 in work with lov objdid file.
1884 Description: lfs find does not continue on file error
1885 Details : Continue other files processing when a file/dir is absent.
1889 Description: Inconsistent usage of lustre_pack_reply()
1890 Details : Standardize the usage of lustre_pack_reply() such that it
1891 always generate a CERROR on failure.
1894 Frequency : occasional
1896 Description: MDS hang or stay in waiting lock
1897 Details : If client receive lock with CBPENDING flag ldlm need send lock
1898 cancel as separate rpc, to avoid situation when cancel request
1899 can't processed due all i/o threads stay in wait lock.
1903 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1907 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1911 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1915 Description: Update to RHEL4 latest kernel.
1917 Severity : enhancement
1919 Description: Build SLES10 patchless client fails
1920 Details : The configure was broken by run ./configure with
1921 --with-linux-obj=.... argument for patchless client. When the
1922 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1923 can't search header adequately. Use absolute path such as
1924 -I($LINUX)/include instead.
1928 Description: interrupt oig_wait produce painc on resend.
1929 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1930 private set, and this produce situation when rq_ptlrpcd_data not
1931 copyed to new allocated request and triggered LBUG on assert
1932 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1933 ptlrpcd set if request is changed and can be safety changed to use
1938 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1939 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1940 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1941 these macros and replace them with their difinitions instead.
1945 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1946 Details : Update which_patch & target file for SLES10 latest kernel.
1949 Description: Remove client patches from SLES 10 kernel.
1950 Details : This causes SLES 10 clients to behave as patchless clients
1951 even on a Lustre-patched (server) kernel.
1953 Severity : enhancement
1955 Description: self-adjustable client's lru lists
1956 Details : use adaptive algorithm for managing client cached locks lru
1957 lists according to current server load, other client's work
1958 pattern, memory activities, etc. Both, server and client
1959 side namespaces provide number of proc tunables for controlling
1962 Severity : enhancement
1964 Description: light-weight GSS support
1965 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1966 and reduce performance overhead.
1968 Severity : enhancement
1970 Description: Linux keyring support
1971 Details : Support using service of Linux keyring for Lustre GSS internal
1972 context refresh/cache mechanism.
1976 Description: Fix errors in lfs documentation
1977 Details : Fixes man pages
1981 Description: don't use GFP_* in generic Lustre code.
1982 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1986 Description: obdclass is limited by single OBD_ALLOC(idarray)
1987 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1991 Description: warnings with build patchless client with vanila 2.6.19 and up
1992 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1993 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1997 Frequency : Only for SLES
1998 Description: sanity_quota fail test_1
1999 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2000 file, which makes TSTID[2] inunique.
2004 Frequency : Only for SLES9
2005 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2006 Details : Update target/ChangeLog/which_patch .
2010 Frequency : Only for test_10 in sanity-quota.sh
2011 Description: a bug in quota test code
2012 Details : checking a test flag in a wrong place
2016 Frequency : Only for test_13 in sanity-quota.sh
2017 Description: a bug in quota test script
2018 Details : shouldn't check overfull allocated quota
2022 Description: jbd statistics
2023 Details : Port older jbd statistics patch for sles10
2025 Severity : enhancement
2027 Description: Add printing inode info into message about error in writepage.
2032 Description: reply_lock_interpret crash due to race with it and lock cancel.
2033 Details : Do not replay locks that are being cancelled. Do not reference
2034 locks by their address during replay, just by their handle.
2038 Frequency : When flocks are used.
2039 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2040 Details : Flock locks might destroy just granted lock if it could be merged
2041 with another existing flock, this is done in completion handler,
2042 so teach ldlm_cli_enquque_fini that this is a valid case for
2048 Description: Oops in read and write path when failing to allocate lock.
2049 Details : Check if lock allocation failed and return error back.
2053 Description: lstripe command fails for valid OST index
2054 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2055 instead of lov->desc.ld_active_tgt_count.
2059 Description: sanity.sh failed test 103
2060 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2065 Description: df doesn't work properly if diskfs blocksize != 4K
2066 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2070 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2071 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2073 Severity : enhancement
2075 Description: Adaptive timeouts
2076 Details : RPC timeouts adapt to changing server load and network
2077 conditions to reduce resend attempts and improve recovery time.
2081 Description: llapi_file_create() does not allow some changes
2082 Details : add llapi_file_open() that allows specifying the mode and
2083 open flags, and also returns an open file handle.
2087 Description: merge and cleanup kernel patches.
2088 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2092 Description: Add journal checksum support.(Kernel part)
2093 Details : The journal checksum feature adds two new flags i.e
2094 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2095 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2096 indicates that the commit block contains the checksum for
2097 the blocks described by the descriptor blocks. Now commit
2098 record can be sent to disk without waiting for descriptor
2099 blocks to be written to disk. This behavior is controlled
2100 using JBD2_FEATURE_ASYNC_COMMIT flag.
2104 Description: OSS needs mutliple precreate threads
2105 Details : Add ability to start more than one create thread per OSS.
2109 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2110 Details : execute lfs setstripe on client
2114 Description: mds_obd_create error creating tmp object
2115 Details : When the user sets quota on root, llog will be affected and can't
2116 create files and write files.
2119 Frequency : Always on ia64 patchless client, and possibly others.
2121 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2122 Details : This allows the patchless client to be loaded on architectures
2123 without this export.
2127 Description: RedHat Update kernel for RHEL5
2128 Details : Add the kernel config file for RHEL5.
2132 Description: RedHat Update kernel for RHEL5
2133 Details : Modify the kernel config file more closer RHEL5.
2137 Description: Build failure against Centos5 (RHEL5)
2138 Details : Define PAGE_SIZE when it isn't present.
2142 Description: test_7 on sanity-quota.sh can't be used on separate machines
2143 Details : add facet to handle it
2147 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2148 Details : with NFS, the anon dentry's parent was set to itself in
2149 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2150 corresponding dentry other than use rec->ur_name.
2152 Severity : enhancement
2154 Description: lfs setstripe enhancement
2155 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2159 Description: enable data checksumming by default
2160 Details : enable checksum by default, allow --disable-checksum
2161 configure option and "-o nochecksum" mount option. Checksums
2162 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2166 Description: lustre support for RHEL5
2167 Details : Add support for RHEL5.
2171 Description: Client eviction due to failover config
2172 Details : after a connection loss, the lustre client should attempt to
2173 reconnect to the last active server first before trying the
2174 other potential connections.
2178 Description: when mds and osts use different quota unit(32bit and 64bit),
2179 quota will be released repeatly.
2180 Details : void sending multiple quota reqs to mds, which will keep the
2181 status between the reqs.
2185 Description: osts not allocated evenly to files
2186 Details : change the condition to increase offset_idx
2189 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2191 Description: Data corruption for OSTs that are formatted larger than 2TB
2193 Details : When generating the bio request for lustre file writes the
2194 sector number would overflow a temporary variable before being
2195 used for the IO. The data reads correctly from Lustre (which
2196 will overflow in a similar manner) but other file data or
2197 filesystem metadata may be corrupted in some cases.
2201 Description: Tune the kernel for good SCSI performance.
2202 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2203 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2208 Description: rewrite ext2-derived code in obdclass/uuid.c
2209 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2210 from scratch preserving functionality.
2214 Description: "lfs find -obd UUID" prints directories
2215 Details : "lfs find -obd UUID" will return all directory names instead
2216 of just file names. It is incorrect because the directories
2217 do not reside on the OSTs.
2221 Description: use i_size_read and i_size_write in 2.6 port
2222 Details : replace inode->i_size access with i_size_read/write()
2225 Frequency : when removing large files
2227 Description: scheduling issue during removal of large Lustre files
2228 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2229 It causes scheduling issues when removing large files (17TB in the
2233 Frequency : only with liblustre clients on XT3
2235 Description: evictions taking too long
2236 Details : allow llrd to evict clients directly on OSTs
2239 Frequency : only on ppc
2241 Description: /proc/fs/lustre/devices broken on ppc
2242 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2243 We should make sure the type of 'index' is loff_t and then cast
2244 explicitly as needed below. Do not assign an explicitly cast
2248 Frequency : only for rhel5
2250 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2251 Details : Modify the target file & which_kernel.
2253 Severity : enhancement
2255 Description: omit set fsid for export NFS
2256 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2257 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2261 Frequency : Always, for kernels after 2.6.16
2262 Description: Fix warning idr_remove called for id=.. which is not allocated.
2263 Details : Last kernels save old s_dev before kill super and not allow
2264 to restore from callback - restore it before call kill_anon_super.
2268 Description: buffer overruns could theoretically occur
2269 Details : llapi_semantic_traverse() modifies the "path" argument by
2270 appending values to the end of the origin string, and a
2271 overrun may occur. Adding buffer overrun check in liblustreapi.
2275 Description: Fix error on 'ls .' at the top of the Lustre mount.
2276 Details : Don't revalidate dentry if it is a root dentry.
2280 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2281 Details : Modify vm-tunables-rhel4.patch.
2285 Description: Kernel config for 2.6.18-vanilla.
2286 Details : Modify targets/2.6-vanilla.target.in.
2287 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2288 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2289 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2290 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2294 Description: improve handling recoverable errors
2295 Details : if request processig with error which can be recoverable on server
2296 request should be resend, otherwise page released from cache and
2301 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2302 Details : Modify target file & which_patch.
2303 A flaw was found in the IA32 system call emulation provided
2304 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2305 value could be stored in the %RAX register, which could trigger an
2306 out-of-bounds system call table access. An untrusted local user
2307 could exploit this flaw to run code in the kernel
2308 (ie a root privilege escalation). (CVE-2007-4573).
2312 Description: O_DIRECT bypasses client statistics.
2313 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2314 read_ahead_stats were updated. copied the stats section from
2315 osc_send_oap_rpc() into async_internal().
2319 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2323 Description: change order of libsysio includes
2324 Details : '#include sysio.h' should always come before '#include xtio.h'
2328 Description: handle "serious error: objid * already exists" more gracefully
2329 Details : If LAST_ID value on disk is smaller than the objects existing in
2330 the O/0/d* directories, it indicates disk corruption and causes an
2331 LBUG(). If the object is 0-length, then we should use the existing
2332 object. This will help to avoid a full fsck in most cases.
2334 Severity : enhancement
2336 Description: adapt the lustre_config script to support the upgrade case
2337 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2338 server targets to 1.6 in parallel.
2342 Description: To avoid grant space > avaible space when the disk is almost
2343 full. Without this patch you might see the error "grant XXXX >
2344 available" or some LBUG about grant, when the disk is almost
2346 Details : In filter_check_grant, for non_grant cache write, we should
2347 check the left space by if (*left > ungranted + bytes), instead
2348 of (*left > ungranted), because only we are sure the left space
2349 is enough for another "bytes", then the ungrant space should be
2350 increase. In client, we should update cl_avail_grant only there
2351 is OBD_MD_FLGRANT in the reply.
2355 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2356 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2357 described in CVE-2007-4573. This problem could allow untrusted
2358 local users to gain root access.
2361 Frequency : when using O_DIRECT and quotas
2363 Description: Incorrect file ownership on O_DIRECT output files
2364 Details : block usage reported by 'lfs quota' does not take into account
2365 files that have been written with O_DIRECT.
2370 Description: touch file failed when fs is not full
2371 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2372 otherwise we can get ENOSP while fs is not full.
2376 Description: parallel lock callbacks
2377 Details : Instead of sending blocking and completion callbacks as separated
2378 requests, adding them to a set and sending in parallel.
2381 Frequency : only for Cray XT3
2382 Bugzilla : 12829/13455
2383 Description: Changing primary group doesn't change the group lustre assigns to
2385 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2386 overridden with the primary group provided by the group upcall,
2387 whereas the supplied fsgid can be trusted if it is in the list of
2388 supplementary groups returned by the group upcall.
2390 Severity : enhancement
2392 Description: Allow masking D_WARNING, D_ERROR messages from console
2393 Details : Console messages can now be disabled via lnet.printk.
2397 Description: User code with malformed file open parameter crashes client node
2398 Details : Before packing join_file req, all the related reference should be
2399 checked carefully in case some malformed flags cause fake
2400 join_file req on client.
2404 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2405 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2406 mode is equal to the granted mode. However, it should also check
2407 that LDLM_FL_CP_REQD is not set.
2411 Description: Heavy nfs access might result in deadlocks
2412 Details : After ELC code landed, it is now improper to enqueue any mds
2413 locks under och_sem, because enqueue might want to decide to
2414 cancel open locks for same inode we are holding och_sem for.
2418 Description: Client eviction while running blogbench
2419 Details : A lot of unlink operations with concurrent I/O can lead to a
2420 deadlock causing evictions. To address the problem, the number of
2421 oustanding OST_DESTROY requests is now throttled to
2422 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2423 callbacks are processed in priority.
2427 Description: enable ACLs on MDS by default
2428 Details : ACLs must be enabled on MDS by default.
2431 Frequency : PPC/PPC64 only
2433 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2434 Details : fix duplicated definitions between asm-ppc64/types.h and
2435 lustre_types.h on PPC.
2438 Frequency : PPC/PPC64 only
2440 Description: asm-ppc/segment.h does not exist
2441 Details : fix compile issue on PPC.
2445 Description: better handle error messages in extents code
2448 Frequency : RHEL4 only
2450 Description: mkfs is very slow on IA64/RHEL4
2451 Details : A performance regression has been discovered in the MPT Fusion
2452 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2453 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2454 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2456 Severity : enhancement
2458 Description: SNMP support enhancement
2459 Details : Adding total number of sampled request for an MDS node in snmp
2462 Severity : enhancement
2464 Description: Optimize ldlm waiting list processing for PR extent locks
2465 Details : When processing waiting list for read extent lock and meeting read
2466 lock that is same or wider to it that is not contended, skip
2467 processing rest of the list and immediatelly return current
2468 status of conflictness, since we are guaranteed there are no
2469 conflicting locks in the rest of the list.
2473 Description: Time out and refuse to reconnect
2474 Details : When the failover node is the primary node, it is possible
2475 to have two identical connections in imp_conn_list. We must
2476 compare not conn's pointers but NIDs, otherwise we can defeat
2477 connection throttling.
2481 Description: port llog fixes from b1_6 into HEAD
2482 Details : Port llog reference couting and some llog cleanups from b1_6
2483 (bug 10800) into HEAD, for protect from panic and access to already
2484 free llog structures.
2488 Description: Detect stride IO mode in read-ahead
2489 Details : When a client does stride read, read-ahead should detect that and
2490 read-ahead pages according to the detected stride pattern.
2494 Description: data checksumming impacts single node performance
2495 Details : add support for several checksum algorithm. Currently, only CRC32
2496 and Adler-32 are supported. The checksum type can be changed on
2497 the fly via /proc/fs/lustre/osc/*/checksum_type.
2501 Description: use adler32 for page checksums
2502 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2507 Description: build for x2 fails
2508 Details : fix compile issue on Cray systems.
2512 Description: Properly match for duplicate locks
2513 Details : Due to different lock order from skiplists code, we need to
2514 traverse entire list for now
2517 Frequency : only on PPC/SLES10
2519 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2520 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2521 be defined. Add a hack in mkfs_lustre.c to work around this compile
2526 Description: LASSERT on MDS when client holding flock lock dies
2527 Details : ldlm pool logic depends on number of granted locks equal to
2528 number of released locks which is not true for flock locks, so
2529 just exclude such locks from consideration.
2533 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2534 Details : Use fsfilt_sync() for both the whole filesystem sync and
2535 individual file sync to eliminate dangerous inode locking
2536 with I_LOCK that can lead to a deadlock.
2540 Description: performance in 1.6.3
2541 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2542 sure that 1MB requests are not fragmented by the block layer.
2544 Severity : enhancement
2546 Description: organize the server-side client stats on per-nid basis
2547 Details : Change the structure of stats under obdfilter and mds to
2555 The "uuid"s file would list the uuids of _active_ exports.
2556 And the clear entry is to clear all stats and stale nids.
2558 Severity : enhancement
2560 Description: eliminate client locks in face of contention
2561 Details : file contention detection and lockless i/o implementation
2562 for contended files.
2566 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2569 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2571 Description: MDS may lose file striping (and hence file data) in some cases
2572 Details : If there are additional extended attributes stored on the MDS,
2573 in particular ACLs, SELinux, or user attributes (if user_xattr
2574 is specified for the client mount options) then there is a risk
2575 of attribute loss. Additionally, the Lustre file striping
2576 needs to be larger than default (e.g. striped over all OSTs),
2577 and an additional attribute must be stored initially in the
2578 inode and then increase in size enough to be moved to the
2579 external attribute block (e.g. ACL growing in size) for file
2584 Description: skiplist implementation simplification
2585 Details : skiplists are used to group compatible locks on granted list
2586 that was implemented as tracking first and last lock of each
2587 lock group the patch changes that to using doubly linked lists
2591 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2592 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2593 because LASSERT on both the data supplied by a client, and the
2594 data on disk is dangerous and incorrect.
2596 Severity : enhancement
2598 Description: Slow truncate/writes to huge files at high offsets.
2599 Details : Directly associate cached pages to lock that protect those pages,
2600 this allows us to quickly find what pages to write and remove
2601 once lock callback is received.
2605 Description: more ldlm soft lockups
2606 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2607 starve other threads from the resource lock for a long time in
2608 case of long waiting queue, so change the debug level from
2609 D_OTHER to the less frequently used D_INFO.
2611 Severity : enhancement
2613 Description: add -gid, -group, -uid, -user options to lfs find
2617 Description: Hung threads in invalidate_inode_pages2_range
2618 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2619 once one is completed. As a result, some RPCs are stuck in the
2620 queue and are never sent.
2624 Description: filter threads hungs on waiting journal commit
2625 Details : Cleanup filter group llog code, then only filter group llog will
2626 be only created in the MDS/OST syncing process.
2630 Description: Procfs and llog threads access destoryed import sometimes.
2631 Details : Sync the import destoryed process with procfs and llog threads by
2632 the import refcount and semaphore.
2634 Severity : enhancement
2636 Description: openlock cache of b1_6 port to HEAD
2641 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2642 Details : If insertion of an extent fails, then discard the inode
2643 preallocation and free data blocks else it can lead to duplicate
2648 Description: don't always update ctime in ext3_xattr_set_handle()
2649 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2650 In some cases the ctime should not be updated, for example for
2651 2.0->1.8 compatibility it is necessary to delete an xattr and it
2652 should not update the ctime.
2656 Bugzilla : 15713/16362
2657 Description: Assertion in iopen_connect_dentry in 1.6.3
2658 Details : looking up an inode via iopen with the wrong generation number can
2659 populate the dcache with a disconneced dentry while the inode
2660 number is in the process of being reallocated. This causes an
2661 assertion failure in iopen since the inode's dentry list contains
2662 both a connected and disconnected dentry.
2666 Description: assertion failure in ldlm_handle2lock()
2667 Details : fix a race between class_handle_unhash() and class_handle2object()
2668 introduced in lustre 1.6.5 by bug 13622.
2673 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2674 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2679 Description: oops in sd_iostats_seq_show()
2680 Details : unloading/reloading the scsi low level driver triggers a kernel
2681 bug when trying to access the sd iostat file.
2686 Description: Kernel panics during QLogic driver reload
2687 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2688 patch, causing memory corruption.
2693 Description: journal_dev option does not work in b1_6
2694 Details : pass mount option during pre-mount.
2696 Severity : enhancement
2698 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2699 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2700 extent information of a file. It can be used to map logical blocks
2701 in a file to physical blocks in the block device.
2705 Description: LDLM soft lockups - improvement
2706 Details : It is be possible to send the lock handle along with each read
2707 or write request because the client is already doing a lock match
2708 itself so there isn't any reason the OST should have to re-do that
2712 Frequency : only on Cray X2
2714 Description: X2 build failures
2715 Details : fix build failures on Cray X2.
2719 Description: xid & resent requests
2720 Details : Initialize RPC XID from clock at startup (randomly if clock is
2723 Severity : enhancement
2725 Description: Add lustre_start utility to start or stop multiple Lustre servers
2730 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2731 Details : In case of memory pressure, list_del() can be called twice on
2732 req->rq_history_list, causing a kernel oops.
2736 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2737 Details : fix stack overflow in the distributed lock manager by defering
2738 export eviction after a failed AST to the elt thread instead of
2739 handling it in the dlm interpret routine.
2743 Description: Convert some comments to new format.
2744 Details : Update documenting comments to match doxygen conventions.
2748 Description: Grammar fixes.
2749 Details : A couple of trivial spelling fixes.
2753 Description: OSD_COUNTERS-mandatory
2754 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2755 to simplify the code.
2759 Description: simplify cmm_device freeing logic.
2760 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2764 Description: Add lockdep support to dt_object_operations locking interface.
2765 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2766 parameter indicating lock ordering. Update mdd code to use new
2771 Description: Introduce failloc constants for lockless IO tests.
2772 Details : Add two new failloc constants to test lockless IO. Only one of
2773 them in implemented---another is checked in yet to be landed
2778 Description: Add lockdep support for inode mutex.
2779 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2780 in the situations where Lustre has to lock more than one inode
2785 Description: Add optional invariants checking support.
2786 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2787 configure switch. This macro is to be used for consistency and
2788 sanity checks that are too expensive to be left in `production'
2793 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2794 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2799 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2800 Details : Use spin_lock_nested() in (the only) situation where more than
2801 one ldlm_lock is locked simultaneously. Also, fix possible
2802 dead-lock in ldlm_lock_change_resource() by enforcing particular
2807 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2808 Details : Instead of specifying each ldlm_lock call-back through separate
2809 parameter, wrap them into struct ldlm_callback_suite.
2813 Description: Kill join_lru obd method and its callers.
2814 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2815 in memory---a requirement implemented through ->o_join_lru() obd
2816 method in HEAD. Remove this method and its users.
2820 Description: Add asynchronous ldlm ENQUEUE completion handler.
2821 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2822 case that stock ldlm_completion_ast() cannot handle as it waits
2823 until lock is granted. Introduce new ldlm_completion_ast_async()
2824 for this. Also comment ldlm_completion_ast().
2828 Description: ldlm_error <-> errno conversion.
2829 Details : Add functions to map (rather arbitrary) between LDLM error codes
2830 and standard errno values. CLIO needs this to prevent LDLM specific
2831 constants from escaping ldlm and osc.
2835 Description: Kill unused ldlm_handle2lock_ns() function.
2836 Details : Kill unused ldlm_handle2lock_ns() function.
2840 Description: Add lu_ref support to ldlm_lock
2841 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2842 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2843 and ldlm_resource. LDLM interface has to be changed, because code that
2844 releases a reference on a lock, has to "know" what reference this is.
2845 In the most frequent case
2847 lock = ldlm_handle2lock(handle);
2849 LDLM_LOCK_PUT(lock);
2851 no changes are required. When any other reference (received _not_ from
2852 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2853 instead of LDLM_LOCK_PUT().
2855 Arguably, changes are pervasive, and interface requires some discipline
2856 for proper use. On the other hand, it was very instrumental in finding
2857 a few leaked lock references.
2861 Description: Add ldlm_lock_addref_try().
2862 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2863 attempts to addref a lock that might be being canceled
2868 Description: Add ldlm_weigh_callback().
2869 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2870 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2871 hard-coded `number of pages' logic.
2875 Description: Add lockdep annotations to llog code.
2876 Details : Use appropriately tagged _nested() locking calls in the places
2877 where llog takes more than one ->lgh_lock lock.
2881 Description: Add loi_kms_set().
2882 Details : Wrap kms updates into a helper function.
2886 Description: Constify instances of struct lsm_operations.
2887 Details : Constify instances of struct lsm_operations.
2891 Description: lu_conf support.
2892 Details : On a server, a file system object is uniquely identified
2893 by a fid, which is sufficient to locate and load all object
2894 state (inode). On a client, on the other hand, more data are
2895 necessary instantiate an object. Change lu_object_find() and
2896 friends to take additional `lu_conf' argument describing object.
2897 Typically this includes layout information.
2901 Description: lu_context fixes.
2902 Details : Introduce new lu_context functions that are needed on the client
2903 side, where some system threads (ptlrpcd) are shared by multiple
2904 modules, and so cannot be stopped during module shutdown.
2908 Description: Add start and stop methods to lu_device_type_operations.
2909 Details : Introduce two new methods in lu_device_type_operations, that are
2910 invoked when first instance of a given type is created and last one
2911 is destroyed respectively. This is need by CLIO.
2915 Description: Add lu_ref support to struct lu_device.
2916 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2917 track leaked references.
2921 Description: Introduce lu_kmem_descr.
2922 Details : lu_kmem_descr and its companion interface allow to create
2923 and destroy a number of kmem caches at once.
2927 Description: Fix lu_object finalization race.
2928 Details : Fix a race between lu_object_find() finding an object and its
2929 concurrent finalization. This race is (most likely) not possible
2930 on the server, but might happen on the client.
2934 Description: Introduce lu_ref interface.
2935 Details : lu_ref is a debugging module allowing to track references to
2936 a given object. It is quite cpu expensive, and has to be
2937 explicitly enabled with --enable-lu_ref. See usage description
2942 Description: Factor lu_site procfs stats into a separate function.
2943 Details : Separate lu_site stats printing code into a separate function
2944 that can be reused on a client.
2948 Description: Constify instances of struct {lu,dt,md}_device_operations.
2949 Details : Constify instances of struct {lu,dt,md}_device_operations.
2953 Description: Introduce struct md_site and move meta-data specific parts of
2954 struct lu_site here.
2955 Details : Move md-specific fields out of struct lu_site into special struct
2956 md_site, so that lu_site can be used on a client.
2960 Description: Kill mdd_lov_destroy().
2961 Details : Remove unused mdd code.
2965 Description: Add st_block checking to multistat.c.
2966 Details : Add st_block checking to multistat.c.
2970 Description: Add lu_ref support to struct obd_device.
2971 Details : Add lu_ref tracking to obd_device.
2975 Description: Kill obd_set_fail_loc().
2976 Details : Remove unused code.
2980 Description: Add special type for ptlrpc_request interpret functions.
2981 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2982 there. Actual usage will be in CLIO.
2986 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2987 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2988 doesn't work with lockdep.
2992 Description: Add rwv.c test program.
2993 Details : New testing program exercising readv(2) and writev(2) (Qian).
2997 Description: Add sendfile.c test program.
2998 Details : New testing program exercising sendfile(2) (Jay).
3002 Description: Ratelimit a message that can be very frequent.
3003 Details : Ratelimit a memory allocation failure message that can
3008 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3009 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3013 Description: Liblustre build fixes.
3014 Details : Liblustre build fixes.
3018 Description: libcfs: add cfs_{need,cond}_resched() interface.
3019 Details : libcfs: add cfs_{need,cond}_resched() definition and
3020 implementations for Linux, NT, and liblustre.
3022 Severity : enhancement
3024 Description: More exported tunables for mballoc
3025 Details : Add support for tunable preallocation window and new tunables for
3026 large/small requests
3030 Description: Detect corruption of block bitmap and checking for preallocations
3031 Details : Checks validity of on-disk block bitmap. Also it does better
3032 checking of number of applied preallocations. When corruption is
3033 found, it turns filesystem readonly to prevent further corruptions.
3037 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3038 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3039 window to 0. If it does hit cache miss, and read-pattern is still
3040 stride-io mode, does not reset the stride window, but also does not
3041 increase the stride window length in this case.
3045 Frequency : only for big-endian servers
3046 Description: Check if system is big-endian while mounting fs with extents feature
3047 Details : Mounting a filesystem with extents feature will fail on big-endian
3048 systems since ext3-based ldiskfs is not supported on big-endian
3049 systems. This can be over-riden with "bigendian_extents" mount option.
3051 Severity : enhancement
3053 Description: The root squash functionality
3054 Details : A security feature, which is to prevent users from being able
3055 to mount lustre on their desktop, run as root, and delete
3056 all of the files in the filesystem. The goal is accomplished by
3057 remapping user id (UID) and group id (GID) of the root user to
3058 a UID and GID specified by the system administartor via Lustre
3059 configuration management server (MGS). The functionality also
3060 allows to specify sets of clients for which the remapping does
3065 Description: Excessive recovery window
3066 Details : With AT enabled, the recovery window can be excessively long (6000+
3067 seconds). To address this problem, we no longer use
3068 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3069 timeout no longer depends on the service time, it is set to
3070 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3071 time via pb_service_time.
3075 Description: Watchdog triggered on MDS failover
3076 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3077 the OSTs know that the MDS "UUID" can be reused for the same export
3078 from a different NID, so we do not need to wait for the export to be
3082 Frequency : rare, only if using MMP with Linux RAID
3084 Description: MMP doesn't work with Linux RAID
3085 Details : While using HA for Lustre servers with Linux RAID, it is possible
3086 that MMP will not detect multiple mounts. To make this work we
3087 need to unplug the device queue in RAID when the MMP block is being
3088 written. Also while reading the MMP block, we should read it from
3089 disk and not the cached one.
3091 Severity : enhancement
3093 Description: open file using fid
3094 Details : A file can be opened using just its fid, like
3095 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3098 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3101 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3102 Details : add proper locking for creation of jbd_4k slab cache
3106 Description: MMP check in ext3_remount() fails without displaying any error
3107 Details : When multiple mount protection fails during remount, proper error
3110 Severity : enhancement
3112 Description: Allow stripe size to be up to 4G-64k
3113 Details : Fix math logic to allow large stripe sizes.
3117 Description: add check for >8TB ldiskfs filesystems
3118 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3119 Don't allow >8TB ldiskfs filesystems to be mounted without
3120 force_over_8tb mount option
3125 Descriptoin: OSS complains "too many blocks" and crashes
3126 Details : In filter_commitrw_write():
3127 1. lquota_chkquota();
3128 2. obdo_from_inode();
3129 3. lquota_pending_commit();
3130 "2." has changed uid and gid in this case, then caused
3131 lquota_chkquota() and lquota_pending_commit() to use
3132 different uid and gid. That is the root of the bug.
3136 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3137 Details : Detect and print out the 'stripe count = -1' case properly.
3141 Description: Out or order replies might be lost on replay
3142 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3143 request with tid smaller than one currently being added, add it
3144 to the start, not end of the list.
3146 --------------------------------------------------------------------------------
3148 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3150 * Support for kernels:
3151 2.6.5-7.283 (SLES 9),
3152 2.6.9-55.EL (RHEL 4),
3153 2.6.16.46-0.14 (SLES 10),
3154 2.6.18.8 vanilla (kernel.org)
3155 * Client support for unpatched kernels:
3156 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3157 2.6.16 - 2.6.22 vanilla (kernel.org)
3158 * Due to recently discovered recovery problems, we do not recommend
3159 using patchless RHEL 4 clients with this or any earlier release.
3160 * Recommended e2fsprogs version: 1.39.cfs8
3161 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3162 * Starting with this release, the ldiskfs backing filesystem required
3163 by Lustre is now in its own package, lustre-ldiskfs. This package
3164 should be installed. It is versioned separately from Lustre and
3165 may be released separately in future.
3169 Description: block reactivating mgc import until all deactivates complete
3170 Details : Fix race when failing back MDT/MGS to itself (testing)
3173 Frequency : at statup only
3175 Description: mds_lov_synchronize race leads to various problems
3176 Details : simultaneous MDT->OST connections at startup can cause the
3177 sync to abort, leaving the OSC in a bad state.
3179 Severity : enhancement
3181 Description: add optional extra BUILD_VERSION info
3182 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3183 to override the lustre version.
3186 Frequency : 2.6.18 servers only
3188 Description: ll_kern_mount() doesn't release the module reference
3189 Details : The ldiskfs module reference count never drops down to 0
3190 because ll_kern_mount() doesn't release the module reference.
3195 Description: server LBUG when using old ost_num_threads parameter
3196 Details : Accept the old ost_num_threads parameter but warn that it
3197 is deprecated, and fix an off-by-one error that caused an LBUG.
3202 Description: Transient SCSI error results in persistent IO issue
3203 Details : iobuf->dr_error is not reinitialized to 0 between two
3207 Frequency : sometimes when underlying device returns I/O errors
3209 Description: OSTs not going read-only during write failures
3210 Details : OSTs are not remounted read-only when the journal commit threads
3211 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3212 instead of the ext3 wrappers.
3216 Description: poor connect scaling with increasing client count
3217 Details : Don't run filter_grant_sanity_check for more than 100 exports
3218 to improve scaling for large numbers of clients.
3221 Frequency : SLES10 only
3223 Description: sanity-quota.sh quotacheck failed: rc = -22
3224 Details : Quotas cannot be enabled on SLES10.
3227 Frequency : liblustre clients only
3229 Description: getdirentries does not give error when run on compute nodes
3230 Details : getdirentries does not fail when the size specified as an argument
3231 is too small to contain at least one entry
3233 Severity : enhancement
3235 Description: Add LNET router traceability for debug purposes
3236 Details : If a checksum failure occurs with a router as part of the
3237 IO path, the NID of the last router that forwarded the bulk data
3238 is printed so it can be identified.
3243 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3244 Details : Due to a race condition, liblustre clients were occasionally
3245 evicted incorrectly.
3247 Severity : enhancement
3249 Description: lfs setstripe use optional parameters instead of postional
3252 Severity : enhancement
3254 Description: Nanosecond timestamp support for ldiskfs
3255 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3256 resolution timestamps. There is not yet support for this at
3257 the Lustre filesystem level.
3260 Frequency : during server recovery
3262 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3263 Details : request with rq_no_resend flag not awake l_wait_event if they get
3267 Frequency : nfs export on patchless client
3269 Description: connectathon hang when test nfs export over patchless client
3270 Details : Disconnected dentry cannot be found with lookup, so we do not need
3271 to unhash it or make it invalid
3274 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3275 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3276 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3281 Description: open req refcounting wrong on reconnect
3282 Details : If reconnect happened between getting open reply from server and
3283 call to mdc_set_replay_data in ll_file_open, we will schedule
3284 replay for unreferenced request that we are about to free.
3285 Subsequent close will crash in variety of ways.
3286 Check that request is still eligible for replay in
3287 mdc_set_replay_data().
3292 Description: disable writes to filesystem when reading health_check file
3293 Details : the default for reading the health_check proc file has changed
3294 to NOT do a journal transaction and write to disk, because this
3295 can cause reads of the /proc file to hang and block HA state
3296 checking on a healthy but otherwise heavily loaded system. It
3297 is possible to return to the previous behaviour during configure
3298 with --enable-health-write.
3300 Severity : enhancement
3302 Description: 64-bit inode version
3303 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3304 to the inode. This will be required for version-based recovery.
3309 Description: MDS fails to start if a duplicate client export is detected
3310 Details : in some rare cases it was possible for a client to connect to
3311 an MDS multiple times. Upon recovery the MDS would detect this
3312 and fail during startup. Handle this more gracefully.
3314 Severity : enhancement
3316 Description: Add -o localflock option to simulate old noflock
3318 Details : This will achieve local-only flock/fcntl locks
3324 Description: log_commit_thread vs filter_destroy race leads to crash
3325 Details : Take import reference before releasing llog record semaphore
3330 Description: Wrong request locking in request set processing
3331 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3332 imp_delayed_list, in this place should be used imp_lock.
3335 Frequency : when reconnection
3337 Description: Grant Leak when osc reconnect to OST
3338 Details : When osc reconnect ost, OST(filter) should check whether it
3339 should grant more space to client by comparing fed_grant and
3340 cl_avail_grant, and return the granted space to client instead
3341 of "new granted" space, because client will call osc_init_grant
3342 to update the client grant space info.
3345 Frequency : when client reconnect to OST
3347 Description: Grant Leak when osc do resend and replay bulk write
3348 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3349 bulk write request, because the grant info while be sync between
3350 OSC and OST when reconnect, and we should ignore the grant info
3351 these of resend/replay write req.
3356 Description: Grant space more than available left space sometimes.
3357 Details : When then OST is about to be full, if two bulk writing from
3358 different clients came to OST. Accord the avaliable space of the
3359 OST, the first req should be permitted, and the second one
3360 should be denied by ENOSPC. But if the seconde arrived before
3361 the first one is committed. The OST might wrongly permit second
3362 writing, which will cause grant space > available space.
3365 Frequency : when client is evicted
3367 Description: Grant might be wrongly erased when osc is evicted by OST
3368 Details : when the import is evicted by server, it will fork another
3369 thread ptlrpc_invalidate_import_thread to invalidate the
3370 import, where the grant will be set to 0. While the original
3371 thread will update the grant it got when connecting. So if
3372 the former happened latter, the grant will be wrongly errased
3373 because of this race.
3378 Description: Checking Stale with correct fid
3379 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3380 to check whether it is stale, because sometimes, we want the
3381 enqueue happened anyway, and op_data.fid2 will not be initialized.
3383 Severity : enhancement
3385 Description: update patchless client
3386 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3389 Frequency : only with 2.4 kernel
3391 Description: random memory corruption
3392 Details : size of struct ll_inode_info is to big for union inode.u and this
3393 can be cause of random memory corruption.
3398 Description: Memory leak in recovery
3399 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3400 It should also check obd_fail before fsfilt_start, otherwise if
3401 fsfilt_start return -EROFS,(failover mds during mds recovery).
3402 then the req will return with repmsg->transno = 0 and rc = EROFS.
3403 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3404 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3405 be freed no matter whether fsfilt_commit success or not.
3408 Frequency : only with huge count clients
3410 Description: Prevents from taking the superblock lock in llap_from_page for
3412 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3413 for llap_from_page call in ll_removepage prevents from taking the
3414 superblock lock for a soon died page.
3419 Description: Not check open intent error before release open handle
3420 Details : in some rare cases, the open intent error is not checked before
3421 release open handle, which may cause
3422 ASSERTION(open_req->rq_transno != 0), because it tries to release
3423 the failed open handle.
3428 Description: Set cat log bitmap only after create log success.
3429 Details : in some rare cases, the cat log bitmap is set too early. and it
3430 should be set only after create log success.
3434 Description: Accessing a block bevice can re-enable I/O when Lustre is
3435 tearing down a device.
3436 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3441 Description: service threads may hog cpus when there are a lot of requests
3442 Details : Insert cond_resched to give other threads a chance to use some CPU
3447 Description: the cat log was not initialized in recovery
3448 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3449 unlink log on mds will not be initialized until mds post
3450 recovery. And also in mds post recovery, the unlink log will
3451 initialization will be done asynchronausly, so there will be race
3452 between add unlink log and unlink log initialization.
3456 Description: brw_stats were being printed incorrectly
3457 Details : brw_stats were being printed as log2 but all of them were not
3458 recorded as log2. Also remove some code duplication arising from
3459 filter_tally_{read,write}.
3463 Frequency : rare, only in recovery.
3464 Description: ASSERTION(req->rq_type != LI_POISON) failed
3465 Details : imp_lock should be held while iterating over imp_sending_list for
3466 prevent destroy request after get timeout in ptlrpc_queue_wait.
3470 Description: replay-single.sh test 52 fails
3471 Details : A lock's skiplist need to be cleanup when it being unlinked
3472 from its resource list.
3476 Description: Short directio read returns full requested size rather than
3478 Details : Direct I/O operations should return actual amount of bytes
3479 transferred rather than requested size.
3481 Severity : enhancement
3483 Description: metadata RPC reduction (e.g. for rm performance)
3484 Details : decrease the amount of synchronous RPC between clients and servers
3485 by canceling conflicing lock before the operation on the client
3486 and packing thier handles into the main operation RPC to server.
3488 Severity : enhancement
3490 Description: Async OSC create to avoid the blocking unnecessarily.
3491 Details : If an OST has no remaining object, system will block on the
3492 creation when it needs to create a new object on this OST. Now,
3493 ways use pre-created objects when available, instead of blocking on
3494 an empty osc while others are not empty. If we must block, we
3495 block for the shortest possible period of time.
3499 Description: improve handling recoverable errors
3500 Details : if request processig with error which can be recoverable on server
3501 request should be resend, otherwise page released from cache and
3504 Severity : enhancement
3506 Description: refine locking for avoid write wrong info into lov_objid file
3507 Details : fix possible races with add new target and write/update data in
3510 --------------------------------------------------------------------------------
3512 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3517 Frequency : on some architectures
3519 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3520 Details : Uninitialized flags sometimes cause configuration commands to
3524 Frequency : patchless clients only
3526 Description: missing __iget() symbol export
3527 Details : The __iget() symbol export is missing. To avoid the need for
3528 this on patchless clients the deathrow inode reaper is turned
3529 off, and we depend on the VM to clean up old inodes. This
3530 dependency was during via the fix for bug 12181.
3532 --------------------------------------------------------------------------------
3534 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3536 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3537 INTEROPERATE with older versions automatically. In many cases a
3538 special upgrade step is needed. Please read the
3539 user documentation before upgrading any part of a live system.
3540 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3541 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3542 * WARNING: Lustre configuration and startup changes are required with
3543 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3547 Severity : enhancement
3549 Description: MountConf
3550 Details : Lustre configuration is now managed via mkfs and mount
3551 commands instead of lmc and lconf. New obd types (MGS, MGC)
3552 are added for dynamic configuration management. See
3553 https://mail.clusterfs.com/wikis/lustre/MountConf for
3556 Severity : enhancement
3558 Description: dynamic OST addition
3559 Details : OSTs can now be added to a live filesystem
3561 Severity : enhancement
3563 Description: startup order invariance
3564 Details : MDTs and OSTs can be started in any order. Clients only
3565 require the MDT to complete startup.
3567 Severity : enhancement
3569 Description: parallel, asynchronous orphan cleanup
3570 Details : orphan cleanup is now performed in separate threads for each
3571 OST, allowing parallel non-blocking operation.
3573 Severity : enhancement
3575 Description: optimized stripe assignment
3576 Details : stripe assignments are now made based on ost space available,
3577 ost previous usage, and OSS previous usage, in order to try
3578 to optimize storage space and networking resources.
3580 Severity : enhancement
3582 Description: Permanently set tunables
3583 Details : All writable /proc/fs/lustre tunables can now be permanently
3584 set on a per-server basis, at mkfs time or on a live system.
3586 Severity : enhancement
3588 Description: Lustre message v2
3589 Details : Add lustre message format v2.
3591 Severity : enhancement
3593 Description: client OST exclusion list
3594 Details : Clients can be started with a list of OSTs that should be
3595 declared "inactive" for known non-responsive OSTs.
3599 Description: SPEC SFS validation failure on NFS v2 over lustre.
3600 Details : Changes the blocksize for regular files to be 2x RPC size,
3601 and not depend on stripe size.
3603 Severity : enhancement
3605 Description: Multiple MD RPCs in flight.
3606 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3607 To avoid overly-overloading MDS, introduce a limit on number of
3608 MDS RPCs in flight for a single client and add /proc controls
3609 to adjust this limit.
3611 Severity : enhancement
3613 Description: client read/write statistics
3614 Details : Add client read/write call usage stats for performance
3615 analysis of user processes.
3616 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3617 file access. extents_stats shows chunk size distribution.
3618 extents_stats_per_process show chunk size distribution per
3621 Severity : enhancement
3623 Description: mds statistics
3624 Details : Add detailed mds operations statistics in
3625 /proc/fs/lustre/mds/*/stats.
3629 Description: Failure of copying files with lustre special EAs.
3630 Details : Client side always return success for setxattr call for lustre
3631 special xattr (currently only "trusted.lov").
3636 Description: Refcount LNET uuids
3637 Details : The global LNET uuid list grew linearly with every startup;
3638 refcount repeated list entries instead of always adding to
3641 Severity : enhancement
3643 Description: Dynamic service threads
3644 Details : Within a small range, start extra service threads
3645 automatically when the request queue builds up.
3648 Frequency : mixed-endian client/server environments
3650 Description: mixed-endian crashes
3651 Details : The new msg_v2 system had some failures in mixed-endian
3654 Severity : enhancement
3656 Description: Easy OST removal
3657 Details : OSTs can be permanently deactivated with e.g. 'lctl
3658 conf_param lustre-OST0001.osc.active=0'
3660 Severity : enhancement
3662 Description: MGS proc entries
3663 Details : Added basic proc entries for the MGS showing what filesystems
3666 Severity : enhancement
3668 Description: provide MGS failover
3669 Details : Added config lock reacquisition after MGS server failover.
3671 Severity : enhancement
3673 Description: add Linux 2.4 support
3674 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3678 Description: a large application tries to do I/O to the same resource and dies
3679 in the middle of it.
3680 Details : Check the req->rq_arrival time after the call to
3681 ost_brw_lock_get(), but before we do anything about
3682 processing it & sending the BULK transfer request. This
3683 should help move old stale pending locks off the queue as
3684 quickly as obd_timeout.
3687 Frequency : when an incorrect nid is specified during startup
3689 Description: ptlrpc connect to non-existant node causes kernel crash
3690 Details : LNET can't be re-entered from an event callback, which
3691 happened when we expire a message after the export has been
3692 cleaned up. Instead, hand the zombie cleanup off to another
3695 Severity : enhancement
3697 Description: plain/inodebits lock performance improvement
3698 Details : Grouping plain/inodebits in granted list by their request modes
3699 and bits policy, thus improving the performance of search through
3703 Frequency : only if OST filesystem is corrupted
3705 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3706 Details : for a short time RPCs with bulk IO are in the replay list,
3707 but replay of bulk IOs is unimplemented. If the OST filesystem
3708 is corrupted due to disk cache incoherency and then replay is
3709 started it is possible to trip an assertion. Avoid putting
3710 committed RPCs into the replay list at all to avoid this issue.
3713 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3715 Description: System hang on startup
3716 Details : This bug allowed the liblustre (e.g. catamount) client to
3717 return to the app before handling all startup RPCs. This
3718 could leave the node unresponsive to lustre network traffic
3719 and manifested as a server ptllnd timeout.
3721 Severity : enhancement
3723 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3724 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3725 boolean to control whether to print peer debug info when a
3726 client's RPC times out.
3729 Frequency : only for kernels with patches from Lustre below 1.4.3
3731 Description: Remove old rdonly API
3732 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3735 Frequency : only for devices with external journals
3737 Description: Set external device read-only also
3738 Details : During a commanded failover stop, we set the disk device
3739 read-only while the server shuts down. We now also set any
3740 external journal device read-only at the same time.
3743 Frequency : when upgrading from 1.4 while trying to change parameters
3745 Description: The wrong (new) MDC name was used when setting parameters for
3746 upgraded MDT's. Also allows changing of OSC (and MDC)
3747 parameters if --writeconf is specified at tunefs upgrade time.
3750 Frequency : when setting specific ost indicies
3752 Description: QOS code breaks on skipped indicies
3753 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3754 created with --index need not be sequential.
3756 Severity : enhancement
3758 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3759 Details : The uninit_groups feature works in conjunction with the kernel
3760 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3761 up the pass1 processing of e2fsck. This is a read-only feature
3762 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3763 mount filesystems that have had this feature enabled.
3765 Severity : enhancement
3767 Description: Improve multi-block allocation algorithm to avoid fragmentation
3768 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3769 allocation locality and avoid filesystem fragmentation.
3771 ------------------------------------------------------------------------------
3773 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3775 * Support for kernels:
3776 2.4.21-47.0.1.EL (RHEL 3)
3777 2.6.5-7.283 (SLES 9)
3778 2.6.9-42.0.10.EL (RHEL 4)
3779 2.6.12.6 vanilla (kernel.org)
3780 2.6.16.27-0.9 (SLES 10)
3781 * Recommended e2fsprogs version: 1.39.cfs5
3783 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3787 Frequency : occasional, depends on client load and configuration
3788 Bugzilla : 12181, 12203
3789 Description: data loss for recently-modified files
3791 Details : In some cases it is possible that recently written or created
3792 files may not be written to disk in a timely manner (this should
3793 normally be within 30s unless client IO load is very high).
3794 The problem appears as zero-length files or files that are a
3795 multiple of 1MB in size after a client crash or client eviction
3796 that are missing data at the end of the file.
3798 This problem is more likely to be hit on clients where files are
3799 repeatedly created and unlinked in the same directory, clients
3800 have a large amount of RAM, have many CPUs, the filesystem has
3801 many OSTs, the clients are rebooted frequently, and/or the files
3802 are not accessed by other nodes after being written.
3804 The presence of the problem can be detected by looking at
3805 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3806 smaller than the second (nr_unused) then dirty files will not
3807 be flushed automatically to disk. "sync; sleep 10" should be
3808 run several times on the node before unmounting it to update
3809 Lustre (this is also safe to run on nodes without this problem).
3811 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3812 that can cause this same problem, so customers using that kernel
3813 also need to update the kernel in addition to Lustre. In order
3814 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3816 It is normal that files written just before a client crash (less
3817 than 30s) may not yet have been flushed to disk, even for local
3821 Frequency : frequent on thin XT3 nodes
3823 Description: UUID collision on thin XT3 Linux nodes
3824 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3825 randomly, since we relied on an insufficiently-seeded PRNG.
3830 Description: OSS hangs after "All ost request buffers busy"
3831 Details : A deadlock between quota and journal operations caused OSS
3832 hangs after printing "All ost request buffers busy."
3835 Frequency : always on liblustre builds
3837 Description: Cleanup compiler warnings on liblustre
3840 Frequency : always on liblustre builds on XT3
3842 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3846 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3847 > 32000 subdirectories
3848 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3849 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3850 subdirectory count crosses 32000. This will aid e2fsck to
3851 correctly handle more than 32000 subdirectories.
3854 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3856 Description: System hang on startup
3857 Details : This bug allowed the liblustre (e.g. catamount) client to
3858 return to the app before handling all startup RPCs. This
3859 could leave the node unresponsive to lustre network traffic
3860 and manifested as a server ptllnd timeout.
3862 Severity : enhancement
3864 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3865 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3866 boolean to control whether to print peer debug info when a
3867 client's RPC times out.
3872 Description: make O_SYNC working on 2.6 kernels
3873 Details : 2.6 kernels use different method for mark pages for write,
3874 so need add a code to lustre for O_SYNC work.
3879 Description: Failure to close file and release space on NFS
3880 Details : Put inode details into lock acquired in ll_intent_file_open.
3881 Use mdc_intent_lock in ll_intent_open to properly
3882 detect all kind of errors unhandled by mdc_enqueue.
3887 Description: proc file read during shutdown sometimes raced obd removal,
3889 Details : Add lock to prevent obd access after proc file removal.
3892 Frequency : Only for files larger than 4GB on 32-bit clients.
3894 Description: improperly doing page alignment of locks
3895 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3896 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3901 Description: RPCs being resent when they shouldn't be
3902 Details : Some RPCs that should not be resent are being resent. This
3903 can cause inconsistencies in the RPC state machine. Do not
3904 resend such requests.
3907 Frequency : rare, only with NFS export
3909 Description: Crash on NFS re-export node
3910 Details : under very unusual load conditions an assertion is hit in
3911 ll_intent_file_open()
3914 Frequency : only if OST filesystem is corrupted
3916 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3917 Details : for a short time RPCs with bulk IO are in the replay list,
3918 but replay of bulk IOs is unimplemented. If the OST filesystem
3919 is corrupted due to disk cache incoherency and then replay is
3920 started it is possible to trip an assertion. Avoid putting
3921 committed RPCs into the replay list at all to avoid this issue.
3926 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3927 Details : Large single O_DIRECT read and write calls can fail to allocate
3928 a sufficiently large buffer to process the request. In case of
3929 allocation failure the allocation is retried with a smaller
3930 buffer and broken into smaller requests.
3932 Severity : enhancement
3934 Description: Add -o localflock option to simulate old noflock behaviour.
3935 Details : This will achieve local-only flock/fcntl locks coherentness.
3940 Description: versioning check is incomplete
3941 Details : Checking the version difference of client vs. server, report
3942 error if the gap is too big.
3947 Description: add support PG_writeback bit
3948 Details : add support for PG_writeback bit for Lustre, for more carefull
3949 work with page cache in 2.6 kernel. This also fix some deadlocks
3950 and remove hack for work O_SYNC with 2.6 kernel.
3952 Severity : enhancement
3954 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3955 Details : The uninit_groups feature works in conjunction with the kernel
3956 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3957 up the pass1 processing of e2fsck. This is a read-only feature
3958 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3959 mount filesystems that have had this feature enabled.
3961 Severity : enhancement
3963 Description: Improve multi-block allocation algorithm to avoid fragmentation
3964 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3965 allocation locality and avoid filesystem fragmentation.
3967 ------------------------------------------------------------------------------
3969 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3971 * Support for kernels:
3972 2.6.9-42.0.3.EL (RHEL 4)
3973 2.6.5-7.276 (SLES 9)
3974 2.4.21-47.0.1.EL (RHEL 3)
3975 2.6.12.6 vanilla (kernel.org)
3976 2.6.16.21-0.8 (SLES10)
3977 * Recommended e2fsprogs version: 1.39.cfs2-0
3979 * The backwards-compatible /proc/sys/portals symlink has been removed
3980 in this release. Before upgrading, please ensure that you change
3981 any configuration scripts or /etc/sysctl.conf files that access
3982 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3983 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3984 in advance of the upgrade on any system running Lustre 1.4.6 or
3985 newer, since /proc/sys/lnet was added in that version.
3986 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3990 Frequency : only when quota is used
3992 Description: avoid scanning export list for quota master
3993 Details : Change the algorithms to avoid scanning export list in order
3994 to improve the efficiency.
3997 Frequency : MDS failover only, very rarely
3999 Description: "went back in time" messages on mds failover
4000 Details : The greatest transno may be lost when the current operation
4001 finishes with an error (transno==0) and the client's last_rcvd
4002 record is over-written. Save the greatest transno in the
4003 mds_last_transno for this case.
4006 Frequency : always for specific kernels and striping counts
4008 Description: client may get "Matching packet too big" without ACL support
4009 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4010 when trying to access files in certain configurations. The
4011 clients should in fact be denied when mounting because they do
4012 not understand ACLs.
4015 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4017 Description: many clients connecting with IO in progress causes connect timeouts
4018 Details : Avoid synchronous journal commits to avoid delays caused by many
4019 clients connecting/disconnecting when bulk IO is in progress.
4020 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4021 OST_IO_PORTAL to avoid delays behind potentially many pending
4025 Frequency : occasionally with multiple writers to a single file
4027 Description: shared writes to file may result in wrong size reported by stat()
4028 Details : Allow growing of kms when extent lock is cancelled
4031 Frequency : always with random mmap IO to multi-striped file
4033 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4034 Details : Only if the hole is at the end of OST object so that kms is too
4035 small. Fix is to increase kms accordingly in ll_nopage.
4038 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4040 Description: writes to a missing object would leak memory on the OST
4041 Details : If there is an inconsistency between the MDS and OST filesystems,
4042 such that the MDS references an object that doesn't exist, writes
4043 to that object will leak memory due to incorrect cleanup in the
4044 error handling path, eventually running out of memory on the OST.
4049 Description: Creating too long symlink causes lustre errors
4050 Details : Check symlink and name lengths before sending requests to MDS.
4053 Frequency : only if flock is enabled (not on by default)
4055 Description: posix locks not released on fd closure on 2.6.9+
4056 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4057 kernels, this caused such locks not to be released on fd close and
4058 then assertions on fs unmount about still used locks.
4061 Frequency : MDS failover only, very rarely
4063 Description: clients may get ASSERTION(granted_lock != NULL)
4064 Details : When request was taking a long time, and a client was resending
4065 a getattr by name lock request. The were multiple lock requests
4066 with the same client lock handle and
4067 mds_getattr_name->fixup_handle_for_resent_request found one of the
4068 lock handles but later failed with ASSERTION(granted_lock != NULL).
4073 Description: handle->h_buffer_credits > 0, assertion failure
4074 Details : h_buffer_credits is zero after truncate, causing assertion
4075 failure. This patch extends the transaction or creates a new
4079 Frequency : NFS re-export or patchless client
4080 Bugzilla : 11179, 10796
4081 Description: Crash on NFS re-export node (__d_move)
4082 Details : We do not want to hash the dentry if we don't have a lock.
4083 But if this dentry is later used in d_move, we'd hit uninitialised
4084 list head d_hash, so we just do this to init d_hash field but
4085 leave dentry unhashed.
4088 Frequency : NFS re-export or patchless client
4090 Description: NFS exports has problem with symbolic link
4091 Details : lustre client didn't properly install dentry when re-exported
4092 to NFS or running patchless client.
4095 Frequency : NFS re-export or patchless client
4097 Description: Various nfs/patchless fixes.
4098 Details : fixes reuse disconected alias for lookup process - this fixes
4099 warning "find_exported_dentry: npd != pd",
4100 fix permission error with open files at nfs.
4101 fix apply umask when do revalidate.
4104 Frequency : occasional
4106 Description: Crash on NFS re-export node
4107 Details : calling clear_page() on the wrong pointer triggered oops in
4108 generic_mapping_read().
4111 Frequency : rarely, using O_DIRECT IO
4113 Description: unaligned directio crashes client with LASSERT
4114 Details : check for unaligned buffers before trying any requests.
4117 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4119 Description: stale data returned from RAID cache
4120 Details : If only a small amount of IO is done to the RAID device before
4121 reading it again it is possible to get stale data from the RAID
4122 cache instead of reading it from disk.
4125 Frequency : always for sles10 kernel
4127 Description: sles10 support
4128 Details : ll_follow_link: compile fixes and using of nd_set_link
4129 under newer kernels.
4132 Frequency : depends on arch, kernel and compiler version, always on sles10
4135 Description: recursive or deep enough symlinks cause stack overflow
4136 Details : getting rid of large stack-allocated variable in
4140 Frequency : depends on hardware
4142 Description: lustre write performance loss in the SLES10 kernel
4143 Details : the performance loss is caused by using of write barriers in the
4144 ext3 code. The SLES10 kernel turns barrier support on by
4145 default. The fix is to undo that change for ldiskfs.
4147 ------------------------------------------------------------------------------
4149 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4151 * Support for kernels:
4152 2.6.9-42.0.3EL (RHEL 4)
4153 2.6.5-7.276 (SLES 9)
4154 2.4.21-47.0.1.EL (RHEL 3)
4155 2.6.12.6 vanilla (kernel.org)
4159 Frequency : quota enabled and large files being deleted
4161 Description: releasing more than 4GB of quota at once hangs OST
4162 Details : If a user deletes more than 4GB of files on a single OST it
4163 will cause the OST to spin in an infinite loop. Release
4164 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4169 Description: statfs data retrieved from /proc may be stale or zero
4170 Details : When reading per-device statfs data from /proc, in the
4171 {kbytes,files}_{total,free,avail} files, it may appear
4172 as zero or be out of date.
4175 Frequency : systems with MD RAID1 external journal devices
4177 Description: lconf's call to blkid is confused by RAID1 journal devices
4178 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4179 of returning all block devices that match the journal UUID.
4182 Frequency : always, for aggregate stripe size over 4GB
4184 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4185 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4186 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4189 Frequency : always if "lfs find" used on a local file/directory
4191 Description: "lfs find" segfaults if used on a local file/directory
4192 Details : The case where a directory component was not specified wasn't
4193 handled correctly. Handle this properly.
4196 Frequency : always on ppc64
4198 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4199 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4200 perform endianness conversion.
4203 Frequency : rarely (truncate to non-zero file size after write under load)
4204 Bugzilla : 10730, 10687
4205 Description: Files padded with zeros to next 4K multiple
4206 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4207 it is possible that files that are truncated to a non-zero size
4208 immediately after being written are filled with zero bytes beyond
4209 the truncated size. No file data is lost.
4211 Severity : enhancement
4213 Description: Allow recovery/failover for liblustre clients.
4214 Details : liblustre clients were unaware of failover configurations until
4217 Severity : enhancement
4219 Description: user file locks should fail when not mounting with flock option
4220 Details : Set up an error-returning stub in ll_file_operations.lock field
4221 to prevent incorrect behaviour when client is mounted without
4222 flock option. Also, set up properly f_op->flock field for
4226 Frequency : always on ia64
4228 Description: "lfs df" loops on printing out MDS statfs information
4229 Details : The obd_ioctl_data was not initialized and in some systems
4230 this caused a failure during the ioctl that did not return
4231 an error. Initialize the struct and return an error on failure.
4234 Frequency : SLES 9 only
4236 Description: Error of copying files with lustre special EAs as root
4237 Details : Client side always return success for setxattr call for lustre
4238 special xattr (currently only "trusted.lov").
4241 Frequency : rarely on clusters with both ia64+i386 clients
4243 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4244 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4245 clients are doing concurrent writes to the same file it is
4246 possible that the ia64 clients may LASSERT because the OST
4247 extent locks are not PAGE_SIZE aligned. Ensure that grown
4248 locks are always aligned on the request boundary.
4251 Frequency : specific use, occasional
4253 Description: Overwriting in use executable truncates on-disk binary image
4254 Details : If one node attempts to overwrite an executable in use by
4255 another node, we now correctly return ETXTBSY instead of
4256 truncating the file.
4258 Severity : enhancement
4260 Description: Async OSC create to avoid the blocking unnecessarily.
4261 Details : If an OST has no remaining object, system will block on the creating
4262 when need to create a new object on this OST. Now, ways use
4263 pre-created objects when available, instead of blocking on an
4264 empty osc while others are not empty. If we must block, we block
4265 for the shortest possible period of time.
4270 Description: chmod on Lustre root is propagated to other clients
4271 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4272 invalid by the follow_mount time.
4275 Frequency : liblustre clients only
4277 Description: Race in 'instant cancel' lock handling could lead to such locks
4278 never to be granted in case of SMP MDS
4279 Details : Do not destroy not yet granted but cbpending locks in
4283 Frequency : replay/resend of open
4285 Description: non null lock assetion failure in mds_intent_policy
4286 Details : Trying to replay/resend lockless open requests resulted in
4287 mds_open() returning 0 with no lock. Now it sets a flag if
4288 a lock is going to be returned.
4290 Severity : enhancement
4292 Description: Checksum enhancements
4293 Details : New checksum enhancements allow for resending RPCs that failed
4296 Severity : enhancement
4298 Description: Tunables on number of dirty pages in cacche
4299 Details : Allow to set limit on number of dirty pages cached.
4304 Description: client crash on unmount - lock still has references
4305 Details : In some error handling cases it was possible to leak a lock
4306 reference on a client while accessing a file. This was not
4307 harmful to the client during operation, but would cause the
4308 client to crash when the filesystem is unmounted.
4311 Frequency : specific case, rare
4313 Description: ETXTBSY on mds though file not in use by client
4314 Details : ETXTBSY is no longer incorrectly returned when attempting to
4315 chmod or chown a directory that the user previously tried to
4316 execute or a currently-executing binary.
4319 Frequency : extremely rare except on liblustre-based clients
4321 Description: Lustre space not freed when files are deleted
4322 Details : Clean up open-unlinked files after client eviction. Previously
4323 the unlink was skipped and the files remained as orphans.
4328 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4329 Details : In some cases it was possible to send a blocking callback to
4330 a client doing a glimpse, even though that client didn't get
4331 a lock granted. When the glimpse lock is cancelled on the OST
4332 the freed lock is left on the waiting list and corrupted the list.
4335 Frequency : all core dumps
4337 Description: Broke core dumps to lustre
4338 Details : Negative dentry may be unhashed if parent does not have UPDATE
4339 lock, but some callers, e.g. do_coredump, expect dentry to be
4340 hashed after successful create, hash it in ll_create_it.
4342 ------------------------------------------------------------------------------
4344 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4346 * Support for kernels:
4347 2.6.9-42.0.2.EL (RHEL 4)
4348 2.6.5-7.276 (SLES 9)
4349 2.4.21-40.EL (RHEL 3)
4350 2.6.12.6 vanilla (kernel.org)
4354 Frequency : always on RHEL 3
4356 Description: Number of open files grows over time
4357 Details : The number of open files grows over time, whether or not
4358 Lustre is started. This was due to a filp leak introduced
4359 by one of our kernel patches.
4361 ------------------------------------------------------------------------------
4363 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4365 * Support for kernels:
4366 2.6.9-42.EL (RHEL 4)
4367 2.6.5-7.276 (SLES 9)
4368 2.4.21-40.EL (RHEL 3)
4369 2.6.12.6 vanilla (kernel.org)
4374 Bugzilla : 5719, 9635, 9792, 9684
4375 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4376 Details : If a server is under heavy load and cannot reply to new
4377 connection requests before the client resends the (re)connect,
4378 the connection handling code can behave badly if two service
4379 threads are concurrently handing separate (re)connections from
4380 the same client. Add better locking to the connection handling
4381 code, and ensure that only a single connection will be processed
4382 for a given client UUID, even if the lock is dropped.
4384 Severity : enhancement
4386 Description: add TCP zero-copy support to kernel
4387 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4388 sends if the hardware supports scatter-gather and checksumming.
4389 This allows socklnd to do client-write and server-read more
4390 efficiently and reduce CPU utilization from skbuf copying.
4393 Frequency : only if NFS exporting from client
4395 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4396 Details : If setting attributes on a file created under NFS that had
4397 never been opened it would be possible to oops the client
4398 if the file had no objects.
4401 Frequency : always for liblustre
4403 Description: liblustre client does MDS+OSTs setattr RPC for each write
4404 Details : When doing a write from a liblustre client, the client
4405 incorrectly issued an RPC to the MDS and each OST the file was
4406 striped over in order to update the timestamps. When writing
4407 with small chunks and many clients this could overwhelm the MDS
4408 with RPCs. In all cases it would slow down the write because
4409 these RPCs are unnecessary.
4411 Severity : enhancement
4413 Description: allow number of MDS service threads to be changed at module load
4414 Details : It is now possible to change the number of MDS service threads
4415 running. Adding "options mds mds_num_threads={N}" to the MDS's
4416 /etc/modprobe.conf will set the number of threads for the next
4417 time Lustre is restarted (assuming the "mds" module is also
4418 reloaded at that time). The default number of threads will
4419 stay the same, 32 for most systems.
4424 Description: OST crash if filesystem is unformatted or corrupt
4425 Details : If an OST is started on a device that has never been formatted
4426 or if the filesystem is corrupt and cannot even mount then the
4427 error handling cleanup routines would dereference a NULL pointer.
4432 Description: NULL pointer deref in llap_from_page.
4433 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4434 incorrect) mapping if the page was truncated/reclaimed while it was
4435 searched for. Check for this condition and skip such pages when
4436 doing readahead. Introduce extra check to llap_from_page() to
4437 verify page->mapping->host is non-NULL (so page is not anonymous).
4440 Frequency : Sometimes when using sys_sendfile
4442 Description: "page not covered by a lock" warnings from ll_readpage
4443 Details : sendfile called ll_readpage without right page locks present.
4444 Now we introduced ll_file_sendfile that does necessary locking
4445 around call to generic_file_sendfile() much like we do in
4449 Frequency : with certain MDS communication failures at client mount time
4451 Description: NULL pointer deref after failed client mount
4452 Details : a client connection request may delayed by the network layer
4453 and not be sent until after the PTLRPC layer has timed out the
4454 request. If the client fails the mount immediately it will try
4455 to clean up before the network times out the request. Add a
4456 reference from the request import to the obd device and delay
4457 the cleanup until the network drops the request.
4460 Frequency : occasionally during client (re)connect
4462 Description: assertion failure during client (re)connect
4463 Details : processing a client connection request may be delayed by the
4464 client or server longer than the client connect timeout. This
4465 causes the client to resend the connection request. If the
4466 original connection request is replied in this interval, the
4467 client may trip an assertion failure in ptlrpc_connect_interpret()
4468 which thought it would be the only running connect process.
4471 Frequency : only with obd_echo servers and clients that are rebooted
4473 Description: kernel BUG accessing uninitialized data structure
4474 Details : When running an obd_echo server it did not start the ping_evictor
4475 thread, and when a client was evicted an uninitialized data
4476 structure was accessed. Start the ping_evictor in the RPC
4477 service startup instead of the OBD startup.
4479 Severity : enhancement
4480 Bugzilla : 10193 (patchless)
4481 Description: Remove dependency on various unexported kernel interfaces.
4482 Details : No longer need reparent_to_init, exit_mm, exit_files,
4483 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4486 Frequency : rare (only users of deprecated and unsupported LDAP config)
4488 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4489 Details : LDAP apparently contains 'lustreName' attributes instead of
4490 'name'. A simple remapping of the name is sufficient.
4493 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4495 Description: waiting_locks_callback trips kernel BUG if client is evicted
4496 Details : Running with the dump_on_timeout debug flag turned on makes
4497 it possible that the waiting_locks_callback() can try to dump
4498 the Lustre kernel debug logs from an interrupt handler. Defer
4499 this log dumping to the expired_lock_main() thread.
4501 Severity : enhancement
4503 Description: Support NFS exporting on 2.6 kernels.
4504 Details : Implement non-rawops metadata methods for NFS server to use without
4505 changing NFS server code.
4508 Frequency : very rare (synthetic metadata workload only)
4510 Description: two racing renames might cause an MDS thread to deadlock
4511 Details : Running the "racer" program may cause one MDS thread to rename
4512 a file from being the source of a rename to being the target of
4513 a rename at exactly the same time that another thread is doing
4514 so, and the second thread has already enqueued these locks after
4515 doing a lookup of the target and is trying to relock them in
4516 order. Ensure that we don't try to re-lock the same resource.
4519 Frequency : only very large systems with liblustre clients
4521 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4522 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4523 all OSTs in parallel. This allows the eviction of stale liblustre
4524 clients to proceed much faster than if they were done in series,
4525 and also offers similar improvements for other set_info RPCs.
4530 Description: excessive CPU usage during initial read phase on client
4531 Details : During the initial read phase on a client, it would agressively
4532 retry readahead on the file, consuming too much CPU and impacting
4533 performance (since 1.4.5.8). Improve the readahead algorithm
4534 to avoid this, and also improve some other common cases (read
4535 of small files in particular, where "small" is files smaller than
4536 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4541 Description: MDS crash when receiving packet with unknown intent.
4542 Details : Do not LBUG in unknown intent case, just return -EFAULT
4544 Severity : enhancement
4545 Bugzilla : 9293, 9385
4546 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4547 Details : Do not serialize getattr (non-intent version) and statfs.
4550 Frequency : occasional, when OST network is overloaded/intermittent
4552 Description: client evicted by OST after bulk IO timeout
4553 Details : If a client sends a bulk IO request (read or write) the OST
4554 may evict the client if it is unresposive to its data GET/PUT
4555 request. This is incorrect if the network is overloaded (takes
4556 too long to transfer the RPC data) or dropped the OST GET/PUT
4557 request. There is no need to evict the client at all, since
4558 the pinger and/or lock callbacks will handle this, and the
4559 client can restart the bulk request.
4562 Frequency : Always when mmapping file with no objects
4564 Description: client crashes when mmapping file with no objects
4565 Details : Check that we actually have objects in a file before doing any
4566 operations on objects in ll_vm_open, ll_vm_close and
4572 Description: Request leak when working with deleted CWD
4573 Details : Introduce advanced request refcount tracking for requests
4574 referenced from lustre intent.
4576 Severity : Enhancement
4578 Description: Cache open file handles on client.
4579 Details : MDS now will return special lock along with openhandle, if
4580 requested and client is allowed to hold openhandle, even if unused,
4581 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4582 closing files for every read/write openration.
4584 Severity : Enhancement
4586 Description: Cache open negative dentries on client when possible.
4587 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4588 negative dentries on lock revocation.
4593 Description: Remounting a client read-only wasn't possible with a zconf mount
4594 Details : It wasn't possible to remount a client read-only with llmount.
4596 Severity : enhancement
4597 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4598 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4599 MPICH is included to add Lustre-specific ADIO interfaces.
4600 This is based closely on the UFS ADIO layer and only differs
4601 in file creation, in order to allow the OST striping to be set.
4602 This is user-contributed code and not supported by CFS.
4607 Description: extended inode attributes (immutable, append-only) work improperly
4608 when 2.4 and 2.6 kernels are used on client/server or vice versa
4609 Details : Introduce kernel-independent values for these flags.
4611 Severity : enhancement
4614 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4615 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4616 granularity. Now, Fractional MB granularity is be supported,
4617 this is very useful for low memory system.
4619 Severity : enhancement
4621 Description: Getattr by fid
4622 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4623 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4624 have locks and avoids getattr rpc requests for it on every stat.
4627 Frequency : Always, for filesystems larger than 2TB
4629 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4630 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4631 filesystems larger than 2TB to be mounted with mballoc enabled.
4634 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4636 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4637 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4638 filesystems larger than 2TB is mounted then the kernel will
4639 silently corrupt the start of the filesystem. CONFIG_LBD is
4640 enabled for all CFS-supported kernels, but the possibility of
4641 this happening with a modified kernel config exists.
4643 Severity : enhancement
4645 Description: add client O_DIRECT support for 2.6 kernels
4646 Details : It is now possible to do O_DIRECT reads and writes to files
4647 in the Lustre client mountpoint on 2.6 kernel clients.
4649 Severity : enhancement
4651 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4652 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4653 parallel, not waiting for response from every OST before sending
4654 a rpc to the next OST.
4659 Description: setattr vs write race when updating file timestamps
4660 Details : Client processes that update a file timestamp into the past
4661 right after writing to the file (e.g. tar) it is possible that
4662 the updated file modification time can be reset to the current
4663 time due to a race between processing the setattr and write RPC.
4665 Severity : enhancement
4667 Description: Bring 'lfs find' closer in line with regular Linux find.
4668 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4669 -print0 options and obtains all the needed info through the lustre
4672 Severity : enhancement
4674 Description: support up to 1024 configured devices on one node
4675 Details : change obd_dev array from statically allocated to dynamically
4676 allocated structs as they are first used to reduce memory usage
4681 Description: Flush dirty partially truncated pages during truncate
4682 Details : Immediatelly flush partially truncated pages in filter_setattr,
4683 this way we completely avoid having any pages in page cache on OST
4684 and can retire ugly workarounds during writes to flush such pages.
4689 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4690 Details : protect inode from truncation within vfs_unlink() context
4691 just take a reference before calling vfs_unlink() and release it
4692 when parent's i_sem is free.
4697 Description: last_id value checked outside lock on OST caused LASSERT failure
4698 Details : If there were multiple MDS->OST object precreate requests in
4699 flight, it was possible that the OST's last object id was checked
4700 outside a lock and incorrectly tripped an assertion. Move checks
4701 inside locks, and discard old precreate requests.
4704 Frequency : always, if extents are used on OSTs
4706 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4707 Details : OSTs using the extents format would not zero the high 16 bits of
4708 the index physical block number. This is not a problem for any
4709 OST filesystems smaller than 16TB, and no kernels support ext3
4710 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4711 new/modified files) and can be fixed for existing filesystems
4712 with e2fsprogs-1.39-cfs1.
4717 Description: import connection selection may be incorrect if timer wraps
4718 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4719 import connections to be ignored if the 32-bit jiffies counter
4720 wraps. Use a 64-bit jiffies counter.
4723 Frequency : very large clusters immediately after boot
4725 Description: LNET request buffers exhausted under heavy short-term load
4726 Details : If a large number of client requests are generated on a service
4727 that has previously never seen so many requests it is possible
4728 that the request buffer growth cannot keep up with the spike in
4729 demand. Instead of dropping incoming requests, they are held in
4730 the LND until the RPC service can accept more requests.
4733 Frequency : Sometimes during replay
4735 Description: Assertion failure in ll_local_open after replay.
4736 Details : If replay happened on an open request reply before we were able
4737 to set replay handler, reply will become not swabbed tripping the
4738 assertion in ll_local_open. Now we set the handler right after
4739 recognising of open request
4742 Frequency : very rare
4744 Description: kernel reports "badness in vsnprintf"
4745 Details : Reading from the "recovery_status" /proc file in small chunks
4746 may cause a negative length in lprocfs_obd_rd_recovery_status()
4747 call to vsnprintf() (which is otherwise harmless). Exit early
4748 if there is no more space in the output buffer.
4750 Severity : enhancement
4752 Description: clear OBD RPC statistics by writing to them
4753 Details : It is now possible to clear the OBD RPC statistics by writing
4754 to the "stats" file.
4759 Description: Client mtime is not the same on different clients after utimes
4760 Details : In some cases, the client was using the utimes() syscall on
4761 a file cached on another node. The clients now validate the
4762 ctime from the MDS + OSTs to determine which one is right.
4767 Description: Inability to activate failout mode
4768 Details : lconf script incorrectly assumed that in pythong string's numeric
4769 value is used in comparisons.
4772 Frequency : always with multiple stripes per file
4774 Description: Inefficient object allocation for mutli-stripe files
4775 Details : When selecting which OSTs to stripe files over, for files with
4776 a stripe count that divides evenly into the number of OSTs,
4777 the MDS is always picking the same starting OST for each file.
4778 Return the OST selection heuristic to the original design.
4783 Description: mount failures may take full timeout to return an error
4784 Details : Under some heavy load conditions it is possible that a
4785 failed mount can wait for the full obd_timeout interval,
4786 possibly several minutes, before reporting an error.
4787 Instead return an error as soon as the status is known.
4789 Frequency : quota enabled and large files being deleted
4791 Description: releasing more than 4GB of quota at once hangs OST
4792 Details : If a user deletes more than 4GB of files on a single OST it
4793 will cause the OST to spin in an infinite loop. Release
4794 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4799 Description: statfs data retrieved from /proc may be stale or zero
4800 Details : When reading per-device statfs data from /proc, in the
4801 {kbytes,files}_{total,free,avail} files, it may appear
4802 as zero or be out of date.
4805 Frequency : systems with MD RAID1 external journal devices
4807 Description: lconf's call to blkid is confused by RAID1 journal devices
4808 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4809 of returning all block devices that match the journal UUID.
4812 Frequency : always, for aggregate stripe size over 4GB
4814 Description: assertion fails when trying to use 4GB stripe size
4815 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4816 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4819 Frequency : always on ppc64
4821 Description: the first write on an ext3 filesystem with mballoc got stuck
4822 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4823 perform endianness conversion.
4825 ------------------------------------------------------------------------------
4827 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4829 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4830 INTEROPERATE with older versions automatically. Please read the
4831 user documentation before upgrading any part of a live system.
4832 * WARNING: Lustre networking configuration changes are required with
4833 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4836 * Support for kernels:
4837 2.6.9-22.0.2.EL (RHEL 4)
4838 2.6.5-7.244 (SLES 9)
4839 2.6.12.6 vanilla (kernel.org)
4842 Severity : enhancement
4843 Bugzilla : 7981/8208
4844 Description: Introduced Lustre Networking (LNET)
4845 Details : LNET is new networking infrastructure for Lustre, it includes
4846 a reorganized network configuration mode (see the user
4847 documentation for full details) as well as support for routing
4848 between different network fabrics. Lustre Networking Devices
4849 (LNDs) for the supported network fabrics have also been
4850 created for this new infrastructure.
4852 Severity : enhancement
4853 Description: Introduced Access control lists
4854 Details : clients can set ACLs on files and directories in order to have
4855 more fine-grained permissions than the standard Unix UGO+RWX.
4856 The MDS must be started with the "-o acl" mount option.
4858 Severity : enhancement
4859 Description: Introduced filesystem quotas
4860 Details : Administrators may now establish per-user quotas on the
4863 Severity : enhancement
4865 Description: Configuration change for the XT3
4866 The PTLLND is now used to run Lustre over Portals on the XT3
4867 The configure option(s) --with-cray-portals are no longer used.
4868 Rather --with-portals=<path-to-portals-includes> is used to
4869 enable building on the XT3. In addition to enable XT3 specific
4870 features the option --enable-cray-xt3 must be used.
4875 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4876 Details : A race between allocating a new llcd and re-getting the llcd_lock
4877 allowed another thread to grab newly-allocated llcd.
4879 Severity : enhancement
4881 Description: 2.6 OST async journal commit and locking fix to improve performance
4882 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4883 2.6 kernels are now async as they already were in 2.4 kernels so
4884 that they can commit concurrently with the network bulk transfer.
4885 For block-allocated files the filter allocation semaphore is held
4886 to avoid filesystem fragmentation during allocation. BKL lock
4887 removed for 2.6 xattr operations where it is no longer needed.
4892 Description: lconf incorrectly determined whether two IP networks could talk
4893 Details : In some more complicated routing and multiple-network
4894 configurations, lconf will avoid trying to make a network
4895 connection to a disjoint part of the IP space. It was doing the
4896 math incorrectly for one set of cases.
4901 Description: Fix for potential infinite loop processing records in an llog.
4902 Details : If an llog record is corrupted/zeroed, it is possible to loop
4903 forever in llog_process(). Validate the llog record length
4904 and skip the remainder of the block on error.
4907 Frequency : occasional (liblustre only)
4909 Description: liblustre could not open files whose last component is a symlink
4910 Details : sysio_path_walk() would incorrectly pass the open intent to
4911 intermediate path components.
4914 Frequency : rare (liblustre only with non-standard tuning)
4915 Bugzilla : 7201 (7350)
4916 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4917 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4918 after it has been released, only do this for referenced locks
4920 Severity : enhancement
4922 Description: specifying an (invalid) directory default stripe_size of -1
4923 would reset the directory default striping
4924 Details : stripe_size -1 was used internally to signal directory stripe
4925 removal, now use "all default" to signal dir stripe removal
4926 as a directory striping of "all default" is not useful
4929 Frequency : common for large clusters running liblustre clients
4931 Description: doing an ls when liblustre clients are running is slow
4932 Details : sending a glimpse AST to a liblustre client waits for every AST
4933 to time out, as liblustre clients will not respond. Since they
4934 cannot cache data we refresh the OST lock LVB from disk instead.
4936 Severity : enhancement
4938 Description: doing an ls at the same time as file IO can be slow
4939 Details : enqueue and other "small" requests can be blocked behind many
4940 large IO requests. Create a new OST IO portal for non-IO
4941 requests so they can be processed faster.
4944 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4946 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4947 created file with mode 000
4948 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4949 allow an ftruncate() as a non-root user to a file with mode 000.
4950 Lustre now respects this flag to disable mode checks when
4951 truncating a file owned by the user
4954 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4957 Description: Revoking locks from clients that went dead or catatonic might take
4959 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4960 cancellation of such locks instant on servers without waiting for
4961 any reply from clients. Clients drops these locks when cancel
4962 notification from server is received without replying.
4965 Frequency : liblustre-only, when liblustre client dies or becomes busy
4967 Description: Doing ls on Linux clients can take a long time with active
4969 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4970 granting such locks to it in the first place if possible. Locks
4971 are taken by proxy on the OST during the read or write and
4972 dropped immediately afterward. Add connect flags handling, do
4973 not grant locks to liblustre clients for glimpse ASTs.
4975 Severity : enhancement
4977 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4978 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4979 sequential reads and behaves badly with applications doing random
4980 reads. Improve it by reading ahead at least read region, and
4981 avoiding excessive large RPC for small reads.
4983 Severity : enhancement
4985 Description: Creating more than 1000 files for a single job may cause a load
4986 imbalance on the OSTs if there are also a large number of OSTs.
4987 Details : qos_prep_create() uses an OST index reseed value that is an
4988 even multiple of the number of available OSTs so that if the
4989 reseed happens in the middle of the object allocation it will
4990 still utilize the OSTs as uniformly as possible.
4995 Description: OST or MDS may oops in ping_evictor_main()
4996 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4997 but doesn't restart at beginning of obd_exports_timed list
5000 Severity : enhancement
5002 Description: improve by-nid export eviction on the MDS and OST
5003 Details : allow multiple exports with the same NID to be evicted at one
5004 time without re-searching the exports list.
5007 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5009 Description: MDS may oops in groups_free()
5010 Details : in rare race conditions a newly allocated group_info struct is
5011 freed again, and this can be NULL. The 2.4 compatibility code
5012 for groups_free() checked for a NULL pointer, but 2.6 did not.
5015 Frequency : common for liblustre clients doing little filesystem IO
5016 Bugzilla : 9352, 7313
5017 Description: server may evict liblustre clients accessing contended locks
5018 Details : if a client is granted a lock or receives a completion AST
5019 with a blocking AST already set it would not reply to the AST
5020 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5023 Frequency : lfs setstripe, only systems with more than 160 OSTs
5025 Description: unable to set striping with a starting offset beyond OST 160
5026 Details : llapi_create_file() incorrectly limited the starting stripe
5027 index to the maximum single-file stripe count.
5030 Frequency : LDAP users only
5032 Description: lconf did not handle in-kernel recovery with LDAP properly
5033 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5035 Severity : enhancement
5037 Description: bind OST threads to NUMA nodes to improve performance
5038 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5039 node and do their allocations there to localize memory access
5041 Severity : enhancement
5043 Description: llmount can determine client NID directly from Myrinet (GM)
5044 Details : the client NID code from gmnalnid was moved directly into
5045 llmount, removing the need to use this or specifying the
5046 client NID explicitly when mounting GM clients with zeroconf
5049 Frequency : if client is started with down MDS
5051 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5052 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5053 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5054 ptlrpc_import_delayed_req() because we didn't check if the
5055 request was interrupted, and we also didn't break out of the
5056 event loop if there was no timeout
5061 Description: data loss during non-page-aligned writes to a single file from
5062 both multiple nodes and multiple threads on one node at same time
5063 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5064 inconsistency led to false short-reads, that were cached and later
5065 used by ->prepare_write() to fill in partially written page,
5066 leading to data loss.
5069 Frequency : always, if lconf --abort_recovery used
5071 Description: lconf --abort_recovery fails with 'Operation not supported'
5072 Details : lconf was attempting to abort recovery on the MDT device and not
5075 Severity : enhancement
5077 Description: remove cleanup logs
5078 Details : replace lconf-generated cleanup logs with lustre internal
5079 cleanup routines. Eliminates the need for client-cleanup and
5082 Severity : enhancement
5084 Description: add support for EAs (user and system) on lustre filesystems
5085 Details : it is now possible to store extended attributes in the Lustre
5086 client filesystem, and with the user_xattr mount option it
5087 is possible to allow users to store EAs on their files also
5089 Severity : enhancement
5091 Description: Add possibility (config option) to show minimal available OST free
5093 Details : When compiled with --enable-mindf configure option, statfs(2)
5094 (and so, df) will return least minimal free space available from
5095 all OSTs as amount of free space on FS, instead of summary of
5096 free spaces of all OSTs.
5098 Severity : enhancement
5100 Description: do not expand extent locks acquired on OST-side
5101 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5102 by server: they are not cached anyway.
5105 Frequency : when mmap is used/binaries executed from Lustre
5107 Description: Unmmap pages before throwing them away from read cache.
5108 Details : llap_shrink cache now attempts to unmap pages before discarding
5109 them (if unmapping failed - do not discard). SLES9 kernel has
5110 extra checks that trigger if this unmapping is not done first.
5115 Description: lconf didn't resolve symlinks before checking to see whether a
5116 given mountpoint was already in use
5119 Frequency : when migrating failover services
5120 Bugzilla : 6395, 9514
5121 Description: When migrating a subset of services from a node (e.g. failback
5122 from a failover service node) the remaining services would
5123 time out and evict clients.
5124 Details : lconf --force (implied by --failover) sets the global obd_timeout
5125 to 5 seconds in order to quickly disconnect, but this caused
5126 other RPCs to time out too quickly. Do not change the global
5127 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5129 Severity : enhancement
5130 Frequency : if MDS is started with down OST
5131 Bugzilla : 9439,5706
5132 Description: Allow startup/shutdown of an MDS without depending on the
5133 availability of the OSTs.
5134 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5135 Add appropriate locking and lov-osc refcounts for safe
5136 cleaning. Add osc abort_inflight calls in case the
5137 synchronize never started.
5140 Frequency : occasional (Cray XT3 only)
5142 Description: root not authorized to access files in CRAY_PORTALS environment
5143 Details : The client process capabilities were not honoured on the MDS in
5144 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5145 been accessed by an authorized user then root was able to access
5146 the file on the local client also. The root user capabilities
5147 are now allowed on the MDS, as this environment has secure UID.
5150 Frequency : occasional
5152 Description: ldiskfs "too long searching" message happens too often
5153 Details : A debugging message (otherwise harmless) prints too often on
5154 the OST console. This has been reduced to only happen when
5155 there are fragmentation problems on the filesystem.
5160 Description: Division by zero in statfs when all OSCs are inactive
5161 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5162 lov_statfs divides by value returned by lov_get_stripecnt().
5166 Bugzilla : 9489, 3273
5167 Description: First write from each client to each OST was only 4kB in size,
5168 to initialize client writeback cache, which caused sub-optimal
5169 RPCs and poor layout on disk for the first written file.
5170 Details : Clients now request an initial cache grant at (re)connect time
5171 and so that they can start streaming writes to the cache right
5172 away and always do full-sized RPCs if there is enough data.
5173 If the OST is rebooted the client also re-establishes its grant
5174 so that client cached writes will be honoured under the grant.
5179 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5180 Details : Now I/O RPCs go to different portal number and (presumably) fast
5181 lock requests (and glimses) and other RPCs get their own service
5182 threads pool that should be able to service those RPCs
5185 Severity : enhancement
5187 Description: Ability to exchange lustre version between client and servers and
5188 issue warnings at client side if client is too old. Also for
5189 liblustre clients there is ability to refuse connection of too old
5191 Details : New 'version' field is added to connect data structure that is
5192 filled with version info. That info is later checked by server and
5196 Frequency : rare, liblustre only.
5197 Bugzilla : 9296, 9581
5198 Description: Two simultaneous writes from liblustre at offset within same page
5199 might proceed at the same time overwriting eachother with stale
5201 Details : I/O lock withing llu_file_prwv was released too early, before data
5202 actually was hitting the wire. Extended lock-holding time until
5203 server acknowledges receiving data.
5206 Frequency : extremely rare. Never observed in practice.
5208 Description: avoid generating lustre_handle cookie of 0.
5209 Details : class_handle_hash() generates handle cookies by incrementing
5210 global counter, and can hit 0 occasionaly (this is unlikely, but
5211 not impossible, because initial value of cookie counter is
5212 selected randonly). Value of 0 is used as a sentinel meaning
5213 "unassigned handle" --- avoid it. Also coalesce two critical
5214 sections in this function into one.
5216 Severity : enhancement
5218 Description: allow liblustre clients to delegate truncate locking to OST
5219 Details : To avoid overhead of locking, liblustre client instructs OST to
5220 take extent lock in ost_punch() on client's behalf. New connection
5221 flag is added to handle backward compatibility.
5223 Severity : enhancement
5224 Bugzilla : 4928, 7341, 9758
5225 Description: allow number of OST service threads to be specified
5226 Details : a module parameter allows the number of OST service threads
5227 to be specified via "options ost ost_num_threads={N}" in the
5228 OSS's /etc/modules.conf or /etc/modprobe.conf.
5232 Bugzilla : 6146, 9635, 9895
5233 Description: servers crash with bad pointer in target_handle_connect()
5234 Details : In rare cases when a client is reconnecting it was possible that
5235 the connection request was the last reference for that export.
5236 We would temporarily drop the export reference and get a new
5237 one, but this may have been the last reference and the export
5238 was just destroyed. Get new reference before dropping old one.
5240 Severity : enhancement
5241 Frequency : if client is started with failover MDS
5243 Description: Allow multiple MDS hostnames in the mount command
5244 Details : Try to read the configuration from all specified MDS
5245 hostnames during a client mount in case the "primary"
5248 Severity : enhancement
5250 Description: Stop sending data to evicted clients as soon as possible.
5251 Details : Check if the client we are about to send or are sending data to
5252 was evicted already. (Check is done every second of waiting,
5253 for which l_wait_event interface was extended to allow checking
5254 of exit condition at specified intervals).
5257 Frequency : rare, normally only when NFS exporting is done from client
5259 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5261 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5264 Frequency : always, if srand() is called before liblustre initialization
5266 Description: Liblustre uses system PRNG disturbing its usage by user application
5267 Details : Introduce internal to lustre fast and high-quality PRNG for
5268 lustre usage and make liblustre and some other places in generic
5269 lustre code to use it.
5271 Severity : enhancement
5272 Bugzilla : 9477, 9557, 9870
5273 Description: Verify that the MDS configuration logs are updated when xml is
5274 Details : Check if the .xml configuration logs are newer than the config
5275 logs stored on the MDS and report an error if this is the case.
5276 Request --write-conf, or allow starting with --old_conf.
5278 Severity : enhancement
5280 Description: Handle symlinks in the path when checking if Lustre is mounted.
5281 Details : Resolve intermediate symlinks when checking if a client has
5282 mounted a filesystem to avoid duplicate client mounts.
5287 Description: lconf can hit an error exception but still return success.
5288 Details : The lconf command catches the Command error exception at the top
5289 level script context and will exit with the associated exit
5290 status, but doesn't ensure that this exit status is non-zero.
5295 Description: failure of ptlrpc thread startup can cause oops
5296 Details : Starting a ptlrpc service thread can fail if there are a large
5297 number of threads or the server memory is very fragmented.
5298 Handle this without oopsing.
5301 Frequency : always, only if liblustre and non-default acceptor port was used
5303 Description: liblustre cannot connect to servers with non-default acceptor port
5304 Details : tcpnal_set_default_params() was not called and was therefore
5305 ignoring the environment varaible TCPNAL_PORT, as well as other
5306 TCPNAL_ environment variables
5311 Description: two objects could be created on the same OST for a single file
5312 Details : If an OST is down, in some cases it was possible to create two
5313 objects on a single OST for a single file. No problems other
5314 than potential performance impact and spurious error messages.
5318 Bugzilla : 5681, 9562
5319 Description: Client may oops in ll_unhash_aliases
5320 Details : Client dcache may become inconsistent in race condition.
5321 In some cases "getcwd" can fail if the current directory is
5327 Description: Inode refcounting problems in NFS export code
5328 Details : link_raw functions used to call d_instantiate without obtaining
5329 extra inode reference first.
5333 Bugzilla : 9942, 9903
5334 Description: Referencing freed requests leading to crash, memleaks with NFS.
5335 Details : We used to require that call to ll_revalidate_it was always
5336 followed by ll_lookup_it. Also with revalidate_special() it is
5337 possible to call ll_revalidate_it() twice for the same dentry
5338 even if first occurence returned success. This fix changes semantic
5339 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5340 reference on a request referred from the intent.
5341 ll_intent_release() then releases such a request.
5344 Frequency : rare, normally benchmark loads only
5346 Description: unlinked inodes were kept in memory on the client
5347 Details : If a client is repeatedly creating and unlinking files it
5348 can accumulate a lot of stale inodes in the inode slab cache.
5349 If there is no other client load running this can cause the
5350 client node to run out of memory. Instead flush old inodes
5351 from client cache that have the same inode number as a new inode.
5354 Frequency : SLES9 2.6.5 kernel and long filenames only
5355 Bugzilla : 9969, 10379
5356 Description: utime reports stale NFS file handle
5357 Details : SLES9 uses out-of-dentry names in some cases, which confused
5358 the lustre dentry revalidation. Change it to always use the
5362 Frequency : rare, unless heavy write-truncate concurrency is continuous
5363 Bugzilla : 4180, 6984, 7171, 9963, 9331
5364 Description: OST becomes very slow and/or deadlocked during object unlink
5365 Details : filter_destroy() was holding onto the parent directory lock
5366 while truncating+unlinking objects. For very large objects this
5367 may block other threads for a long time and slow overall OST
5368 responsiveness. It may also be possible to get a lock ordering
5369 deadlock in this case, or run out of journal credits because of
5370 the combined truncate+unlink. Solution is to do object truncate
5371 first in one transaction without parent lock, and then do the
5372 final unlink in a new transaction with the parent lock. This
5373 reduces the lock hold time dramatically.
5376 Frequency : rare, 2.4 kernels only
5378 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5379 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5380 before unmounting their filesystem in order to avoid deadlock.
5381 The kernel_locked() function in 2.4 kernels only checks whether
5382 the kernel lock is held, not whether it is this process that is
5383 holding it as 2.6 kernels do.
5388 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5389 Details : The client ptlrpc code may be trying to reconnect to a down
5390 server before a previous connection attempt has timed out.
5391 Increase the reconnect interval to be longer than the connection
5392 timeout interval to avoid sending duplicate connections to
5396 Frequency : echo_client brw_test command
5398 Description: fix echo_client to work with OST preallocated code
5399 Details : OST preallocation code (5137) didn't take echo_client IO path
5400 into account: echo_client calls filter methods outside of any
5401 OST thread and, hence, there is no per-thread preallocated
5402 pages and buffers to use. Solution: hijack pga pages for IO. As
5403 a byproduct, this avoids unnecessary data copying.
5407 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5408 Description: Client can oops in mdc_commit_close() after open replay
5409 Details : It was possible for the MDS to return an open request with no
5410 transaction number in mds_finish_transno() if the client was
5411 evicted, but without actually returning an error. Clients
5412 would later try to replay that open and may trip an assertion
5413 Simplify the client close codepath, and always return an error
5414 from the MDS in case the open is not successful.
5417 Frequency : rare, 2.6 OSTs only
5419 Description: OST may deadlock under high load on fragmented files
5420 Details : If there was a heavy load and highly-fragmented OST filesystems
5421 it was possible to have all the OST threads deadlock waiting on
5422 allocation of biovecs, because the biovecs were not released
5423 until the entire RPC IO was completed. Instead, release biovecs
5424 as soon as they are complete to ensure forward IO progress.
5426 Severity : enhancement
5428 Description: Support for specifying external journal device at mount
5429 Details : If an OST or MDS device is formatted with an external journal
5430 device, this device major/minor is stored in the ext3 superblock
5431 and may not be valid for failover. Allow detecting and
5432 specifying the external journal at mount time.
5437 Description: Mounting an MDS with pending unlinked files may cause oops
5438 Details : target_finish_recovery() calls mds_postrecov() which returned
5439 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5440 considers this an error and immediately begins cleaning up the
5441 lov, just after starting the mds_lov process
5443 Severity : enhancement
5445 Description: Implement 'lfs df' to report actual free space on per-OST basis
5446 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5447 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5448 usage of objects; '-h' to report in human readable format.
5450 ------------------------------------------------------------------------------
5452 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5459 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5460 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5461 at startup during orphan unlinking. Instead, skip buddy update
5462 before setup, as it will be regenerated after recovery is complete.
5467 Description: If an OST is inactive, its locks might reference stale inodes.
5468 Details : lov_change_cbdata() must iterate over all namespaces, even if
5469 they are inactive to clear inode references from the lock.
5471 Severity : enhancement
5472 Frequency : occasional, if non-standard max_dirty_mb used
5474 Description: Client will block write RPCs if not enough grant
5475 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5476 then the client will block writes while waiting for another RPC
5477 to complete instead of consuming its dirty limit. With change
5478 we get improved performance when max_dirty_mb is small.
5480 Severity : enhancement
5481 Bugzilla : 3389, 6253
5482 Description: Add support for supplementary groups on the MDS.
5483 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5484 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5485 lookups for user supplementary groups into a cache.
5489 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5490 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5491 for liblustre O_EXCL is enforced by MDS.
5495 Description: readdir never returns NULL in liblustre.
5496 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5497 dentry in struct dirent.
5501 Frequency : liblustre only, depends on application IO pattern
5502 Description: liblustre clients evicted if not contacting servers
5503 Details : Don't put liblustre clients into the ping_evictor list, so
5504 they will not be evicted by the pinger ever.
5506 Severity : enhancement
5508 Description: Add ability to evict clients by NID from MDS.
5509 Details : By echoing "nid:$NID" string into
5510 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5511 $NID would be instantly evicted from this MDS and from all active
5512 OSTs connected to it.
5516 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5517 Details : lookup_it_finish() used to query file size from OSTs that was not
5522 Description: service threads change working directory to that of init
5523 Details : Starting lustre service threads may pin the working directory
5524 of the parent thread, making that filesystem busy. Threads
5525 now change to the working directory of init to avoid this.
5529 Frequency : during shutdown only
5530 Description: shutdown with a failed MDS or OST can cause unmount to hang
5531 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5536 Frequency : 2.6 only
5537 Description: chmod/chown may include an extra supplementary group
5538 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5539 supplementary group and if none is specified this is used.
5542 Bugzilla : 5479 (6816)
5544 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5545 Details : If lookup is in progress on a file that is unlinked we might try
5546 to revalidate the inode and fail in revalidate after lookup is
5547 complete and ll_file_open() enqueues the open again but
5548 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5551 Frequency : always, if lconf --abort_recovery used
5553 Description: lconf --abort_recovery fails with 'Operation not supported'
5554 Details : lconf was attempting to abort recovery on the MDT device and not
5557 ------------------------------------------------------------------------------
5559 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5564 Frequency : rare (only unsupported configurations with a node running as an
5566 Bugzilla : 6514, 5137
5567 Description: Mounting a Lustre file system on a node running as an OST could
5569 Details : OSTs now preallocates memory needed to write out data at
5570 startup, instead of when needed, to avoid having to
5571 allocate memory in possibly low memory situations.
5572 Specifically, if the file system is mounted on on OST,
5573 memory pressure could force it to try to write out data,
5574 which it needed to allocate memory to do. Due to the low
5575 memory, it would be unable to do so and the node would
5576 become unresponsive.
5578 Severity : enhancement
5580 Description: Addition of lconf --service command line option
5581 Details : lconf now accepts a '--service <arg>' option, which is
5582 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5584 Severity : enhancement
5586 Description: Failover mode is now the default for OSTs.
5587 Details : By default, OSTs will now run in failover mode. To return to
5588 the old behaviour, add '--failout' to the lmc line for OSTs.
5590 Severity : enhancement
5592 Description: Health checks are now provided for MDS and OSTs
5593 Details : Additional detailed health check information on MSD and OSTs
5594 is now provided through the procfs health_check value.
5597 Frequency : occasional, depends on IO load
5599 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5600 after numerous create/delete cycles
5601 Details : The ext3 inode allocation policy would not allocate new inodes
5602 very well on the OSTs because there are no new directories
5603 being created. Instead we look for groups with free space if
5604 the parent directories are nearly full.
5609 Description: Network or server problems during mount may cause partially
5610 mounted clients instead of returning an error.
5611 Details : The config llog parsing code may overwrite the error return
5612 code during mount error handling, returning success instead
5618 Description: MDS can fail to allocate large reply buffers
5619 Details : After long uptimes the MDS can fail to allocate large reply
5620 buffers (e.g. zconf client mount config records) due to memory
5621 fragmentation or consumption by the buffer cache. Preallocate
5622 some large reply buffers so that these replies can be sent even
5623 under memory pressure.
5627 Frequency : rare (liblustre)
5628 Description: fsx running with liblustre complained that using truncate() to
5629 extend the file doesn't work. This patch corrects that issue.
5630 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5631 ATTR_SIZE and lsm use in llu_setattr_raw.
5635 Frequency : rare, only 2.6 kernels
5636 Description: Unusual file access patterns on the MDS may result in inode
5637 data being lost in very rare circumstances.
5638 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5639 code in the 2.6 kernel caused the inode and/or EA data not to
5640 be read from disk, causing single-file corruption.
5644 Frequency : rare, only 2.6 filesystems using extents
5645 Description: Heavy concurrent write and delete load may cause data corruption.
5646 Details : It was possible under high-load situations to have an extent
5647 metadata block in the block device cache from a just-unlinked
5648 file overwrite a newly-allocated data block. We now unmap any
5649 metadata buffers that alias just-allocated data blocks.
5653 Frequency : filesystems with default stripe_count larger than 77
5654 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5655 Details : lconf specifies an inode size of 4096 bytes when the default
5656 stripe_count is larger than 77. This conflicts with the default
5657 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5658 this case to avoid pinning too much memory for large EAs.
5660 ------------------------------------------------------------------------------
5662 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5667 Frequency : rare (extremely heavy IO load with hundreds of clients)
5669 Description: Client is evicted, gets IO error writing to file
5670 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5671 caused clients to be evicted for AST timeouts. The fixes in
5672 bug 5192 mean we no longer need to have such short AST timeouts
5673 so ldlm_timeout has been increased.
5676 Frequency : occasional during --force or --failover shutdown under load
5677 Bugzilla : 5949, 4834
5678 Description: Server oops/LBUG if stopped with --force or --failover under load
5679 Details : a collection of import/export refcount and cleanup ordering
5680 issues fixed for safer force cleanup
5683 Frequency : only filesystems larger than 120 OSTs
5684 Bugzilla : 5990, 6223
5685 Description: lfs getstripe would oops on a very large filesystem
5686 Details : lov_getconfig used kfree on vmalloc'd memory
5689 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5690 Bugzilla : 6242, 6243
5691 Description: reading from files that had been truncated to a non-zero size
5692 but never opened returned no data
5693 Details : ll_file_read() reads zeros from no-object files to EOF
5698 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5699 Details : There's little harm in aborting MDS/OSS recovery and letting it
5700 try again, so I removed the LASSERT and return an error instead.
5702 Severity : enhancement
5704 Description: New debugging infrastructure for tracking down data corruption
5705 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5706 (b) cover more of the client-side code path, and (c) try to narrow
5707 down where problems occurred
5711 Bugzilla : 3819, 4364, 4397, 6313
5712 Description: Racing close and eviction MDS could cause assertion in mds_close
5713 Details : It was possible to get multiple mfd references during close and
5714 client eviction, leading to one thread referencing a freed mfd.
5716 Severity: : enhancement
5717 Bugzilla : 3262, 6359
5718 Description: Attempts to reconnect to servers are now more aggressive.
5719 Details : This builds on the enhanced upcall-less recovery that was added
5720 in 1.4.2. When trying to reconnect to servers, clients will
5721 now try each server in the failover group every 10 seconds. By
5722 default, clients would previously try one server every 25 seconds.
5727 Description: After recovery, certain operations trigger a failed
5728 assertion on a client.
5729 Details : Failing over an mds, using lconf -d --failover, while a
5730 client was doing a readdir() call would cause the client to
5731 LBUG after recovery completed and the readdir() was resent.
5733 Severity : enhancement
5735 Description: Default groups are now added by lconf
5736 Details : You can now run lconf --group <servicename> without having to
5737 manually add groups with lmc.
5740 Frequency : occasional
5742 Description: Nodes with an elan id of 0 trigger a failed assertion
5745 Frequency : always when accessing e.g. tty/console device nodes
5747 Description: tty and some other devices nodes cannot be used on lustre
5748 Details : file's private_data field is used by device data and lustre
5749 values in there got lost. New field was added to struct file to
5750 store fs-specific private data.
5753 Frequency : when exporting Lustre via NFS
5755 Description: NFSD failed occasionally when looking up a path component
5756 Details : NFSD is looking up ".." which was broken in ext3 directories
5757 that had grown large enough to become hashed.
5760 Frequency : Clusters with multiple interfaces not on the same subnet
5762 Description: Nodes will repeatedly try to reconnect to an interface which it
5763 cannot reach and report an error to the log.
5764 Details : Extra peer list entries will be created by lconf with some peers
5765 unreachable. lconf now validates the peer before adding it.
5768 Frequency : Only if a default stripe is set on the filesystem root.
5770 Description: Setting a default stripe on the filesystem root prevented the
5771 filesystem from being remounted.
5772 Details : The client was sending extra request flags in the root getattr
5773 request and did not allocate a reply buffer for the dir EA.
5776 Frequency : occasional, higher if lots of files are accessed by one client
5777 Bugzilla : 6159, 6097
5778 Description: Client trips assertion regarding lsm mismatch/magic
5779 Details : While revalidating inodes the VFS looks up inodes with ifind()
5780 and in rare cases can find an inode that is being freed.
5781 The ll_test_inode() code will free the lsm during ifind()
5782 when it finds an existing inode and then the VFS later attaches
5783 this free lsm to a new inode.
5787 Bugzilla : 6422, 7030
5788 Description: MDS deadlock between mkdir and client eviction
5789 Details : Creating a new file via mkdir or mknod (starting a transaction
5790 and getting the ns lock) can deadlock with client eviction
5791 (gets ns lock and trying to finish a synchronous transaction).
5794 Frequency : occasional
5795 Description: While starting a server, the fsfilt_ext3 module could not be
5797 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5798 kernels. Previously, lconf would still use the ext3 name
5799 when trying to load modules. Now, it will correctly use
5800 ext3 on 2.4 and ldiskfs on 2.6.
5802 Severity : enhancement
5803 Description: The default stripe count has been changed to 1
5804 Details : The interpretation of the default stripe count (0, to lfs
5805 or lmc) has been changed to mean striping across a single
5806 OST, rather than all available. For general usage we have
5807 found a stripe count of 1 or 2 works best.
5809 Severity : enhancement
5810 Description: Add support for compiling against Cray portals.
5811 Details : Conditional compiling for some areas that are different
5815 Frequency : occasional
5816 Bugzilla : 6409, 6834
5817 Description: Creating files with an explicit stripe count may lead to
5818 a failed assertion on the MDS
5819 Details : If some OSTs are full or unavailable, creating files may
5820 trigger a failed assertion on the MDS. Now, Lustre will
5821 try to use other servers or return an error to the
5825 Frequency : occasional
5827 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5828 caused lock timeout during readahead (since 1.4.2).
5829 Details : Processes doing readahead might match a lock that hasn't been
5830 granted yet if there are overlapping and conflicting lock
5831 requests. The readahead process waits on ungranted lock
5832 (original lock is CBPENDING), while OST waits for that process
5833 to cancel CBPENDING read lock and eventually evicts client.
5835 Severity : enhancement
5837 Description: Initial enabling of flock support for clients
5838 Details : Implements fcntl advisory locking and file status functions.
5839 This feature is provided as an optional mount flag (default
5840 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5841 locking are implemented yet, and those that are are not guaranteed
5842 to be completely correct in production environments.
5843 mount -t lustre -o [flock|noflock] ...
5846 Frequency : occasional
5848 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5849 assertion in the ext3 JBD (journaling) layer.
5850 Details : The b_committed_data struct is protected by the big kernel lock
5851 in 2.4 kernels, serializing journal_commit_transaction() and
5852 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5853 access to this struct. In 2.6 kernels there is finer grained
5854 locking to improve SMP performance of the JBD layer.
5858 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5863 Description: merge and cleanup kernel patches.
5866 -----------------------------------------------------------------------------
5868 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5870 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5871 versions of Lustre. Please update all servers and clients to
5872 version 1.4.2 or later at the same time. You must also run
5873 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5874 to update the configuration logs.
5876 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5877 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5878 - swab llog records of type '0' so we get proper header size/idx (5861)
5879 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5880 - fix rename of one directory over another leaking an inode (5953)
5881 - avoid SetPageDirty on 2.6 (5981)
5882 - don't re-add just-being-destroyed locks to the waiting list (5653)
5883 - when creating new directories, inherit the parent's custom
5884 striping settings if present parent (3048)
5885 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5886 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5887 - don't hold client locks on temporary worklist from l_lru (5666)
5888 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5889 - automatically evict dead clients (5921)
5890 - Update file size properly in create+truncate+fstat case (6196)
5891 - Do not unhash mountpoint dentries, do not allow removal of
5893 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5894 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5895 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5896 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5897 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5898 - Correctly handle reads of files with no objects (6243)
5899 - lctl recover will also mark a device active if deactivate used (5933)
5901 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5902 - allow --write-conf on an MDS with different nettype than client (5619)
5903 - don't write config llogs to MDS for mounts not from that MDS (5617)
5904 - lconf should create multiple TCP connections from a client (5201)
5905 - init scripts are now turned off by default; run chkconfig --on
5906 lustre and chkconfig --on lustrefs to use them
5907 - upcalls are no longer needed for clients to recover to failover
5909 - add --abort-recovery option to lconf to abort recovery on device
5911 - add support for an arbitrary number of OSTs (3026)
5912 - Quota support protocol changes.
5913 - forward compatibility changes to wire structs (6007)
5914 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5915 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5916 - improved "lustre status" script
5917 - initialize blocksize for non-regular files (6062)
5918 - added --disable-server and --disable-client configure options (5782)
5919 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5920 - Vanilla 2.4.29 support
5921 - increase maximum number of obd devices to 520 (6242)
5922 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5923 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5924 SLES 9 2.6.5 kernels
5925 - limit stripes per file to 160 (the maximum EA size) (6093)
5927 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5930 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5931 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5932 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5933 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5934 - small ext3 extents cleanups and fixes (5733)
5935 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5936 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5937 - avoid needless client->OST connect, fix handle mismatch (5317)
5938 - fix DLM error path that led to out-of-sync client, long delays (5779)
5939 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5940 - fix several locking issues related to i_size (5492,5624,5654,5672)
5941 - don't move pending lock onto export if it is already evicted (5683)
5942 - fix kernel oops when creating .foo in unlinked directory (5548)
5943 - fix deadlock in obdfilter statistics vs. object create (5811)
5944 - use time_{before,after} to avoid timer jiffies wrap (5882)
5945 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5946 - Do not leak request if server was not able to process it (5154)
5947 - If mds_open unable to find parent dir, make that negative lookup(5154)
5948 - don't create new directories with extent-mapping (5909, 5936)
5950 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5951 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5952 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5953 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5954 - support for RHEL4 kernel on i686 (5773)
5955 - provide error messages when incompatible logs are encountered (5898)
5957 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5958 * version 1.4.0.10 (1.4.1 release candidate 1)
5960 - don't keep a lock reference when lock is not granted (4238)
5961 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5962 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5963 - drop import inflight refcount on signal_completed_replay error (5255)
5964 - unlock page after async write error during send (3677)
5965 - handle missing objects in filter_preprw_read properly (5265)
5966 - no transno return for symlink open, don't save no-trasno open (3440)
5967 - don't try to complete elan receive that already failed (4012)
5968 - free RPC server reply state on error (5406)
5969 - clean up thread from ptlrpc_start_thread() on error (5160)
5970 - readahead could read extra page into cache that wasn't ejected (5388)
5971 - prevent races in class_attach/setup/cleanup/detach (5260)
5972 - don't dereference de->d_inode after l_dput of de (5458)
5973 - use "int" for stripe value returned from lock_to_stripe (5544)
5974 - mballoc allocation and error-checking fixes in 2.6 (5504)
5975 - block device patches to fix I/O request sizes in 2.6 (5482)
5976 - look up hostnames for IB nals (5602)
5977 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5978 - don't start multiple acceptors for the same port (5277)
5979 - fix incorrect LASSERT in mds_getattr_name (5635)
5980 - export a proc file for general "ping" checking (5628)
5981 - fix "lfs check" to not block when the MDS is down (5628)
5983 - service request history (4965)
5984 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5985 - create an "evict_client" /proc entry on OSTs, like the MDS has
5986 - fix mount usage message, return errors per mount(8) (5168)
5987 - change grep [] to grep "[]" in tests so they work in more UMLs
5988 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5989 - remove ancient LOV_MAGIC_V0 EA support (5047)
5990 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5991 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5992 - allow readahead allocations to fail when low on memory (5383)
5993 - mmap locking landed again, after considerable improvement (2828)
5994 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5996 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5999 - send OST transaction number in read/write reply to free req (4966)
6000 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6001 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6002 - the watchdog thread now runs as interruptible (5246)
6003 - flock/lockf fixes (but it's still disabled, pending 5135)
6004 - don't use EXT3 constants in llite code (5094)
6005 - memory shortage at startup could cause assertion (5176)
6007 - reorganization of lov code
6008 - single portals codebase
6010 - add extents/mballoc support (5025)
6011 - direct I/O reads in the obdfilter (4048)
6012 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6014 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6017 - send OST transaction number in read/write reply to free req (4966)
6018 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6019 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6020 - the watchdog thread now runs as interruptible (5246)
6021 - handle missing objects in filter_preprw_read properly (5265)
6022 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6023 - drop import inflight refcount on signal_completed_replay error (5255)
6024 - unlock page after async write error during send (3677)
6025 - return original error code on reconstructed replies (3761)
6026 - no transno return for symlink open, don't save no-trasno open (3440)
6028 - add pid to ldlm debugging output (4922)
6029 - bump the watchdog timeouts -- we can't handle 30sec yet
6030 - extra debugging for orphan dentry/inode bug (5259)
6032 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6035 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6036 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6037 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6038 - fix NULL dereference in /proc/sys/portals/routes (4827)
6039 - allow failed mdc_close() operations to be interrupted (4561)
6040 - stop precreate on OST before MDS would time out on it (4778)
6041 - don't send partial-page writes before EOF from client (4410)
6042 - discard client grant for sub-page writes on large-page clients (4520)
6043 - don't free dentries not owned by NFS code, check generation (4806)
6044 - fix lsm leak if mds_create_objects() fails (4801)
6045 - limit debug_daemon file size, always print CERROR messages (4789)
6046 - use transno after validating reply (3892)
6047 - process timed out requests if import state changes (3754)
6048 - update mtime on OST during writes, return in glimpse (4829)
6049 - add mkfsoptions to LDAP (4679)
6050 - use ->max_readahead method instead of zapping global ra (5039)
6051 - don't interrupt __l_wait_event() during strace
6053 - add software watchdogs to catch hung threads quickly (4941)
6054 - make lustrefs init script start after nfs is mounted
6055 - fix CWARN/ERROR duplication (4930)
6056 - return async write errors to application if possible (2248)
6057 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6058 - print NAL number in %x format (4645)
6059 - update barely-supported suse-2.4.21-171 series (4842)
6060 - support for sles 9 %post scripts
6061 - support for building 2.6 kernel-source packages
6062 - support for sles km_* packages
6064 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6067 - ignore -ENOENT errors in osc_destroy (3639)
6068 - notify osc create thread that OSC is being cleaned up (4600)
6069 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6070 - reconstruct ost_handle() like mds_handle() (4657)
6071 - create a new thread to do import eviction to avoid deadlock (3969)
6072 - let lconf resolve symlinked-to devices (4629)
6073 - don't unlink "objects" from directory with default EA (4554)
6074 - hold socknal file ref over connect in case target is down (4394)
6075 - allow more than 32000 subdirectories in a single directory (3244)
6076 - fix blocks count for O_DIRECT writes (3751)
6077 - OST returns ENOSPC from object create when no space left (4539)
6078 - don't send truncate RPC if file size isn't changing (4410)
6079 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6080 - bind to privileged port in socknal and tcpnal (3689)
6082 - rate limit CERROR/CWARN console message to avoid overload (4519)
6083 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6084 - basic mmap support (3918)
6085 - kernel patch series update from b1_4 (4711)
6087 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6090 - avoid crash during MDS cleanup with OST shut down (2775)
6091 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6092 - don't use bad inodes on the MDS (3744)
6093 - dynamic object preallocation to improve recovery speed (4236)
6094 - don't hold spinlock over lock dumping or change debug flags (4401)
6095 - don't zero obd_dev when it is force cleaned (3651)
6096 - print grants to console if they go negative (4431)
6097 - "lctl deactivate" will stop automatic recovery attempts (3406)
6098 - look for existing locks in ldlm_handle_enqueue() (3764)
6099 - don't resolve lock handle twice in recovery avoiding race (4401)
6100 - revalidate should check working dir is a directory (4134)
6102 - don't always mark "slow" obdfilter messages as errors (4418)
6104 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6107 - don't close LustreDB during write_conf until it is done (3860)
6108 - fix typo in lconf for_each_profile (3821)
6109 - allow dumping logs from multiple threads at one time (3820)
6110 - don't allow multiple threads in OSC recovery (3812)
6111 - fix debug_size parameters (3864)
6112 - fix mds_postrecov to initialize import for llog ctxt (3121)
6113 - replace config semaphore with spinlock (3306)
6114 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6115 - don't allow enqueue to complete on a destroyed export (3822)
6116 - down write_lock before checking llog header bitmap (3825)
6117 - recover from lock replay timeout (3764)
6118 - up llog sem before sending rpc (3652)
6119 - reduce ns lock hold times when setting kms (3267)
6120 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6121 - fix NULL deref and obd_dev leak on setup error (3312)
6122 - replace some LBUG about llog ops with error handling (3841)
6123 - don't match INVALID dentries from d_lookup and spin (3784)
6124 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6125 - fix invalid assertion in ptlrpc_set_wait (3880)
6127 - add libwrap support for the TCP acceptor (3996)
6128 - add /proc/sys/portals/routes for non-root route listing (3994)
6129 - allow setting MDS UUID in .xml (2580)
6130 - print the stack of a process that LBUGs (4228)
6132 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6135 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6136 - make sure to unset replay flag from failed open requests (3440)
6137 - if default stripe count is 0, use OST count for inode size (3636)
6138 - update parent mtime/ctime on client for create/unlink (2611)
6139 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6140 - free recovery state on server during a forced cleanup (3571)
6141 - unregister_reply for resent reqs (3063)
6142 - loop back devices mounting and status check on 2.6 (3563)
6143 - fix resource-creation race that can provoke i_size == 0 (3513)
6144 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6145 - more debugging for page-accounting assertion (3746)
6146 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6147 - avoid deadlock after precreation failure (3758)
6148 - fix race and lock order deadlock in orphan handling (3450, 3750)
6149 - add validity checks when grabbing inodes from l_ast_data (3599)
6151 - add /proc/.../recovery_status to obdfilter (3428)
6152 - lightweight CDEBUG infrastructure, debug daemon (3668)
6153 - change default OSC RPC parameters to be better on small clusters
6154 - turn off OST read cache for files smaller than 32MB
6155 - install man pages and include them in rpms (3100)
6156 - add new init script for (un)mounting lustre filesystems (2593)
6157 - run chkconfig in %post for init scripts (3701)
6158 - drop scimac NAL (unmaintained)
6160 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6163 - clean kiobufs before and after use (3485)
6164 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6165 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6166 - change init script to fail more clearly if not run as root (1528)
6167 - allow clients to reconnect during replay (1742)
6168 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6169 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6170 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6171 - fix cond_resched (3554)
6172 - only evict unfinished clients after recovery (3515)
6173 - allow bulk resend, prevent data loss (3570)
6174 - dynamic ptlrpc request buffer allocation (2102)
6175 - don't allow unlinking open directory if it isn't empty (2904)
6176 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6177 - remove extraneous obd dereference causing LASSERT failure (3334)
6178 - don't use get_cycles() when creating temp. files on the mds (3156)
6179 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6180 - handle EEXIST for set-stripe, set proper directory name (3336)
6182 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6183 - fix ksocknal_fmb_callback() error messages (2918)
6185 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6188 - don't copy lvb into (possibly NULL) reply on error (2983)
6189 - don't deref dentry after dput, don't free lvb on error (2922)
6190 - use the kms to determine writeback rpc length (2947)
6191 - increment oti_logcookies when osc is inactive (2948)
6192 - update client's i_blocks count via lvb messages (2543)
6193 - handle intent open/close of special files properly (1557)
6194 - mount MDS with errors=remount-ro, like obdfilter (2009)
6195 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6196 - don't use cancelling-locks' kms values (2947)
6197 - use highest lock extent for kms, not last one (2925)
6198 - don't dereference ERR_PTR() dentry in error handling path (3107)
6199 - fix thread race in portals_debug_dumplog() (3122)
6200 - create lprocfs device entries at setup instead of at attach (1519)
6201 - common AST error handler, don't evict client on completion race (3145)
6202 - zero nameidata in detach_mnt in 2.6 (3118)
6203 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6204 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6205 - initialize RPC timeout timer earlier for 2.6 (3219)
6206 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6207 - print nal/nid for unknown nid (3258)
6208 - additional checks for oscc recovery before doing precreate (3284)
6209 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6210 - don't crash in mdc_close for bad permissions on open (3285)
6211 - zero i_rdev for non-device files (3147)
6212 - clear page->private before handing to FS, better assertion (3119)
6213 - tune the read pipeline (3236)
6214 - fix incorrect decref of invalidated dentry (2350)
6215 - provide read-ahead stats and refine rpc in flight stats (3328)
6216 - don't hold journal transaction open across create RPC (3313)
6217 - update atime on MDS at close time (3265)
6218 - close LDAP connection when recovering to avoid server load (3315)
6219 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6220 - don't leak open file on MDS after open resend (3325)
6221 - serialize filter_precreate and filter_destroy_precreated (3329)
6222 - loop device shouldn't call sync_dev() for nul device (3092)
6223 - clear page cache after eviction (2766)
6224 - resynchronize MDS->OST in background (2824)
6225 - refuse to mount the same filesystem twice on same mountpoint (3394)
6226 - allow llmount to create routes for mounting behind routers (3320)
6227 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6228 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6229 - fix rare problem with rename on htree directories (3417)
6231 - allow default OST striping configuration per directory (1414)
6232 - fix compilation for qswnal for 2.6 kernels (3125)
6233 - increase maximum number of MDS request buffers for large systems
6234 - change liblustreapi to be useful for external progs like lfsck (3098)
6235 - increase local configuration timeout for slow disks (3353)
6236 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6238 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6241 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6242 - don't overwrite extent policy data in reply if lock was blocked (2901)
6243 - drop filter export grants atomically with removal from device (2663)
6244 - del obd_self_export from work_list in class_disconnect_exports (2908)
6245 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6246 - swab reply message in mdc_close, other PPC fixes (2464)
6247 - fix destroying of named logs (2325)
6248 - overwrite old logs when running lconf --write_conf (2264)
6249 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6250 - fix race in target_handle_connect (2898)
6251 - mds_reint_create() should take same inode create lock (2926)
6252 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6253 - don't close files for self_export to avoid uninitialized obd (2936)
6254 - allow MDS with the same name as client node (2939)
6255 - hold dentry reference for closed log files for unlink (2325)
6256 - reserve space for all logs during transactions (2059)
6257 - don't evict page beyond end of stripe extent (2925)
6258 - don't oops on a deleted current working directory (2399)
6259 - handle hard links to targets without a parent properly (2517)
6260 - don't dereference NULL lock when racing during eviction (2867)
6261 - don't grow lock extents when lots of conflicting locks (2919)
6263 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6266 - account for cache space usage on clients to avoid data loss (974)
6267 - lfsck support in lustre kernel code (2349)
6268 - reduce journal credits needed for BRW writes (2370)
6269 - orphan handling to avoid losing space on client/server crashes
6270 - ptlrpcd can be blocked, stopping ALL progress (2477)
6271 - use lock value blocks to assist in proper KMS, faster stat (1021)
6272 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6273 - recovery for initial connections (2355)
6274 - fixes for mds_cleanup_orphans (1934)
6275 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6276 - block all file creations until orphan recovery completes (1901)
6277 - client remove rq_connection from request struct (2423)
6278 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6279 - recovery timer race (2670)
6280 - mdc_close recovey bug (2532)
6281 - ptlrpc cleanup bug (2710)
6282 - mds timeout on local locks (2588)
6283 - namespace lock held during RPCs (2431)
6284 - handle interrupted sync write properly (2503)
6285 - don't try to handle a message that hasn't been replied to (2699)
6286 - client assert failure during cleanup after abort recovery (2701)
6287 - leak mdc device after failed mount (2712)
6288 - ptlrpc_check_set allows timedout requests to complete (2714)
6289 - wait for inflight reqs when ptlrpcd finishes (2710)
6290 - make sure unregistered services are removed from the srv_list
6291 - reset bulk XID's when resending them (caught by 1138 test)
6292 - unregister_bulk after timeout
6293 - fix lconf error (2694)
6294 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6295 - readahead locks pages, leaves pending causing memory pressure (2673)
6296 - increase OST request buffers to 4096 on large machines (2729)
6297 - fix up permission of existing directories in simple_mkdir (2661)
6298 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6299 - don't assign transno to errored transactions (2742)
6300 - don't delete objects on OST if given a bogus objid from MDS (2751)
6301 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6302 - if rq_no_resend, then timeout request after recovery (2432)
6303 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6304 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6305 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6306 - MDS recovery shouldn't delete objects that it has given out (2730)
6307 - if enqueue arrives after completion, don't clobber LVB (2819)
6308 - don't unlock pages twice when trigger_group_io returns error (2814)
6309 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6310 - don't write pages to disk if there was an error (1450)
6311 - don't ping imports that have recovery disabled (2676)
6312 - take buffered bytes into account when balancing socknal conn (2817)
6313 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6314 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6315 - remove little-endian swabbing of llog records (1987)
6316 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6317 - retry reposting request buffers if they fail (1191)
6318 - grow extent at grant time to avoid granting a revoked lock (2809)
6319 - lock revoke doesn't evict page if covered by a second lock (2765)
6320 - disable VM readahead to avoid reading outside lock extents (2805)
6322 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6323 - updated kernel patches for hp-2.4.20 kernel (2681)
6325 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6328 - fix truncated write corruption (2366)
6329 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6331 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6332 - ensure len(uuid) < 37 in lmc (1171)
6333 - fix ia64 OOPS in llog_test (2255)
6334 - zero end of page at obdfilter for partial page writes (2648)
6335 - don't leave stale dentries around after renames (bug 2428)
6336 - fix timeouts when evicting a client with a single lock held (2642)
6337 - set deadline for the initial HELLO message to drain (2634)
6338 - print out dotted-quad IP addresses in the socknal (2302)
6340 - additional debugging for MDS client eviction problem (2443)
6341 - fix mkfsoptions support for osts (2603, 2604)
6343 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6346 - add series for the vanilla 2.6.0 kernel
6347 - add series for the vanilla 2.4.24 kernel
6348 - add series for a cray x86/64 UL kernel drop
6349 - fix xattr patches for the vanilla 2.4.19 series
6351 - generate true UUIDs in lmc (1171)
6352 - have portals stack dumping break in UML (2466)
6353 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6354 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6355 - wake callback waiting threads on client eviction (2460)
6356 - Add --ptldebug and --subsystem to lmc (1719)
6357 - update assertion to allow safe interrupt allocation
6358 - set rq_no_resend for cancel requests (2432)
6359 - recalculate ptlrpcd timeout after resend (2494)
6360 - call vfs_rmdir when removing pending directories (2368)
6361 - fix renaming a file to itself (2429)
6362 - lmc creates a default one-stripe lov (2454)
6363 - expand procfs space to handle large clusters (2326)
6364 - increase UML stack to avoid overflow
6365 - update lconf's list of debug and subsystem masks
6366 - fix lfs find --obd (2510)
6367 - /proc tunable for disabling filter read caching (2591)
6368 - stop rpm packages from altering slapd.conf (2301)
6369 - disable nagle in the socknal under 0conf (2578)
6370 - choose mds inode size based on stripe count (2572)
6371 - fix kernel-source rpm problems (2516)
6373 - add --disable-doc to avoid pdf generation (2421)
6374 - update documentation, tests, type-os, comments
6375 - avoid format warnings on ia64
6376 - remove the TOE NAL
6377 - tiny code cleanups by removing unused fields
6379 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6382 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6383 - remove the most blatant lies from BUILDING file (2371)
6384 - change default debug level to reasonable production setting
6385 - reduce client side cache size to reduce cache flush time
6386 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6387 - make TCP zerocopy and pinger support enabled by default (2476)
6388 - sync writes completed after process exits caused crashes (2319)
6389 - maintain correct mount count on the MDS (2356)
6390 - backout 1557, because 2316 wasn't really fixed
6391 - better file I/O statistics gathering in /proc
6392 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6393 - another kernel patch to fix zero-copy TCP function export
6394 - don't take duplicate lock when processing re-sent getattr (2420)
6395 - lctl uses obd_self_export instead of creating new conn (2353)
6396 - MDS/OST recovery case which requires object creation asserted (2425)
6397 - move lfs from /usr/sbin to /usr/bin in packages
6398 - fix race between mds_client_add and mds_client_free (2417)
6399 - use kmalloc instead of slabs in portals (2430)
6400 - don't create duplicate records when a failover MDS is present (2442)
6401 - remove unnecessary mount age check (2332)
6402 - don't remove directory inodes from locks prematurely (2451)
6403 - don't break if MDS service name is the same as hostname (2103)
6404 - fix races in client write RPC generation when cache full (2482)
6406 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6409 - remove now-unused request->rq_obd (278)
6410 - if an allocation fails, print out how much memory we've used (1933)
6411 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6412 - add the "configurable stack size" patch to most series files (1256)
6413 - ability to write large log records, for 100+ OST configs (2306)
6414 - fix NULL deref when filter_prep fails (2314)
6415 - fix operator precedence error in filter_sync
6416 - dynamic allocation of socknal TX descriptors (2315)
6417 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6418 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6419 - ensure that CWARN messages in Portals always get to the syslog
6420 - __init/__exit are not for prototype decls (ldlm_init/exit)
6421 - x86-64 compile warning fixes
6422 - fix gateway LMC keyword conflict (2318)
6423 - fix MDS lock inversions in getattr/reint paths (1844)
6424 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6425 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6426 - DLM race condition prevented some lock evictions (2328)
6427 - ENOMEM detection and retry on socknal sends (2230)
6428 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6429 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6430 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6431 - better error messages when a client is rejected during recovery (1505)
6432 - avoid cancelling locks which were never granted, after failure (2330)
6433 - fix i_sem/journal inversion in mds_client_add (2333)
6434 - fix truncate/getattr lock cycle deadlock (2334)
6435 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6436 - fix two rare exit paths which could leak an l_lock() ref (2321)
6437 - fencepost error in MDS/OST orphan recovery (2226)
6438 - make log record alignment 8 bytes (1988)
6439 - lstripe now fails when requested offset > ost_count (2237)
6440 - ensure that all kernel series have a complete list.h (1607)
6441 - fix crashes in special-file operations (2316)
6442 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6444 - allow configurable automake binary, for testing new versions
6445 - small update to the lfs documentation
6447 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6449 * fix negative export reference count in fsfilt_sync (2312)
6451 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6452 * release candidate 0.9.1
6454 - orphans are moved into the PENDING directory for possible recovery
6455 - replayed opens now open by fid for orphan/rename safety (1042)
6456 - last close of an orphan inode generates a transno (683)
6457 - chdir() and mount() now pin the directory entry (1020)
6458 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6459 - discard very old requests without processing them (1502)
6460 - remove some common, well-understood CERRORs (1505)
6461 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6462 - clear "grant" flags in OST replies until OST grant code lands (1644)
6463 - fix read performance by not clobbering i_blksize on client (1598)
6464 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6465 - make LRU size a /proc tunable, clears locks when reduced (707)
6466 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6467 - server locks take references on exports now (1558)
6468 - build fixes for 2.4.20-rh trees (1663)
6469 - return an error from lov_create if all OSCs are inactive (1751)
6470 - fix import levels when a reconnect happens without a timeout (1597)
6471 - exit early from mds_open if we get a lookup error (1749)
6472 - partial page read at EOF wouldn't wait for disk before sending (1642)
6473 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6474 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6475 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6476 - write/truncate lock inversion (1639)
6477 - Don't auto-load obdclass, portals modules during cleanup (1495)
6478 - fix timestamps from jumping to "now" (1763)
6479 - extra journal assertions (1648)
6480 - add an extra multiunlink test (1771)
6481 - fix read_record/write_record API (1776)
6482 - fix leak of offset_extent, possible incorrect i_size later (1772)
6483 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6484 - Debugging for the kqswnal_get_idle_tx problems (1820)
6485 - Allow recovery to be attempted multiple times (1536)
6486 - Write out MDS last_rcvd file after it is first created (1600)
6487 - Fix tx_descriptor leak in failed transmit situations (1827)
6488 - ext3 journaling fixes for assertion failure after IO error (1871)
6489 - class_export_put() on freed export after completion AST error (1896)
6490 - Fix revalidate looping in VFS (1322)
6491 - Don't access a freed export during MDS_REINT timeout (1521)
6492 - Add open-unlink recovery support on the MDS (1673,1764)
6493 - Return an error if no MDS data was read from last_rcvd (1946)
6494 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6495 - Don't setup a disk device that doesn't match exported UUID (317)
6496 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6497 - avoid committing NULL handle in force close
6498 - local.sh is now a one-stripe LOV configuration
6499 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6500 - fix ext3 htree duplicate directory entry corruption (1516)
6501 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6502 - update to new LOV EA format (2097)
6503 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6505 2003-06-15 Phil Schwan <phil@clusterfs.com>
6508 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6509 - per-import/export recovery handling (958, 931, 959)
6510 - multiple last-rcvd slots, for serving multiple FSes (949)
6511 - connections are again shared between multiple imp/exports (963, 964)
6512 - "umount -f" would hang if any requests needed to be sent (393, 978)
6513 - avoid pinning large req buffer by copying for queued messages (989)
6514 - add "uuid" to "lctl device" command to help upcalls (991)
6515 - "open" RPCs with transnos would confuse recovery counters (1037)
6516 - do proper endian conversion of all wire messages (288, 340, 891)
6517 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6518 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6519 - fix ptlrpc_connection leak in target_handle_connect (1174)
6520 - fix import refcounting bug in OST and MDS cleanup (1134)
6521 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6522 - fix very unlikely obd_types race condition (501)
6523 - remove osc_open hack for echo_client (1187)
6524 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6525 - a failure in read_inode2 leads to deadlock (1139)
6526 - cancel ack-locks as soon as transaction is committed (1072)
6527 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6528 - make sure to commitrw after any preprw to avoid deadlock (1162)
6529 - failing to execute a file in a lustre FS would lock inode (1203)
6530 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6531 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6532 - obd_uuid2tgt would walk off the end of the list (1255)
6533 - on IA64 the osc would give portals incorrect bulk size (1258)
6534 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6535 - fix lock inversion caused by new llite matching code (1282)
6536 - limit the number of dirty pages on a client to 10MB (1286)
6537 - timed out locks were not being corrected cancelled (1289)
6538 - fix O_DIRECT above 4GB on IA-32 (1292)
6539 * major user-visible changes
6540 - fail out/fail over policy now controlled by the upcall (993)
6542 - add OBD_PING to check server availability and failure (954)
6543 - lustre messages are now sent in sending host order (288, 340, 891)
6544 - add eadatalen to MDS getattr reply (340)
6545 - OST read replies may contain second buffer, with per-page status (593)
6547 2003-03-11 Phil Schwan <phil@clusterfs.com>
6550 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6551 - failed open()s could cause deadlock; fixed (867, 869)
6552 - stop cancelling OST locks when files are closed (481)
6553 - overlapping XID spaces caused network corruption (851, 853)
6554 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6555 - setattr_raw added, to do single-RPC, server-side setattrs
6556 - lmc/lconf syntax change for OST UUIDs
6557 - fix crashy race condition between ptlrpc_free_req and osc_close
6558 - don't use request in mdc_enqueue if we hit a timeout (889)
6559 - don't set the inode i_size for regular files from the MDS (896)
6560 - handle out of order completion AST (842)
6561 - don't LBUG if a lock request times out after receiving AST (913)
6562 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6563 - if a bad lock AST arrives, send an error instead of dropping entirely
6564 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6565 - fix leak in bulk IO when only partially completed (899, 900, 926)
6566 - fix O_DIRECT for ia64 (55)
6567 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6568 - C-z after timeout could hang a process forever; fixed (977)
6570 - client-side I/O cache (678, 924, 929, 941, 970)
6572 - READPAGE and SETATTRs which don't take server-side locks get
6575 2003-02-11 Phil Schwan <phil@clusterfs.com>
6578 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6579 - Fix fsfilt_extN_readpage() to read a full page of directory
6580 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6581 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6582 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6583 - Most intent operations were removed, in favour of a new RPC mode
6584 that does a single RPC to the server and bypasses most of the VFS
6585 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6586 - Aggressively cancel local locks on DLM servers
6587 - mds_reint_unlink sends EA to the client if it's the last nlink.
6588 client uses that EA to unlink OST objects.
6589 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6590 - recursive symlinks were fixed (439)
6591 - fixed NULL deref in DEBUG_REQ
6592 - filter_update_lastobjid no longer calls sync, which annoyed extN
6593 - fixed multi-client small-writes to a single file problem (445)
6594 - fixed mtime updates during file writes (607)
6595 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6596 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6597 - fixed Linux OST size reporting problem (444, 656)
6598 - OST now updates object mtime with writes or setattr (607, 619)
6599 - client verifies file size before zeroing page past EOF (445)
6600 - OST now writes last allocated objid to disk with allocation (108)
6601 - LOV on echo now works (409)
6603 - mds_reint_unlink sends a new buffer, with the EA included. this
6604 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6605 set if a regular file was being unlinked, and it was the last link
6606 - use PtlGet from the target for bulk writes (315)
6607 - OST now updates object mtime with writes or setattr (607, 619)
6608 - LDLM now has a grant-time callback to revalidate locked items, if
6610 - Many MDS operations were reorganized to combat race conditions
6612 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6613 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6615 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6618 - Fully reactivate OST imports after reconnection (512, others)
6619 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6620 - More graceful error handling for truncating on dead OST (515)
6621 - Don't error out unless we're actually accessing dead stripes (474)
6622 - Fix garbage sizes when stripes are missing (410)
6623 - LRU counters were broken, causing constant lock purge (433, 432)
6624 - garbage on read from stripes with failed OSTs (441)
6625 - mark OSCs as active before reconnecting during recovery (438)
6626 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6627 - lfind did not preserve OST order in output (443)
6628 - symlinks cause hung clients, incorrect data (439)
6629 - stop dereferencing request after dropping refcount (457)
6630 - don't LASSERT(spin_is_locked) on non-SMP (455)
6631 - fixes for many rename() bugs
6632 - fstat didn't correctly synchronize attributes (399)
6633 - server must handle lock cancellation during blocking AST prep (487)
6634 - bulk descriptors were free()d too soon (511)
6635 - fix paths in lconf, which would load incorrect modules (451, 507)
6636 - fix confusing lconf 'host not found' error message (386)
6637 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6638 - fix race condition in mdc_blocking_ast() for inode access (526)
6639 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6640 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6641 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6642 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6643 - fix lstripe in several ways (kernel side) (527)
6644 - fix request leak in ldlm_cli_enqueue (262)
6645 - incorrect OSC was marked inactive after OST failure
6646 - call mds_fs_cleanup before unmounting filesystem (524)
6647 - fix races between taking ns_lock and ldlm_lock_change_resource
6648 - fix races updating LOV export open file list
6649 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6650 - fix recovery NULL deref in ldlm_cli_cancel_unused
6651 - fix some DLM races by using new hash table for lock handles (419)
6652 - permit the client to specify desired inodes, at replay
6653 - duplicate requests when we queue them for replay reintegration
6654 - fix last_rcvd offset calculation
6655 - sync after each recovered transaction, so we always make progress
6656 - never, not always, ERESTART requests without transnos
6657 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6659 - skip replay if the MDS says that the client is already connected
6660 - don't check for a recovery-enabled export to match lctl's UUID
6661 - don't INC_USE_COUNT for phantom exports
6662 - don't crash when cleaning up phantom exports (567)
6663 - don't double-finish or set replay data for errored mdc_open requests
6664 - abort requests when they time out, so we don't get old replies
6665 - send/receive replies for AST messages again
6666 - if the client says that it doesn't have the lock, cancel it on the
6668 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6669 mark it as destroyed, it will all work out when decref is called
6670 - fix module use counts (22, 581)
6672 - ASTs now expect a reply (server cancels lock on error reply)
6674 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6677 - fix many simultaneous client startup (392)
6678 - fix dentry->d_it clobbering
6679 - credentials weren't being shipped for readdir/getattr operations
6680 - remove invalid assertions triggered during some concurrent MD
6682 - proper Lustre versions added (336, 389)
6683 - fix memory leak for create error case (398)
6684 - fix LOV locking bug that would get cli/srv out of sync
6685 - fix echo client over LOV (409)
6686 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6687 - fix double-O_EXCL intent crash (424)
6688 - avoid sending multiple lock CANCELs (352)
6690 - MDS can do multi-client recovery (modulo bugs in new code)
6692 - many updates, edits, cleanups
6694 2002-11-18 Phil Schwan <phil@clusterfs.com>
6697 - fix null d_it dereference (346)
6698 - fix full OST/dbench hang (333)
6699 - fix permission problem with file removal (286)
6700 - fix removal of OSCs from LOV when they fail
6701 - fix NULL deref during bulk timeout (214)
6702 - fix problems related to multiple filesystems on one MDS (241)
6703 - fixed serious subtle metadata locking bugs
6704 - free locks on clients when inodes are removed due to memory
6706 - fix inode pointer in lock data (285)
6707 - partial support for multiple MDS on a single host (241)
6708 - data locks weren't cancelled at clear_inode time (290, 311)
6709 - intent locks could lead to unbounded lock growth (205)
6710 - added a maximum lock count, an LRU list, and a flusher
6711 - fix multiple rename (365)
6712 - properly abstracted the echo client
6713 - OSC locked 1 byte too many; fixed
6714 - rewrote brw callback code:
6715 - fixed recovery bugs related to LOVs (306)
6716 - fixed too-many-pages-in-one-write crash (191)
6717 - fixed (again) crash in sync_io_timeout (214)
6718 - probably fixed callback-related race (385)
6720 - Add capability to MDS protocol
6721 - LDLM cancellations and callbacks on different portals
6723 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6726 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6727 - defer open object destruction to close time (601981, 138)
6728 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6729 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6730 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6731 - don't call obd_setattr in ll_file_release for destroyed objects
6732 * protocol change to lustre_msg: move |version| and add |flags|
6733 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6734 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6735 * added replay of create, unlink, link and rename operations during
6736 MDS failover; recovery should be much more robust now
6737 * remove failed OSCs from LOVs (only lov_create uses this so far)
6738 * the lustre-HOWTO was brought (more) up to date (582544)
6740 2002-10-23 Phil Schwan <phil@clusterfs.com>
6743 - in-use dentries weren't being reused properly (617851)
6744 - prevent multiple LDLM setup (599178)
6745 - fix LOV size calculations for truncate (617853)
6746 - fix client handling of MDS intent errors (POSIX)
6747 - fix permission bug in lovstripe.c test (624321)
6748 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6749 - truncate past end of file could corrupt data
6750 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6751 - a race in recovery could return ETIMEDOUT to apps (623947)
6752 - building outside the source directory was fixed
6753 * the lustre-HOWTO was brought (more) up to date (582544)
6754 * major progress was made on recovery functionality
6756 2002-10-10 Phil Schwan <phil@clusterfs.com>
6759 - recovery deadlock fix
6760 - rm -rf causes LBUG fix (617817)
6761 - file open by multiple tasks fix (618962)
6762 - directory permissions bugs (602707 and 620007)
6763 - journal_stop fixed with locking (611313)
6764 - O_APPEND failures resolved (618273, perhaps 614459)
6765 - lconf PATH fix (619770)
6766 - IA64 build fix (621450)
6767 - RPC buffer sizes scale with amount of memory
6769 2002-10-01 Phil Schwan <phil@clusterfs.com>
6772 - locks would be cancelled without throwing away data pages,
6773 resulting in inconsistent data (605627)
6774 - inode attributes were not always being refreshed (605627, 612449)
6775 - lconf now continues to cleanup after lctl reports an error
6776 - MDS now enforces user permissions (602707)
6777 - lprocfs cleanup fixed, but not yet enabled (614157)
6778 - fixed infinite server hang, should a client not respond to an AST
6779 - avoid going into recovery if user calls readlink() with a buffer
6780 that's too small (613941)
6781 - AST RPCs no longer require replies (614867) -- this may be changed
6782 - don't crash server if client sends an IOV that's too big (611336)
6783 - fixed lock conversion deadlock (611892)
6784 - fixed the following of symlinks (614622)
6785 * recovery: the server can remove locks from a client that dies, other
6786 clients can make progress
6787 * more extN patch fixes
6788 * compile-time configurable ptlrpc buffer allocations
6790 - collaborative read cache document
6791 - Lustre Lite Performance CDR document-in-progress
6793 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6796 - fix typo in patch-2.4.18
6798 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6801 - clear ptlrpc request each time in handle_incoming_request()
6802 - unlink of files now destroys the object on the OST
6804 2002-09-19 Peter Braam <braam@clusterfs.com>
6806 * add hard link support
6807 * change obdfile creation method
6808 * kernel patch changed
6810 2002-09-19 Peter Braam <braam@clusterfs.com>
6813 - stack overflow bug in extN fixed
6815 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6817 * documentation updates
6818 - add man pages for config tools
6819 - update tests/README to describe testing with new config tools
6820 - finish metadata API descriptions
6821 * bug fixes and cleanups
6822 - statfs workaround for 16TB limit
6823 - LOV stripe allocation improved, can stripe on subset of OSTs
6824 - LOV file size/IO offset was wrong for files > 4GB in size
6825 - object EA data was being dropped, caused files to be unreadable
6826 - memory overflow with non-LOV OST caused memory corruption
6827 - fixed regression tests to work with new config tools, obdfilter
6828 - fixed bug when directory size became larger than 1 block
6829 - fixed bug (for single client case) when PWD was deleted
6830 - invalidate local directory pages when doing intent-based ops
6831 - avoid LDLM oops when lock callback contained bad data
6833 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6835 * documentation updates
6836 * bug fixes and cleanups
6837 - configuration tools
6840 - 64-bit compile warnings
6841 - 64-bit internal statfs data
6843 * test_brw on persistent OST devices
6845 * lprocfs (disabled)
6847 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6849 * documentation updates
6850 * bug fixes and cleanups
6851 * configuration tools
6853 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6856 * many small fixes to 0_5_4
6857 * io/network handling
6858 * thinkos in MDS operations
6860 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6863 * crucial basic fixes to 0.5.3
6864 * IOR, Iozone work over Elan
6867 2002-08-07 Phil Schwan <phil@clusterfs.com>
6868 * version 0_5_3, our first alpha
6869 * we use the new Portals iovs
6870 * documentation updates
6871 * bug fixes and cleanups
6872 * small changes in the DLM wire protocol
6874 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6875 * version 0_5_1 with some initial stability,
6876 * locking on MD and file I/O.
6877 * documentation updates
6878 * several bug fixes since 0.5.0
6879 * small changes in wire protocol
6881 2002-07-18 Phil Schwan <phil@clusterfs.com>
6883 * delivered as Lustre Light Alpha
6884 * fixed a crash after handling invalid MDS requests
6885 * fixed directory pages for architectures with non-4k pages sizes
6887 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6888 * release version v0_4_4
6889 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6891 * Fixes a number of interruption problems with OST operations.
6892 * Update documentation for portals header changes
6893 * Move all wire protocol structs/defines to lustre_idl.h
6894 * Fixes symlink length bug.
6895 * Add tcpdump to repository.
6897 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6898 * release version v0_4_3
6899 * Fixes statfs for inodes on extN.
6900 * Fixes bug in runtests which would delete /etc/hosts.
6901 * Use 64-bit object IDs wherever possible (not into VFS though)
6902 Remove ost_get_info, which is unused by lustre, and out of date.
6904 2002-07-03 Peter Braam <braam@clusterfs.com>
6905 * release version v0_4_2 Fixes a lookup error (type not passed)
6906 * move forward to head of Portals
6907 * move forward to latest Lustre kernel
6909 2002-06-25 Peter Braam <braam@clusterfs.com>
6910 * release version v0_4_1. Hopefully stable on single node use.