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 5.14.0-362.18.1.el9 (RHEL9.3)
7 5.14.0-284.30.1.el9 (RHEL9.2)
8 5.14.0-162.23.1.el9 (RHEL9.1)
9 4.18.0-513.18.1.el8 (RHEL8.9)
10 4.18.0-477.27.1.el8 (RHEL8.8)
11 * Other server kernels known to build and work at some point (others may also work):
12 3.10.0-862.14.4.el7 (RHEL7.5)
13 3.10.0-957.27.2.el7 (RHEL7.6)
14 3.10.0-1062.18.1.el7 (RHEL7.7)
15 3.10.0-1127.19.1.el7 (RHEL7.8)
16 3.10.0-1160.88.1.el7 (RHEL7.9)
17 4.18.0-80.11.2.el8 (RHEL8.0)
18 4.18.0-147.8.1.el8 (RHEL8.1)
19 4.18.0-193.28.1.el8 (RHEL8.2)
20 4.18.0-240.22.1.el8 (RHEL8.3)
21 4.18.0-305.25.1.el8 (RHEL8.4)
22 4.18.0-348.23.1.el8 (RHEL8.5)
23 4.18.0-372.32.1.el8 (RHEL8.6)
24 4.18.0-425.10.1.el8 (RHEL8.7)
25 4.4.120-92.70 (SLES12 SP2)
26 4.4.180-94.100 (SLES12 SP3)
27 4.4.0-131 (Ubuntu 16.04)
28 4.15.0-32 (Ubuntu 18.04)
29 5.4.0-48 (Ubuntu 20.04)
30 vanilla linux 5.4.0 (ZFS + ldiskfs)
31 vanilla linux 5.4.21 (ZFS + ldiskfs)
32 vanilla linux 5.4.136 (ZFS + ldiskfs)
33 5.10.0-60.94.0.118.oe2203 (openEuler 22.03 LTS)
34 5.10.0-136.32.0.108.oe2203sp1 (openEuler 22.03 LTS SP1)
35 5.10.0-153.19.0.95.oe2203sp2 (openEuler 22.03 LTS SP2)
36 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
37 * Client primary kernels built and tested during release cycle:
38 5.14.0-362.18.1.el9 (RHEL9.3)
39 5.14.0-284.30.1.el9 (RHEL9.2)
40 4.18.0-513.18.1.el8 (RHEL8.9)
41 4.18.0-477.27.1.el8 (RHEL8.8)
42 5.4.0-37 (Ubuntu 20.04)
43 5.14.21-150500.55.39 (SLES15 SP5)
44 * Other clients known to build on these kernels at some point (others may also work):
45 3.10.0-862.14.4.el7 (RHEL7.5)
46 3.10.0-957.27.2.el7 (RHEL7.6)
47 4.14.0-49.13.1.el7a (RHEL7.5)
48 4.14.0-115.2.2.el7a (RHEL7.6)
49 3.10.0-1062.18.1.el7 (RHEL7.7)
50 3.10.0-1127.19.1.el7 (RHEL7.8)
51 3.10.0-1160.88.1.el7 (RHEL7.9)
52 4.18.0-80.11.2.el8 (RHEL8.0)
53 4.18.0-147.8.1.el8 (RHEL8.1)
54 4.18.0-193.28.1.el8 (RHEL8.2)
55 4.18.0-240.22.1.el8 (RHEL8.3)
56 4.18.0-305.25.1.el8 (RHEL8.4)
57 4.18.0-348.23.1.el8 (RHEL8.5)
58 4.18.0-372.32.1.el8 (RHEL8.6)
59 4.18.0-425.10.1.el8 (RHEL8.7)
60 5.14.0-70.30.1.el9 (RHEL9.0)
61 5.14.0-162.23.1.el9 (RHEL9.1)
62 4.4.120-92.70 (SLES12 SP2)
63 4.4.180-94.100 (SLES12 SP3)
64 4.12.14-95.48 (SLES12 SP4)
65 4.12.14-122.133 (SLES12 SP5)
66 4.12.14-197.75 (SLES15 SP1)
67 5.3.18-24.96 (SLES15 SP2)
68 5.3.18-150300.59.93 (SLES15 SP3)
69 5.14.21-150400.24.92 (SLES15 SP4)
70 4.4.0-131 (Ubuntu 16.04)
71 4.15.0-48 (Ubuntu 18.04)
72 5.8.0-53 (Ubuntu 20.04.2 HWE)
73 5.11.0-31 (Ubuntu 20.04.3 HWE)
74 5.11.0 (vanilla kernel.org)
75 5.10.0 (openEuler 22.03 LTS)
76 * Recommended e2fsprogs version: 1.47.0-wc6 or newer
77 * Recommended ZFS version: 2.1.11
78 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
79 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
80 stack overflow. For more information, please refer to bugzilla 17630.
81 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
82 the re-exporting nodes to be mounted with "32bitapi" mount option
84 --------------------------------------------------------------------------------
89 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
90 for currently supported client and server kernel versions.
91 * Server primary kernels built and tested during release cycle:
92 4.18.0-348.2.1.el8 (RHEL8.5)
93 * Other server kernels known to build and work at some point (others may also work):
94 3.10.0-862.14.4.el7 (RHEL7.5)
95 3.10.0-957.27.2.el7 (RHEL7.6)
96 3.10.0-1062.18.1.el7 (RHEL7.7)
97 3.10.0-1127.19.1.el7 (RHEL7.8)
98 3.10.0-1160.49.1.el7 (RHEL7.9)
99 4.18.0-80.11.2.el8 (RHEL8.0)
100 4.18.0-147.8.1.el8 (RHEL8.1)
101 4.18.0-193.28.1.el8 (RHEL8.2)
102 4.18.0-240.22.1.el8 (RHEL8.3)
103 4.18.0-305.25.1.el8 (RHEL8.4)
104 4.4.120-92.70 (SLES12 SP2)
105 4.4.180-94.100 (SLES12 SP3)
106 4.4.0-131 (Ubuntu 16.04)
107 4.15.0-32 (Ubuntu 18.04)
108 5.4.0-48 (Ubuntu 20.04)
109 vanilla linux 5.4.0 (ZFS + ldiskfs)
110 vanilla linux 5.4.21 (ZFS + ldiskfs)
111 vanilla linux 5.4.136 (ZFS + ldiskfs)
112 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
113 * Client primary kernels built and tested during release cycle:
114 4.18.0-348.2.1.el8 (RHEL8.5)
115 5.4.0-37 (Ubuntu 20.04)
116 5.3.18-59.27 (SLES15 SP3)
117 * Other clients known to build on these kernels at some point (others may also work):
118 3.10.0-862.14.4.el7 (RHEL7.5)
119 3.10.0-957.27.2.el7 (RHEL7.6)
120 4.14.0-49.13.1.el7a (RHEL7.5)
121 4.14.0-115.2.2.el7a (RHEL7.6)
122 3.10.0-1062.18.1.el7 (RHEL7.7)
123 3.10.0-1127.19.1.el7 (RHEL7.8)
124 3.10.0-1160.49.1.el7 (RHEL7.9)
125 4.18.0-80.11.2.el8 (RHEL8.0)
126 4.18.0-147.8.1.el8 (RHEL8.1)
127 4.18.0-193.28.1.el8 (RHEL8.2)
128 4.18.0-240.22.1.el8 (RHEL8.3)
129 4.18.0-305.25.1.el8 (RHEL8.4)
130 4.4.120-92.70 (SLES12 SP2)
131 4.4.180-94.100 (SLES12 SP3)
132 4.12.14-95.48 (SLES12 SP4)
133 4.12.14-122.91 (SLES12 SP5)
134 4.12.14-197.75 (SLES15 SP1)
135 5.3.18-24.96 (SLES15 SP2)
136 4.4.0-131 (Ubuntu 16.04)
137 4.15.0-48 (Ubuntu 18.04)
138 5.8.0-53 (Ubuntu 20.04.2 HWE)
139 5.11.0-31 (Ubuntu 20.04.3 HWE)
140 5.11.0 (vanilla kernel.org)
141 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
142 * Recommended ZFS version: 2.0.7
143 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
144 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
145 stack overflow. For more information, please refer to bugzilla 17630.
146 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
147 the re-exporting nodes to be mounted with "32bitapi" mount option
149 --------------------------------------------------------------------------------
153 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
154 for currently supported client and server kernel versions.
155 * Server primary kernels built and tested during release cycle:
156 3.10.0-1062.18.1.el7 (RHEL7.7)
157 3.10.0-1127.19.1.el7 (RHEL7.8)
158 3.10.0-1160.6.1.el7 (RHEL7.9)
159 * Other server kernels known to build and work at some point (others may also work):
160 3.10.0-862.14.4.el7 (RHEL7.5)
161 3.10.0-957.27.2.el7 (RHEL7.6)
162 4.18.0-80.11.2.el8 (RHEL8.0)
163 4.18.0-147.8.1.el8 (RHEL8.1)
164 4.18.0-193.28.1.el8 (RHEL8.2)
165 4.18.0-240.1.1.el8 (RHEL8.3)
166 4.4.120-92.70 (SLES12 SP2)
167 4.4.180-94.100 (SLES12 SP3)
168 4.4.0-131 (Ubuntu 16.04)
169 4.15.0-32 (Ubuntu 18.04)
170 5.4.0-48 (Ubuntu 20.04)
171 vanilla linux 5.4.0 (ZFS + ldiskfs)
172 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
173 * Client primary kernels built and tested during release cycle:
174 3.10.0-1062.18.1.el7 (RHEL7.7)
175 3.10.0-1127.19.1.el7 (RHEL7.8)
176 3.10.0-1160.6.1.el7 (RHEL7.9)
177 4.12.14-95.48 (SLES12 SP4)
178 4.12.14-122.20 (SLES12 SP5)
179 4.12.14-197.75 (SLES15 SP1)
180 5.3.18-24.24 (SLES15 SP2)
181 4.15.0-48 (Ubuntu 18.04)
182 * Other clients known to build on these kernels at some point (others may also work):
183 3.10.0-862.14.4.el7 (RHEL7.5)
184 3.10.0-957.27.2.el7 (RHEL7.6)
185 4.14.0-49.13.1.el7a (RHEL7.5)
186 4.14.0-115.2.2.el7a (RHEL7.6)
187 4.18.0-80.11.2.el8 (RHEL8.0)
188 4.18.0-147.8.1.el8 (RHEL8.1)
189 4.18.0-193.28.1.el8 (RHEL8.2)
190 4.18.0-240.1.1.el8 (RHEL8.3)
191 4.4.120-92.70 (SLES12 SP2)
192 4.4.180-94.100 (SLES12 SP3)
193 4.4.0-131 (Ubuntu 16.04)
194 5.4.0-37 (Ubuntu 20.04)
195 5.4.0 (vanilla kernel.org)
196 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
197 * Recommended ZFS version: 2.0.0
198 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
199 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
200 stack overflow. For more information, please refer to bugzilla 17630.
201 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
202 the re-exporting nodes to be mounted with "32bitapi" mount option
204 --------------------------------------------------------------------------------
208 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
209 for currently supported client and server kernel versions.
210 * Server primary kernels built and tested during release cycle:
211 3.10.0-1062.1.1.el7 (RHEL7.7)
212 * Other server kernels known to build and work at some point (others may also work):
213 3.10.0-862.14.4.el7 (RHEL7.5)
214 3.10.0-957.27.2.el7 (RHEL7.6)
215 4.18.0-80.7.1.el8 (RHEL8.0)
216 4.4.120-92.70 (SLES12 SP2)
217 4.4.180-94.100 (SLES12 SP3)
218 4.4.0-131 (Ubuntu 16.04)
219 4.15.0-32 (Ubuntu 18.04)
220 vanilla linux 4.6.7 (ZFS only)
221 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
222 * Client primary unpatched kernels built and tested during release cycle:
223 3.10.0-1062.1.1.el7 (RHEL7.7)
224 4.12.14-95.29 (SLES12 SP4)
225 4.15.0-48 (Ubuntu 18.04)
226 * Other clients known to build on unpatched kernels at some point (others may also work):
227 3.10.0-862.14.4.el7 (RHEL7.5)
228 3.10.0-957.27.2.el7 (RHEL7.6)
229 4.14.0-49.13.1.el7a (RHEL7.5)
230 4.14.0-115.2.2.el7a (RHEL7.6)
231 4.18.0-80.7.1.el8 (RHEL8.0)
232 4.4.120-92.70 (SLES12 SP2)
233 4.4.180-94.100 (SLES12 SP3)
234 4.4.0-131 (Ubuntu 16.04)
235 vanilla linux 4.15.0 (ZFS only)
236 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
237 * Recommended ZFS / SPL version: 0.7.13
238 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
239 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
240 stack overflow. For more information, please refer to bugzilla 17630.
241 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
242 the re-exporting nodes to be mounted with "32bitapi" mount option
243 * Note for Ubuntu 14.04.5:
244 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
245 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
247 --------------------------------------------------------------------------------
251 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
252 for currently supported client and server kernel versions.
253 * Server known to build on the following kernels (others may also work):
254 2.6.32-696.18.7.el6 (RHEL6.9)
255 3.10.0-862.14.4.el7 (RHEL7.5)
256 3.10.0-957.el7 (RHEL7.6)
257 4.4.120-92.70 (SLES12 SP2)
258 4.4.162-94.69 (SLES12 SP3)
259 4.4.0-131 (Ubuntu 16.04)
260 4.15.0-32 (Ubuntu 18.04)
261 vanilla linux 4.6.7 (ZFS only)
262 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
263 * Client known to build on unpatched kernels (others may also work):
264 2.6.32-696.18.7.el6 (RHEL6.9)
265 2.6.32-754.6.3.el6 (RHEL6.10)
266 3.10.0-862.14.4.el7 (RHEL7.5)
267 3.10.0-957.el7 (RHEL7.6)
268 4.14.0-49.13.1.el7a (RHEL7.5)
269 4.14.0-115.2.2.el7a (RHEL7.6)
270 4.4.120-92.70 (SLES12 SP2)
271 4.4.162-94.69 (SLES12 SP3)
272 4.4.0-131 (Ubuntu 16.04)
273 4.15.0-32 (Ubuntu 18.04)
274 vanilla linux 4.15.0 (ZFS only)
275 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
276 * Recommended ZFS / SPL version: 0.7.9
277 * Tested with ZFS / SPL version: 0.7.9
278 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
279 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
280 stack overflow. For more information, please refer to bugzilla 17630.
281 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
282 the re-exporting nodes to be mounted with "32bitapi" mount option
283 * Note for Ubuntu 14.04.5:
284 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
285 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
287 --------------------------------------------------------------------------------
289 2018-04-03 Intel Corporation
291 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
292 for currently supported client and server kernel versions.
293 * Server known to build on patched kernels:
294 2.6.32-431.29.2.el6 (RHEL6.5)
295 2.6.32-504.30.3.el6 (RHEL6.6)
296 2.6.32-573.26.1.el6 (RHEL6.7)
297 2.6.32-642.15.1.el6 (RHEL6.8)
298 2.6.32-696.18.7.el6 (RHEL6.9)
299 3.10.0-693.21.1.el7 (RHEL7.4)
300 3.0.101-0.47.71 (SLES11 SP3)
301 3.0.101-107 (SLES11 SP4)
302 3.12.74-60.64.40 (SLES12 SP1)
303 4.4.114-92.64 (SLES12 SP2)
304 4.4.114-94.11 (SLES12 SP3)
305 3.13.0-101 (Ubuntu 14.04, ZFS only)
306 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
307 4.4.0-116 (Ubuntu 16.04)
308 vanilla linux 4.6.7 (ZFS only)
309 * Client known to build on unpatched kernels:
310 2.6.32-431.29.2.el6 (RHEL6.5)
311 2.6.32-504.30.3.el6 (RHEL6.6)
312 2.6.32-573.26.1.el6 (RHEL6.7)
313 2.6.32-642.15.1.el6 (RHEL6.8)
314 2.6.32-696.18.7.el6 (RHEL6.9)
315 3.10.0-693.21.1.el7 (RHEL7.4)
316 3.0.101-0.47.71 (SLES11 SP3)
317 3.0.101-107 (SLES11 SP4)
318 3.12.74-60.64.40 (SLES12 SP1)
319 4.4.114-92.64 (SLES12 SP2)
320 4.4.114-94.11 (SLES12 SP3)
321 3.13.0-101 (Ubuntu 14.04)
322 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
323 4.4.0-116 (Ubuntu 16.04)
325 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
326 * Recommended ZFS / SPL version: 0.7.6
327 * Tested with ZFS / SPL version: 0.7.6
328 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
329 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
330 stack overflow. For more information, please refer to bugzilla 17630.
331 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
332 the re-exporting nodes to be mounted with "32bitapi" mount option
333 * Note for Ubuntu 14.04.5:
334 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
335 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
337 --------------------------------------------------------------------------------
339 2017-07-05 Intel Corporation
341 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
342 for currently supported client and server kernel versions.
343 * Server known to build on patched kernels:
344 2.6.32-431.29.2.el6 (RHEL6.5)
345 2.6.32-504.30.3.el6 (RHEL6.6)
346 2.6.32-573.26.1.el6 (RHEL6.7)
347 2.6.32-642.15.1.el6 (RHEL6.8)
348 2.6.32-696.3.1.el6 (RHEL6.9)
349 3.10.0-514.21.1.el7 (RHEL7.3)
350 3.0.101-0.47.71 (SLES11 SP3)
351 3.0.101-100 (SLES11 SP4)
352 3.12.74-60.64.40 (SLES12 SP1)
353 4.4.59-92.17 (SLES12 SP2)
354 vanilla linux 4.6.7 (ZFS only)
355 * Client known to build on unpatched kernels:
356 2.6.32-431.29.2.el6 (RHEL6.5)
357 2.6.32-504.30.3.el6 (RHEL6.6)
358 2.6.32-573.26.1.el6 (RHEL6.7)
359 2.6.32-642.15.1.el6 (RHEL6.8)
360 2.6.32-696.3.1.el6 (RHEL6.9)
361 3.10.0-514.21.1.el7 (RHEL7.3)
362 3.0.101-0.47.71 (SLES11 SP3)
363 3.0.101-100 (SLES11 SP4)
364 3.12.74-60.64.40 (SLES12 SP1)
365 4.4.59-92.17 (SLES12 SP2)
367 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
368 * Recommended ZFS / SPL version: 0.6.5.9
369 * Tested with ZFS / SPL version: 0.6.5.9
370 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
371 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
372 stack overflow. For more information, please refer to bugzilla 17630.
373 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
374 the re-exporting nodes to be mounted with "32bitapi" mount option
376 --------------------------------------------------------------------------------
378 12-07-2016 Intel Corporation
380 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
381 for currently supported client and server kernel versions.
382 * Server known to build on patched kernels:
383 2.6.32-431.29.2.el6 (RHEL6.5)
384 2.6.32-504.30.3.el6 (RHEL6.6)
385 2.6.32-573.26.1.el6 (RHEL6.7)
386 2.6.32-642.11.1.el6 (RHEL6.8)
387 3.10.0-514.el7 (RHEL7.3)
388 3.0.101-0.47.71 (SLES11 SP3)
389 3.0.101-80 (SLES11 SP4)
390 3.12.62-60.64.8 (SLES12 SP1)
391 vanilla linux 4.5.7 (ZFS only)
392 * Client known to build on unpatched kernels:
393 2.6.32-431.29.2.el6 (RHEL6.5)
394 2.6.32-504.30.3.el6 (RHEL6.6)
395 2.6.32-573.26.1.el6 (RHEL6.7)
396 2.6.32-642.11.1.el6 (RHEL6.8)
397 3.10.0-514.el7 (RHEL7.3)
398 3.0.101-0.47.71 (SLES11 SP3)
399 3.0.101-80 (SLES11 SP4)
400 3.12.62-60.64.8 (SLES12 SP1)
402 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
403 * Recommended ZFS / SPL version: 0.6.5.8
404 * Tested with ZFS / SPL version: 0.6.5.8
405 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
406 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
407 stack overflow. For more information, please refer to bugzilla 17630.
408 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
409 the re-exporting nodes to be mounted with "32bitapi" mount option
411 --------------------------------------------------------------------------------
413 02-29-2016 Intel Corporation
415 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
416 for currently supported client and server kernel versions.
417 * Server known to build on patched kernels:
418 2.6.32-431.29.2.el6 (RHEL6.5)
419 2.6.32-504.30.3.el6 (RHEL6.6)
420 2.6.32-573.12.1.el6 (RHEL6.7)
421 3.10.0-327.3.1.el7 (RHEL7.2)
422 3.0.101-0.47.71 (SLES11 SP3)
423 3.0.101-68 (SLES11 SP4)
424 vanilla linux 4.2.1 (ZFS only)
425 * Client known to build on unpatched kernels:
426 2.6.32-431.29.2.el6 (RHEL6.5)
427 2.6.32-504.30.3.el6 (RHEL6.6)
428 2.6.32-573.12.1.el6 (RHEL6.7)
429 3.10.0-327.3.1.el7 (RHEL7.2)
430 3.0.101-0.47.71 (SLES11 SP3)
431 3.0.101-68 (SLES11 SP4)
434 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
435 * Recommended ZFS / SPL version: 0.6.4.2
436 * Tested with ZFS / SPL version: 0.6.4.2
437 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
438 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
439 stack overflow. For more information, please refer to bugzilla 17630.
440 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
441 the re-exporting nodes to be mounted with "32bitapi" mount option
443 --------------------------------------------------------------------------------
445 03-10-2015 Intel Corporation
447 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
448 for currently supported client and server kernel versions.
449 * Server known to build on patched kernels:
450 2.6.32-431.29.2.el6 (RHEL6.5)
451 2.6.32-504.8.1.el6 (RHEL6.6)
452 3.0.101-0.46 (SLES11 SP3)
453 * Client known to build on unpatched kernels:
454 2.6.32-431.29.2.el6 (RHEL6.5)
455 2.6.32-504.8.1.el6 (RHEL6.6)
456 3.10.0-123.20.1.el7 (RHEL7)
457 3.0.101-0.46 (SLES11 SP3)
458 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
459 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
460 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
461 stack overflow. For more information, please refer to bugzilla 17630.
462 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
463 the re-exporting nodes to be mounted with "32bitapi" mount option
465 Severity : enhancement
467 Description: control OST-index in IDIF via ROCOMPAT flag.
468 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
469 last_rcvd file. For new formatted OST device, it will be auto set;
470 for the case of upgrading from old OST device, you can enable it
471 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
472 enabled, for new created OST-object, its IDIF-in-LMA will contain
473 the OST-index; for the existing OST-object, the OSD will convert
474 old format IDIF as new format IDIF with OST-index stored in the
475 LMA EA when accessing such OST-object or via OI scrub. Once such
476 flag is enabled, it cannot be reverted back, so the system cannot
477 be downgraded to the orignal incompatible version.
479 --------------------------------------------------------------------------------
481 07-30-2014 Intel Corporation
483 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
484 for currently supported client and server kernel versions.
485 * Server known to build on patched kernels:
486 2.6.32-431.20.3.el6 (RHEL6.5)
487 3.0.101-0.31 (SLES11 SP3)
488 * Client known to build on unpatched kernels:
489 2.6.32-431.20.3.el6 (RHEL6.5)
490 3.0.101-0.31 (SLES11 SP3)
493 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
494 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
495 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
496 stack overflow. For more information, please refer to bugzilla 17630.
497 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
498 the re-exporting nodes to be mounted with "32bitapi" mount option
500 --------------------------------------------------------------------------------
502 10-31-2013 Intel Corporation
504 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
505 * Server builds for kernels:
506 2.6.32-358.18.1.el6 (RHEL6.4)
507 * Client builds for unpatched kernels:
508 2.6.32-358.18.1.el6 (RHEL6.4)
509 3.0.92-0.8 (SLES11 SP3)
511 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
512 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
513 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
514 stack overflow. For more information, please refer to bugzilla 17630.
515 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
516 the re-exporting nodes to be mounted with "32bitapi" mount option
518 --------------------------------------------------------------------------------
520 05-30-2013 Intel Corporation
522 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
523 * Server builds for kernels:
524 2.6.32-358.6.2.el6 (RHEL6.4)
525 * Client builds for unpatched kernels:
526 2.6.32-358.6.2.el6 (RHEL6.4)
527 3.0.74-0.6.6 (SLES11 SP2)
529 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
530 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
531 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
532 stack overflow. For more information, please refer to bugzilla 17630.
533 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
534 the re-exporting nodes to be mounted with "32bitapi" mount option
536 --------------------------------------------------------------------------------
538 9-30-2012 Whamcloud, Inc.
540 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
541 * Server support for kernels:
542 2.6.32-279.5.1.el6 (RHEL6)
543 * Client support for unpatched kernels:
544 2.6.32-279.5.1.el6 (RHEL6)
546 2.6.18-238.19.1.el5 (RHEL5)
547 2.6.16.60-0.69.1 (SLES10)
548 * Recommended e2fsprogs version: 1.41.90.wc4
549 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
550 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
551 stack overflow. For more information, please refer to bugzilla 17630.
552 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
553 the re-exporting nodes to be mounted with "32bitapi" mount option
555 Severity : enhancement
557 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
559 Severity : enhancement
561 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
563 Severity : enhancement
565 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
567 Severity : enhancement
569 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
571 Severity : enhancement
573 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
575 --------------------------------------------------------------------------------
577 04-30-2012 Whamcloud, Inc.
579 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
580 * Server support for kernels:
581 2.6.32-220.4.2.el6 (RHEL6)
582 * Client support for unpatched kernels:
583 2.6.32-220.4.2.el6 (RHEL6)
585 2.6.18-238.19.1.el5 (RHEL5)
586 2.6.16.60-0.69.1 (SLES10)
587 * Recommended e2fsprogs version: 1.41.90.wc4
588 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
589 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
590 stack overflow. For more information, please refer to bugzilla 17630.
591 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
592 the re-exporting nodes to be mounted with "32bitapi" mount option
594 Severity : enhancement
596 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
598 Severity : enhancement
600 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
602 Severity : enhancement
604 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
606 Severity : enhancement
608 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
610 --------------------------------------------------------------------------------
612 09-30-2011 Whamcloud, Inc.
614 * Server support for kernels:
615 2.6.18-238.19.1.el5 (RHEL5)
616 2.6.32-131.6.1.el6 (RHEL6)
617 * Client support for unpatched kernels:
618 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
619 2.6.18-238.19.1.el5 (RHEL5)
620 2.6.32-131.6.1.el6 (RHEL6)
621 2.6.29.4-167.fc11 (FC11)
622 2.6.16.60-0.69.1 (SLES10)
624 * Recommended e2fsprogs version: 1.41.90.wc3
625 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
626 Lustre filesystem with 4K stack may cause a stack overflow. For more
627 information, please refer to bugzilla 17630.
628 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
629 mounted with "32bitapi" mount option
631 Severity : enhancement
632 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
634 Severity : enhancement
635 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
637 Severity : enhancement
638 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
641 Frequency : only with SLES10
643 Description: Use OFED "KMP" provided by Novell
644 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
645 Lustre is now built against this package. That means you need to
646 install the ofed-kmp package from Novell for the patchless client.
647 Note that the ofed-kmp that Novell ships may not exactly match the
648 kernel version but should still be compatible.
650 Severity : enhancement
652 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
653 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
655 --------------------------------------------------------------------------------
657 2010-07-15 Oracle, Inc.
659 * Server support for kernels:
660 2.6.18-164.11.1.el5 (RHEL 5)
661 2.6.18-164.11.1.0.1.el5 (OEL 5)
662 * Client support for unpatched kernels:
663 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
664 2.6.18-164.11.1.el5 (RHEL 5),
665 2.6.18-164.11.1.0.1.el5 (OEL 5)
666 2.6.16.60-0.42.8 (SLES 10),
668 2.6.29.4-167.fc11 (FC11)
669 * Recommended e2fsprogs version: 1.41.10-sun2
671 Severity : enhancement
673 Description: Distribute lfs_migrate script with lustre
674 Details : The lfs_migrate mentioned in the manual is now included in the
675 lustre tree in lustre/scripts and will distributed in lustre RPMs.
677 Severity : enhancement
679 Description: Failover nid list added to import proc file.
681 Severity : enhancement
683 Description: "som_preview" mount option required to use SOM
684 Details : The clients will need to use som_preview mount option
688 Description: Lustre does not do 1MB IOs to HW RAID
689 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
691 Severity : enhancement
693 Description: lfs quota output is not very convenient for awk/sed-parsing
694 Details : Some positions in lfs quota output table could be empty or
695 non-empty which made it hard to parse it with scripts, now
696 a dash is put instead of space where there is not supposed
701 Description: Update to OFED 1.5.1
702 Details : OFED 1.5.1 will be used in environments that don't use
703 vendor supplied OFED.
707 Description: Remove lnet. prefix from lctl params display
708 Details : Also renamed lnet.memused to lnet_memused to
713 Description: Use conf_param -d to remove permanant settings.
717 Description: Add list_param -R
718 Details : lctl list_param now supports recursive mode.
721 Bugzilla : 21829, 22312
722 Description: Fix llobdstat.
723 Details : 1. Now updating the write_bytes counter
724 2. Port script fixes from 1.8.
728 Description: decrease the usage of memory on clients.
729 Details : 1. On clients, recycle dentries and inodes unused.
730 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
735 Description: llobdstat fix and enhancement
736 Details : add a counter to set a limit to how many samples will be returned
737 fix a wildcard in the path to limit to obdfilter stats only
741 Description: lfs check only allowed for root
742 Details : Code cleanups and added a regresion test of non-root use of
747 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
748 Details : Fix hr->hr_index overload.
752 Description: Fix FIEMAP functionality.
756 Description: FC11 patchless client support added
760 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
761 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
765 Description: New acc-sm script and various test-framework enhancments.
769 Description: Previously landed patch was reverted.
773 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
777 Description: Add ldiskfs maxdirsize mount option.
781 Description: Fix mkfs.lustre support for 16TB LUNs.
785 Description: Print mmp_check_interval
786 Details : print mmp_check_interval and make it possible to abort
787 mount operation in case it takes too long.
791 Description: Add cascading_rw to lustre mpi tests.
795 Description: Store SOM data in EA.
796 Details : Also several SOM related cleanups.
800 Description: Add refcount to md_open_data
804 Description: Use refcount on lov_request_set to prevent memory corruption.
808 Description: Statahead: use parent dir's i_mutex for synchronization.
812 Description: loadgen improvements.
816 Description: Enable peer health detection in ksocklnd and ko2iblnd.
820 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
824 Description: Fix race on obd_devcie:md_stats
828 Description: Fix SLES 11 client builds.
829 Details : Resolved uninitalized pointer gcc warnings.
833 Description: Allow non-root access for "lfs check".
834 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
838 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
840 Severity : enhancement
842 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
843 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
847 Description: Sanity test_27p fails.
848 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
849 Sanity tests 27 were cleaned up and improved.
853 Description: Unify req format on client/servers
854 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
855 buffers and swabbers approach..
860 Description: drop join file stuff.
862 Severity : enhancement
864 Description: Add 2.6.27 and SLES11 (patchless client) support
868 Description: Fix in ptlrpc_expire_one_request() to print the signed time
871 Severity : enhancement
873 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
876 Severity : enhancement
878 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
880 Severity : enhancement
882 Description: correctly handle big reply message.
883 Details : send LNet event if reply is bigger than buffer and adjust this buffer
888 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
890 Severity : enhancement
892 Description: Add support for OEL5.
896 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
898 Severity : enhancement
900 Description: "weak-modules" support
901 Details : Implement "weak-modules" support which enables kernel modules
902 to be used with any kernel that implements the same kABI. In
903 order to achieve this modules are now installed in
904 /lib/modules/$(uname -r)/updates/kernel on all distributions.
908 Description: client could not reconnect to OST because of an active request.
909 Details : The client now retries to reconnect to the same server, if a
910 connect request failed with EBUSY or -EAGAIN. Test case 35b
911 conf-sanity.sh has been created.
915 Description: Don't allow make backward step on assignin osc next id.
916 Details : race between allocation next id and ll_sync thread can be cause
917 of set wrong osc next id and can be kill valid ost objects.
921 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
925 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
929 Description: Error handling in osc_statfs_interpret() has been improved.
930 Details : Check in osc_statfs_interpret() for EBADR.
934 Description: Do not send statfs() requests to OSTs disabled by administrator.
935 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
937 Severity : enhancement
939 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
941 Severity : enhancement
943 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
945 Severity : enhancement
947 Description: Enhancement for lfs(1) command to use numeric uid/gid.
951 Description: Improved cleanup procedure for conf-sanity test_46a.
955 Description: Do not start bulk transfer, if client tried to reconnect.
956 Details : Do not start bulk transfer, if client tried to reconnect,
957 but failed due to active rpcs. Check for exp_abort_active_req
958 has been implemented.
962 Description: Prevent inconsistences between linux and lustre mount structures.
963 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
964 Make the sleep interruptible.
966 Severity : enhancement
968 Description: provide server to client comms path
969 Details : server to client communications path via new LDLM_SET_INFO rpc
973 Description: truncate starts GFP_FS allocation under transaction causing deadlock
974 Details : ldiskfs_truncate calls grab_cache_page which may start page
975 allocation under an open transaction. This may lead to
976 calling prune_icache with consequent lustre reentrance.
980 Description: send correctly lsm on open replay
981 Details : MDS is trust to LSM size on replay open, but client can set wrong size
986 Description: enable client interface failover
987 Details : When a child reconnects from another NID, properly update export
988 nid hash position and ldlm reverse import.
990 Severity : enhancement
992 Description: Communicate OST degraded/readonly state via statfs to MDS
993 Details : Flags in the statfs returned from OSTs indicate whether the
994 OST is in a degraded RAID state, or if the filesystem has
995 turned read-only after a filesystem error is detected.
1000 Description: don't panic if EPROTO was hit when reading symlink
1001 Details : correctly handling request reference in error cases.
1003 Severity : enhancement
1005 Description: Add LustreNetLink, a kernel-userspace communcation path.
1007 Severity : enhancement
1009 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1014 Description: access to llog context before init.
1015 Details : move handling CATALOGS file at osc layer and forbid access to llog
1016 context before init.
1019 Frequency : rare, only when client is writing to OST on local node
1021 Description: Avoid deadlock for local client writes
1022 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1023 memory freeing context. This allows OST threads to set the
1024 PF_MEMALLOC flag on task structures in order to allocate memory
1025 from reserved pools and complete IO.
1026 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1027 so that the OST threads generate memory pressure and allow
1028 inactive pages to be reclaimed.
1030 Severity : enhancement
1032 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1037 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1038 Details : move ldlm namespace creation in setup phase to avoid grab
1039 _lprocfs_lock with cli_sem held.
1043 Description: Temporarily disable grant shrink.
1044 Details : Disable the feature for debugging.
1047 Frequency : only when formatting test filesystems
1049 Description: Unable to run several mkfs.lustre on loop devices at the same time
1050 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1051 formatting. The solution is to proper handle the error.
1053 Severity : enhancement
1055 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1057 Severity : enhancement
1059 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1064 Description: resolve race between obd_disconnect and class_disconnect_exports
1065 Details : if obd_disconnect will be called to already disconnected export he
1066 forget release one reference and osc module can't unloaded.
1068 Severity : enhancement
1070 Description: Allow tuning service thread via /proc
1071 Details : For each service a new
1072 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1073 created that can be used to set min/max thread counts, and get the
1074 current number of running threads.
1079 Descriptoin: don't return error if have partially created objects for file.
1080 Details : lov_update_create_set uses set->set_success as index for created
1081 objects, so if some requests will be failed, they will have hole
1082 at end of array and qos_shrink_lsm for allocate correct lsm.
1084 Severity : enhancement
1086 Description: Update OFED support to 1.4.1
1091 Description: Reduce small size read RPC
1092 Details : Set read-ahead limit for every file and only do read-ahead when
1093 available read-ahead pages are bigger than 1M to avoid small size
1096 Severity : enhancement
1098 Description: move AT tunable parameters for more consistent usage
1099 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1101 Severity : enhancement
1103 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1104 Details : allow skip disconnected ost for send statfs request and hide error
1110 Description: fix racy locking of mballoc block bitmaps causing BUG
1111 Details : The locking of the mballoc buddy bitmap and the in-memory
1112 block bitmap was using two different spin locks in some
1113 cases. This made it possible to incorrectly access the
1114 mballoc bitmap while another process was modifying it,
1115 causing a sanity assertion to fail. While no on-disk corruption
1116 was reported, there was some risk of this happening.
1119 Frequency : rare, on llog test 6
1121 Descriptoin: don't allow connect to already connected import
1122 Details : allowing connect to already connected import is hide connecting problem.
1127 Descriptoin: allow kill process which wait statahead result
1128 Details : for some reasons 'ls' can stick in waiting result from statahead,
1129 in this case need way for kill this process.
1131 Severity : enhancement
1133 Description: Add state history info file, enhance import info file
1134 Details : Track import connection state changes in a new osc/mdc proc file;
1135 add overview-type data to the osc/mdc import proc file.
1137 Severity : enhancement
1139 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1141 Severity : enhancement
1143 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1148 Descriptoin: OST grant too much space to client even there are not enough space.
1149 Details : Client will shrink its grant cache to OST if there are no write
1150 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1151 this grant cache if there are already not enough avaible space
1152 (left_space < total_clients * 32M).
1155 Frequency : start MDS on uncleanly shutdowned MDS device
1157 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1158 Details : stay in waiting mds<>ost recovery finished produce random bugs
1159 due race between two ll_sync thread for one lov target. send
1160 ACTIVATE event only if connect realy finished and import have
1164 Frequency : rare, connect and disconnect target at same time
1166 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1167 Details : don't call obd_disconnect under lov_lock. this long time
1168 operation and can block ptlrpcd which answer to connect request.
1173 Descriptoin: don't lose wakeup for imp_recovery_waitq
1174 Details : recover_import_no_retry or invalidate_import and import_close can
1175 both sleep on imp_recovery_waitq, but we was send only one wakeup
1179 Frequency : always with long access acl
1181 Descriptoin: mds can't pack reply with long acl.
1182 Details : mds don't control size of acl but they limited by reint/getattr
1185 Severity : enhancement
1187 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1189 Severity : enhancement
1191 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1194 Frequency : start MDS on uncleanly shutdowned MDS device
1196 Descriptoin: aborting recovery hang on MDS
1197 Details : don't throttle destroy RPCs for the MDT.
1200 Frequency : on remount
1202 Description: external journal device not working after the remount
1203 Details : clear dev_rdonly flag for external journal devices in
1209 Description: shutdown vs evict race
1210 Details : client_disconnect_export vs connect request race.
1211 if client will evicted at this time - we start invalidate
1212 thread without referece to import and import can be freed
1216 Frequency : rare, need acl's on inode.
1218 Description: client can't handle ost additional correctly
1219 Details : if ost was added after client connected to mds client can have
1220 hit lnet_try_match_md ... to big messages to wide striped files.
1221 in this case need teach client to handle config events about add
1222 lov target and update client max ea size at that event.
1224 Severity : enhancement
1226 Description: Changelogs
1227 Details : Changelogs are a lightweight mechanism to track filesystem
1228 metadata and namespace changes. The changelog is recorded
1229 permanently on the MDTs, and is periodically "consumed" / purged
1230 when records are no longer needed.
1232 Severity : enhancement
1234 Description: compact fld format with extents
1235 Details : Store range of seq rather than every seq in FLD. Seq
1236 controller update FLD rather than clients. In Case of CMD, mdt0
1237 has FLD, all other metadata server act as non persistent proxy
1238 for FLD queries and cache fld entries in fld cache.
1243 Description: don't skip ost target if they assigned to file
1244 Details : Drop slow OSCs if we can, but not for requested start idx.
1245 This means "if OSC is slow and it is not the requested
1246 start OST, then it can be skipped, otherwise skip it only
1247 if it is inactive/recovering/out-of-space.
1251 Description: more cleanup in mds_lov
1252 Details : not send LOV EA under replay, we can't know about they size at this
1253 time. Don't allow client connect to mds before any ost connected,
1254 for avoid problems with LOV EA size and returning EIO to client.
1256 Severity : enhancement
1258 Description: Interoperability at server side (Disk interoperability)
1260 Severity : enhancement
1262 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1264 Severity : enhancement
1266 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1268 Severity : enhancement
1270 Description: New client IO stack (CLIO).
1272 Severity : enhancement
1274 Description: Commit on sharing. Eliminate inter-client dependencies between
1275 uncommitted transactions by doing transaction commits.
1276 Thereby clients may recovery independently.
1279 Frequency : Create a symlink file with a very long name
1281 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1282 Details : If there is no extra space in the request for early cancels,
1283 ldlm_req_handles_avail() returns 0 instead of a negative value.
1285 Severity : enhancement
1287 Description: Add /proc entry for import status
1288 Details : The mdc, osc, and mgc import directories now have
1289 an import directory that contains useful import data for debugging
1290 connection problems.
1292 Severity : enhancement
1294 Description: Re-disable certain /proc logging
1295 Details : Enable and disable client's offset_stats, extents_stats and
1296 extents_stats_per_process stats logging on the fly.
1299 Frequency : Only on FC kernels 2.6.22+
1301 Description: oops in statahead
1302 Details : Do not drop reference count for the dentry from VFS when lookup,
1303 VFS will do that by itself.
1305 Severity : enhancement
1307 Description: Generic /proc file permissions
1308 Details : Set /Proc file permissions in a more generic way to enable non-
1309 root users operate on some /proc files.
1313 Description: Hitting mdc_commit_close() ASSERTION
1314 Details : Properly handle request reference release in
1315 ll_release_openhandle().
1319 Description: quota recovery deadlock during mds failover
1320 Details : This patch includes att18982, att18236, att18237 in bz14840.
1322 1. fix osts hang when mds does failover with quotaon
1323 2. prevent watchdog storm when osts threads wait for the
1328 Frequency : only patchless client
1329 Description: add workaround for race between add/remove dentry from hash
1331 Severity : enhancement
1333 Description: Allow OST glimpses to return PW locks
1337 Description: LBUG when llog conf file is full
1338 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1342 Description: Prevent import from entering FULL state when server in recovery
1346 Description: service mount cannot take device name with ":"
1347 Details : Only when device name contains ":/" will mount treat it as
1353 Description: replace ptlrpcd with the statahead thread to interpret the async
1354 statahead RPC callback
1358 Frequency : on recovery
1359 Description: I/O failures after umount during fail back
1360 Details : if client reconnected to restarted server we need join to recovery
1361 instead of find server handler is changed and process self
1362 eviction with cancel all locks.
1364 Severity : enhancement
1366 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1368 Severity : enhancement
1370 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1372 Severity : enhancement
1374 Description: Upcall on Lustre log has been dumped
1375 Details : Allow for a user mode script to be called once a Lustre log has
1376 been dumped. It passes the filename of the dumped log to the
1377 script, the location of the script can be specified via
1378 /proc/sys/lnet/debug_log_upcall.
1383 Description: avoid idr_remove called for id which is not allocated.
1384 Details : Move assigment s_dev for clustered nfs to end of initialization,
1385 to avoid problem with error handling.
1390 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1391 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1392 into CONN_USED_HASH and this prodice warning when put connection
1393 again in unused hash.
1396 Severity : enhancement
1398 Description: File striping can now be set to use an arbitrary pool of OSTs.
1400 Severity : enhancement
1402 Description: Export bytes_read/bytes_write count on OSC/OST.
1406 Description: Early reply size mismatch, MGC loses connection
1407 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1408 the connect flags are properly negotiated.
1413 Description: quotas are not honored with O_DIRECT
1414 Details : all writes with the flag O_DIRECT will use grants which leads to
1415 this problem. Now using OBD_BRW_SYNC to guard this.
1419 Description: add quota statistics
1420 Details : 1. sort out quota proc entries and proc code.
1421 2. add quota statistics
1423 Severity : enhancement
1425 Description: enable quota support for HEAD.
1429 Description: Properly propagate oinfo flags from lov to osc for statfs
1430 Details : restore missing copy oi_flags to lov requests.
1432 Severity : enhancement
1434 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1435 : l_getgroups(8), lst(8), routerstat(8)
1436 Details : included man pages for llobdstat(8), llstat(8),
1437 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1439 Severity : enhancement
1441 Description: configure's --enable-quota should check the
1442 : kernel .config for CONFIG_QUOTA
1443 Details : configure is terminated if --enable-quota is passed but
1444 : no quota support is in kernel
1448 Description: Remove portals compatibility
1449 Details : Remove portals compatibility, not interoperable with releases
1454 Description: Resolve device initialization race
1455 Details : Prevent proc handler from accessing devices added to the
1456 obd_devs array but yet be intialized.
1458 Severity : enhancement
1460 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1462 Severity : enhancement
1464 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1469 Description: Using wrong pointer in osc_brw_prep_request
1470 Details : Access to array[-1] can produce panic if kernel compiled with
1471 CONFIG_PAGE_ALLOC enabled
1475 Description: Client runs out of low memory
1476 Details : Consider only lowmem when counting initial number of llap pages
1480 Description: *optional* service tags registration
1481 Details : if the "service tags" package is installed on a Lustre node
1482 When the filesystem is mounted, a local-node service tag will
1483 be created. See http://inventory.sun.com/ for more information
1484 about the Service Tags asset management system.
1488 Description: Kernel BUG tries to release flock
1489 Details : Lustre does not destroy flock lock before last reference goes
1490 away. So always drop flock locks when client is evicted and
1491 perform unlock regardless of successfulness of speaking to MDS.
1495 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1499 Description: sanity test 65a fails if stripecount of -1 is set
1500 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1505 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1506 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1509 Severity : enhancement
1511 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1515 Description: do not process already freed flock
1516 Details : flock can possibly be freed by another thread before it reaches
1517 to ldlm_flock_completion_ast.
1521 Description: LBUG during stress test
1522 Details : Need properly lock accesses the flock deadlock detection list.
1526 Description: oops in page fault handler
1527 Details : kernel page fault handler can return two special 'pages' in error
1528 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1532 Description: timeout with invalidate import.
1533 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1534 should be handled by ptlrpcd. This produce long age waiting and
1535 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1537 Severity : enhancement
1539 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1543 Description: enable MGS and MDT services start separately
1544 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1545 a co-located MGS without starting the MGS, which is a complement
1546 to 'nosvc' mount option.
1550 Frequency : after recovery
1551 Description: precreate to many object's after del orphan.
1552 Details : del orphan st in oscc last_id == next_id and this triger growing
1553 count of precreated objects. Set flag LOW to skip increase count
1554 of precreated objects.
1558 Frequency : rare, on clear nid stats
1559 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1560 Details : when clean nid stats sometimes try destroy live entry,
1561 and this produce panic in free.
1565 Description: Stack overflow during MDS log replay
1566 ease stack pressure by using a thread dealing llog_process.
1570 Description: wait until IO finished before start new when do lock cancel.
1571 Details : VM protocol want old IO finished before start new, in this case
1572 need wait until PG_writeback is cleared until check dirty flag and
1573 call writepages in lock cancel callback.
1575 Severity : enhancement
1577 Description: using special macro for print time and cleanup in includes.
1581 Description: mds_mfd_close() ASSERTION(rc == 0)
1582 Details : In mds_mfd_close(), we need protect inode's writecount change
1583 within its orphan write semaphore to prevent possible races.
1587 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1591 Frequency : rare, on shutdown ost
1592 Description: don't hit live lock with umount ost.
1593 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1594 use shrink_dcache_sb instead.
1598 Description: don't panic with use echo client
1599 Details : echo client pass NULL as client nid pointer and this produce null
1600 pointer dereference.
1604 Description: fix build on ppc32
1605 Details : compile code with -m64 flag produce wrong object file for ppc32.
1609 Description: add message levels for liblustreapi
1613 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1614 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1615 will result in an -EAGAIN and in the exit call path we call
1616 alloc_rr() which will with increasing aggressiveness attempt to
1617 aquire precreated objects on the minimum number of required OSCs.
1621 Description: Use old size assignment to avoid deadlock
1622 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1623 the scheduling while holding a spinlock. We do not need locking
1624 for size in ll_update_inode() because size is only updated from
1625 the MDS for directories or files without objects, so there is no
1626 other place to do the update, and concurrent access to such inodes
1627 are protected by the inode lock.
1631 Description: resolve "_IOWR redefined" build error on SLES10
1635 Description: dump the memory debugging after all modules are unloaded to
1636 suppress false negative in conf_sanity test 39
1638 Severity : enhancement
1640 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1645 Description: fix SLES kernel versioning
1646 Details : the kernel version for our SLES 10 kernel did not include a "-"
1647 before the "smp" at the end. while this was not a problem in
1648 general, it did mean that software trying to use the kernel
1649 version to try to detect a vendor specific kernel would fail.
1650 this was most evident by the OFED build scripts.
1654 Description: Don't update lov_desc members until making sure they are valid
1655 Details : When updating lov_desc members via proc fs, need fix their
1656 validities before doing the real update.
1660 Description: don't put request into delay list while invalidate in flight.
1661 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1662 import in flight. this produce timeout for invalidate and sometimes
1663 can cause stale data.
1667 Frequency : on ppc only
1668 Description: not convert ost objects for directory because it's not exist.
1669 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1673 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1675 Severity : enhancement
1677 Description: Update to support 2.6.22.14 vanilla kernel.
1681 Frequency : rare, on recovery
1682 Description: read procfs can produce deadlock in some situation
1683 Details : Holding lprocfs lock which send rpc can produce block for destroy
1684 obd objects and this also block reconnect with -EALREADY. This
1685 isn't fix all lprocfs bugs - but make it rare.
1687 Severity : enhancement
1689 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1692 Frequency : frequent on X2 node
1694 Description: mdc_set_open_replay_data LBUG
1695 Details : Set replay data for requests that are eligible for replay.
1699 Description: lustre_mgs: operation 101 on unconnected MGS
1700 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1701 MGC, and late on MGC cannot successfully connect to MGS and a lot
1702 of the error messages complaining that MGS is not connected.
1706 Frequency : on network error
1707 Description: panic with double free request if network error
1708 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1709 but it's true only for synchronus enqueue, for async enqueue
1710 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1713 Severity : enhancement
1715 Description: client-side metadata stat-ahead during readdir(directory readahead)
1716 Details : perform client-side metadata stat-ahead when the client detects
1717 readdir and sequential stat of dir entries therein
1720 Frequency : on start mds
1722 Description: Implement get_info(last_id) in obdfilter.
1725 Frequency : occasional
1727 Description: Correctly check stale fid, not start epoch if ost not support SOM
1728 Details : open with flag O_CREATE need set old fid in op_fid3 because
1729 op_fid2 was overwritten with new generated fid, but mds can answer
1730 with one of these two fids and both is not stale. Setattr
1731 incorrectly started an epoch and assume will be called
1732 done_writing, but without SOM done_writing ever being called.
1735 Frequency : rare, depends on device drivers and load
1737 Description: MDS or OSS nodes crash due to stack overflow
1738 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1739 In some cases, in conjunction with device drivers that use a lot
1740 of stack the MDS (or possibly OSS) service threads could overflow
1741 the stack. One change which was identified to consume additional
1742 stack has been reworked to avoid the extra stack usage.
1745 Frequency : occasional
1747 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1748 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1749 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1750 oscc_has_objects and see OSC has no objests and call
1751 oscc_internal_create to resend create request.
1753 Severity : enhancement
1755 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1757 Severity : enhancement
1759 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1762 Frequency : very rare
1764 Description: Fix replay if there is an un-replied request and open
1765 Details : In some cases, older replay request will revert the
1766 mcd->mcd_last_xid on MDS which is used to record the client's
1767 latest sent request.
1769 Severity : enhancement
1771 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1773 Severity : enhancement
1775 Description: Add rhel5 support to HEAD.
1777 Serverity : enhancement
1779 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1784 Description: Don't allow skipping OSTs if index has been specified.
1785 Details : Don't allow skipping OSTs if index has been specified, make
1786 locking in internal create lots better.
1790 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1791 Details : access to bitfield in structure is always rounded to long
1792 and this produce problem with not atomic change any bit.
1796 Description: Lustre make rpms failed.
1797 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1798 builds Lustre rpms from tarball.
1801 Frequency : rare on shutdown ost
1803 Description: If llog cancel was not send before clean_exports phase, this can
1804 produce deadlock in llog code.
1805 Details : If llog thread has last reference to obd and call class_import_put
1806 this produce deadlock because llog_cleanup_commit_master wait when
1807 last llog_commit_thread exited, but this never success because was
1808 called from llog_commit_thread.
1812 Description: allow userland application know is lost one of stripes.
1813 Details : fill lvb_blocks with error code on ost and return it to
1814 application if error flag found.
1818 Description: NULL lov_tgts causing MDS oops
1819 Details : more safe checks for NULL lov_tgts for avoid oops.
1821 Severity : enhancement
1823 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1827 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1828 Details : If an OST is down the MDS will hang indefinitely in
1829 obd_statfs_rqset() waiting for the statfs data. While for
1830 MDS QOS usage of statfs, it should not stuck in waiting.
1832 Severity : enhancement
1834 Description: remote_acl support
1835 Details : Support ACL-based permission check for remote user.
1836 Support setfacl/getfacl for remote user with the utils
1837 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1838 the system "{s,g}etfacl" utils.
1840 Severity : enhancement
1842 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1844 Severity : enhancement
1846 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1850 Description: make mgs_setparam() handle fsname containing dash
1851 Details : fsname containing a dash does not work with lctl conf_param
1853 Severity : enhancement
1855 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1857 Severity : enhancement
1859 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1863 Frequency : rare, at shutdown
1864 Description: access already free / zero obd_namespace.
1865 Details : if client_disconnect_export was called without force flag set,
1866 and exist connect request in flight, this can produce access to
1867 NULL pointer (or already free pointer) when connect_interpret
1868 store ocd flags in obd_namespace.
1872 Frequency : only at startup
1873 Description: not alloc memory with spinlock held.
1874 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1875 if any spinlock held.
1877 Severity : enhancement
1879 Description: make lustre randomly fail allocating memory
1880 Details : Make lustre randomly failed allocating memory for testing purpose.
1882 Severity : enhancement
1884 Description: lost problems with lov objid file
1885 Details : Fixes some scability and access to not inited memory problems
1886 in work with lov objdid file.
1891 Description: lfs find does not continue on file error
1892 Details : Continue other files processing when a file/dir is absent.
1896 Description: Inconsistent usage of lustre_pack_reply()
1897 Details : Standardize the usage of lustre_pack_reply() such that it
1898 always generate a CERROR on failure.
1901 Frequency : occasional
1903 Description: MDS hang or stay in waiting lock
1904 Details : If client receive lock with CBPENDING flag ldlm need send lock
1905 cancel as separate rpc, to avoid situation when cancel request
1906 can't processed due all i/o threads stay in wait lock.
1910 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1914 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1918 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1922 Description: Update to RHEL4 latest kernel.
1924 Severity : enhancement
1926 Description: Build SLES10 patchless client fails
1927 Details : The configure was broken by run ./configure with
1928 --with-linux-obj=.... argument for patchless client. When the
1929 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1930 can't search header adequately. Use absolute path such as
1931 -I($LINUX)/include instead.
1935 Description: interrupt oig_wait produce painc on resend.
1936 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1937 private set, and this produce situation when rq_ptlrpcd_data not
1938 copyed to new allocated request and triggered LBUG on assert
1939 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1940 ptlrpcd set if request is changed and can be safety changed to use
1945 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1946 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1947 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1948 these macros and replace them with their difinitions instead.
1952 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1953 Details : Update which_patch & target file for SLES10 latest kernel.
1956 Description: Remove client patches from SLES 10 kernel.
1957 Details : This causes SLES 10 clients to behave as patchless clients
1958 even on a Lustre-patched (server) kernel.
1960 Severity : enhancement
1962 Description: self-adjustable client's lru lists
1963 Details : use adaptive algorithm for managing client cached locks lru
1964 lists according to current server load, other client's work
1965 pattern, memory activities, etc. Both, server and client
1966 side namespaces provide number of proc tunables for controlling
1969 Severity : enhancement
1971 Description: light-weight GSS support
1972 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1973 and reduce performance overhead.
1975 Severity : enhancement
1977 Description: Linux keyring support
1978 Details : Support using service of Linux keyring for Lustre GSS internal
1979 context refresh/cache mechanism.
1983 Description: Fix errors in lfs documentation
1984 Details : Fixes man pages
1988 Description: don't use GFP_* in generic Lustre code.
1989 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1993 Description: obdclass is limited by single OBD_ALLOC(idarray)
1994 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1998 Description: warnings with build patchless client with vanila 2.6.19 and up
1999 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2000 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2004 Frequency : Only for SLES
2005 Description: sanity_quota fail test_1
2006 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2007 file, which makes TSTID[2] inunique.
2011 Frequency : Only for SLES9
2012 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2013 Details : Update target/ChangeLog/which_patch .
2017 Frequency : Only for test_10 in sanity-quota.sh
2018 Description: a bug in quota test code
2019 Details : checking a test flag in a wrong place
2023 Frequency : Only for test_13 in sanity-quota.sh
2024 Description: a bug in quota test script
2025 Details : shouldn't check overfull allocated quota
2029 Description: jbd statistics
2030 Details : Port older jbd statistics patch for sles10
2032 Severity : enhancement
2034 Description: Add printing inode info into message about error in writepage.
2039 Description: reply_lock_interpret crash due to race with it and lock cancel.
2040 Details : Do not replay locks that are being cancelled. Do not reference
2041 locks by their address during replay, just by their handle.
2045 Frequency : When flocks are used.
2046 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2047 Details : Flock locks might destroy just granted lock if it could be merged
2048 with another existing flock, this is done in completion handler,
2049 so teach ldlm_cli_enquque_fini that this is a valid case for
2055 Description: Oops in read and write path when failing to allocate lock.
2056 Details : Check if lock allocation failed and return error back.
2060 Description: lstripe command fails for valid OST index
2061 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2062 instead of lov->desc.ld_active_tgt_count.
2066 Description: sanity.sh failed test 103
2067 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2072 Description: df doesn't work properly if diskfs blocksize != 4K
2073 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2077 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2078 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2080 Severity : enhancement
2082 Description: Adaptive timeouts
2083 Details : RPC timeouts adapt to changing server load and network
2084 conditions to reduce resend attempts and improve recovery time.
2088 Description: llapi_file_create() does not allow some changes
2089 Details : add llapi_file_open() that allows specifying the mode and
2090 open flags, and also returns an open file handle.
2094 Description: merge and cleanup kernel patches.
2095 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2099 Description: Add journal checksum support.(Kernel part)
2100 Details : The journal checksum feature adds two new flags i.e
2101 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2102 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2103 indicates that the commit block contains the checksum for
2104 the blocks described by the descriptor blocks. Now commit
2105 record can be sent to disk without waiting for descriptor
2106 blocks to be written to disk. This behavior is controlled
2107 using JBD2_FEATURE_ASYNC_COMMIT flag.
2111 Description: OSS needs mutliple precreate threads
2112 Details : Add ability to start more than one create thread per OSS.
2116 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2117 Details : execute lfs setstripe on client
2121 Description: mds_obd_create error creating tmp object
2122 Details : When the user sets quota on root, llog will be affected and can't
2123 create files and write files.
2126 Frequency : Always on ia64 patchless client, and possibly others.
2128 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2129 Details : This allows the patchless client to be loaded on architectures
2130 without this export.
2134 Description: RedHat Update kernel for RHEL5
2135 Details : Add the kernel config file for RHEL5.
2139 Description: RedHat Update kernel for RHEL5
2140 Details : Modify the kernel config file more closer RHEL5.
2144 Description: Build failure against Centos5 (RHEL5)
2145 Details : Define PAGE_SIZE when it isn't present.
2149 Description: test_7 on sanity-quota.sh can't be used on separate machines
2150 Details : add facet to handle it
2154 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2155 Details : with NFS, the anon dentry's parent was set to itself in
2156 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2157 corresponding dentry other than use rec->ur_name.
2159 Severity : enhancement
2161 Description: lfs setstripe enhancement
2162 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2166 Description: enable data checksumming by default
2167 Details : enable checksum by default, allow --disable-checksum
2168 configure option and "-o nochecksum" mount option. Checksums
2169 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2173 Description: lustre support for RHEL5
2174 Details : Add support for RHEL5.
2178 Description: Client eviction due to failover config
2179 Details : after a connection loss, the lustre client should attempt to
2180 reconnect to the last active server first before trying the
2181 other potential connections.
2185 Description: when mds and osts use different quota unit(32bit and 64bit),
2186 quota will be released repeatly.
2187 Details : void sending multiple quota reqs to mds, which will keep the
2188 status between the reqs.
2192 Description: osts not allocated evenly to files
2193 Details : change the condition to increase offset_idx
2196 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2198 Description: Data corruption for OSTs that are formatted larger than 2TB
2200 Details : When generating the bio request for lustre file writes the
2201 sector number would overflow a temporary variable before being
2202 used for the IO. The data reads correctly from Lustre (which
2203 will overflow in a similar manner) but other file data or
2204 filesystem metadata may be corrupted in some cases.
2208 Description: Tune the kernel for good SCSI performance.
2209 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2210 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2215 Description: rewrite ext2-derived code in obdclass/uuid.c
2216 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2217 from scratch preserving functionality.
2221 Description: "lfs find -obd UUID" prints directories
2222 Details : "lfs find -obd UUID" will return all directory names instead
2223 of just file names. It is incorrect because the directories
2224 do not reside on the OSTs.
2228 Description: use i_size_read and i_size_write in 2.6 port
2229 Details : replace inode->i_size access with i_size_read/write()
2232 Frequency : when removing large files
2234 Description: scheduling issue during removal of large Lustre files
2235 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2236 It causes scheduling issues when removing large files (17TB in the
2240 Frequency : only with liblustre clients on XT3
2242 Description: evictions taking too long
2243 Details : allow llrd to evict clients directly on OSTs
2246 Frequency : only on ppc
2248 Description: /proc/fs/lustre/devices broken on ppc
2249 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2250 We should make sure the type of 'index' is loff_t and then cast
2251 explicitly as needed below. Do not assign an explicitly cast
2255 Frequency : only for rhel5
2257 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2258 Details : Modify the target file & which_kernel.
2260 Severity : enhancement
2262 Description: omit set fsid for export NFS
2263 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2264 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2268 Frequency : Always, for kernels after 2.6.16
2269 Description: Fix warning idr_remove called for id=.. which is not allocated.
2270 Details : Last kernels save old s_dev before kill super and not allow
2271 to restore from callback - restore it before call kill_anon_super.
2275 Description: buffer overruns could theoretically occur
2276 Details : llapi_semantic_traverse() modifies the "path" argument by
2277 appending values to the end of the origin string, and a
2278 overrun may occur. Adding buffer overrun check in liblustreapi.
2282 Description: Fix error on 'ls .' at the top of the Lustre mount.
2283 Details : Don't revalidate dentry if it is a root dentry.
2287 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2288 Details : Modify vm-tunables-rhel4.patch.
2292 Description: Kernel config for 2.6.18-vanilla.
2293 Details : Modify targets/2.6-vanilla.target.in.
2294 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2295 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2296 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2297 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2301 Description: improve handling recoverable errors
2302 Details : if request processig with error which can be recoverable on server
2303 request should be resend, otherwise page released from cache and
2308 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2309 Details : Modify target file & which_patch.
2310 A flaw was found in the IA32 system call emulation provided
2311 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2312 value could be stored in the %RAX register, which could trigger an
2313 out-of-bounds system call table access. An untrusted local user
2314 could exploit this flaw to run code in the kernel
2315 (ie a root privilege escalation). (CVE-2007-4573).
2319 Description: O_DIRECT bypasses client statistics.
2320 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2321 read_ahead_stats were updated. copied the stats section from
2322 osc_send_oap_rpc() into async_internal().
2326 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2330 Description: change order of libsysio includes
2331 Details : '#include sysio.h' should always come before '#include xtio.h'
2335 Description: handle "serious error: objid * already exists" more gracefully
2336 Details : If LAST_ID value on disk is smaller than the objects existing in
2337 the O/0/d* directories, it indicates disk corruption and causes an
2338 LBUG(). If the object is 0-length, then we should use the existing
2339 object. This will help to avoid a full fsck in most cases.
2341 Severity : enhancement
2343 Description: adapt the lustre_config script to support the upgrade case
2344 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2345 server targets to 1.6 in parallel.
2349 Description: To avoid grant space > avaible space when the disk is almost
2350 full. Without this patch you might see the error "grant XXXX >
2351 available" or some LBUG about grant, when the disk is almost
2353 Details : In filter_check_grant, for non_grant cache write, we should
2354 check the left space by if (*left > ungranted + bytes), instead
2355 of (*left > ungranted), because only we are sure the left space
2356 is enough for another "bytes", then the ungrant space should be
2357 increase. In client, we should update cl_avail_grant only there
2358 is OBD_MD_FLGRANT in the reply.
2362 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2363 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2364 described in CVE-2007-4573. This problem could allow untrusted
2365 local users to gain root access.
2368 Frequency : when using O_DIRECT and quotas
2370 Description: Incorrect file ownership on O_DIRECT output files
2371 Details : block usage reported by 'lfs quota' does not take into account
2372 files that have been written with O_DIRECT.
2377 Description: touch file failed when fs is not full
2378 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2379 otherwise we can get ENOSP while fs is not full.
2383 Description: parallel lock callbacks
2384 Details : Instead of sending blocking and completion callbacks as separated
2385 requests, adding them to a set and sending in parallel.
2388 Frequency : only for Cray XT3
2389 Bugzilla : 12829/13455
2390 Description: Changing primary group doesn't change the group lustre assigns to
2392 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2393 overridden with the primary group provided by the group upcall,
2394 whereas the supplied fsgid can be trusted if it is in the list of
2395 supplementary groups returned by the group upcall.
2397 Severity : enhancement
2399 Description: Allow masking D_WARNING, D_ERROR messages from console
2400 Details : Console messages can now be disabled via lnet.printk.
2404 Description: User code with malformed file open parameter crashes client node
2405 Details : Before packing join_file req, all the related reference should be
2406 checked carefully in case some malformed flags cause fake
2407 join_file req on client.
2411 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2412 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2413 mode is equal to the granted mode. However, it should also check
2414 that LDLM_FL_CP_REQD is not set.
2418 Description: Heavy nfs access might result in deadlocks
2419 Details : After ELC code landed, it is now improper to enqueue any mds
2420 locks under och_sem, because enqueue might want to decide to
2421 cancel open locks for same inode we are holding och_sem for.
2425 Description: Client eviction while running blogbench
2426 Details : A lot of unlink operations with concurrent I/O can lead to a
2427 deadlock causing evictions. To address the problem, the number of
2428 oustanding OST_DESTROY requests is now throttled to
2429 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2430 callbacks are processed in priority.
2434 Description: enable ACLs on MDS by default
2435 Details : ACLs must be enabled on MDS by default.
2438 Frequency : PPC/PPC64 only
2440 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2441 Details : fix duplicated definitions between asm-ppc64/types.h and
2442 lustre_types.h on PPC.
2445 Frequency : PPC/PPC64 only
2447 Description: asm-ppc/segment.h does not exist
2448 Details : fix compile issue on PPC.
2452 Description: better handle error messages in extents code
2455 Frequency : RHEL4 only
2457 Description: mkfs is very slow on IA64/RHEL4
2458 Details : A performance regression has been discovered in the MPT Fusion
2459 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2460 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2461 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2463 Severity : enhancement
2465 Description: SNMP support enhancement
2466 Details : Adding total number of sampled request for an MDS node in snmp
2469 Severity : enhancement
2471 Description: Optimize ldlm waiting list processing for PR extent locks
2472 Details : When processing waiting list for read extent lock and meeting read
2473 lock that is same or wider to it that is not contended, skip
2474 processing rest of the list and immediatelly return current
2475 status of conflictness, since we are guaranteed there are no
2476 conflicting locks in the rest of the list.
2480 Description: Time out and refuse to reconnect
2481 Details : When the failover node is the primary node, it is possible
2482 to have two identical connections in imp_conn_list. We must
2483 compare not conn's pointers but NIDs, otherwise we can defeat
2484 connection throttling.
2488 Description: port llog fixes from b1_6 into HEAD
2489 Details : Port llog reference couting and some llog cleanups from b1_6
2490 (bug 10800) into HEAD, for protect from panic and access to already
2491 free llog structures.
2495 Description: Detect stride IO mode in read-ahead
2496 Details : When a client does stride read, read-ahead should detect that and
2497 read-ahead pages according to the detected stride pattern.
2501 Description: data checksumming impacts single node performance
2502 Details : add support for several checksum algorithm. Currently, only CRC32
2503 and Adler-32 are supported. The checksum type can be changed on
2504 the fly via /proc/fs/lustre/osc/*/checksum_type.
2508 Description: use adler32 for page checksums
2509 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2514 Description: build for x2 fails
2515 Details : fix compile issue on Cray systems.
2519 Description: Properly match for duplicate locks
2520 Details : Due to different lock order from skiplists code, we need to
2521 traverse entire list for now
2524 Frequency : only on PPC/SLES10
2526 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2527 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2528 be defined. Add a hack in mkfs_lustre.c to work around this compile
2533 Description: LASSERT on MDS when client holding flock lock dies
2534 Details : ldlm pool logic depends on number of granted locks equal to
2535 number of released locks which is not true for flock locks, so
2536 just exclude such locks from consideration.
2540 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2541 Details : Use fsfilt_sync() for both the whole filesystem sync and
2542 individual file sync to eliminate dangerous inode locking
2543 with I_LOCK that can lead to a deadlock.
2547 Description: performance in 1.6.3
2548 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2549 sure that 1MB requests are not fragmented by the block layer.
2551 Severity : enhancement
2553 Description: organize the server-side client stats on per-nid basis
2554 Details : Change the structure of stats under obdfilter and mds to
2562 The "uuid"s file would list the uuids of _active_ exports.
2563 And the clear entry is to clear all stats and stale nids.
2565 Severity : enhancement
2567 Description: eliminate client locks in face of contention
2568 Details : file contention detection and lockless i/o implementation
2569 for contended files.
2573 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2576 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2578 Description: MDS may lose file striping (and hence file data) in some cases
2579 Details : If there are additional extended attributes stored on the MDS,
2580 in particular ACLs, SELinux, or user attributes (if user_xattr
2581 is specified for the client mount options) then there is a risk
2582 of attribute loss. Additionally, the Lustre file striping
2583 needs to be larger than default (e.g. striped over all OSTs),
2584 and an additional attribute must be stored initially in the
2585 inode and then increase in size enough to be moved to the
2586 external attribute block (e.g. ACL growing in size) for file
2591 Description: skiplist implementation simplification
2592 Details : skiplists are used to group compatible locks on granted list
2593 that was implemented as tracking first and last lock of each
2594 lock group the patch changes that to using doubly linked lists
2598 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2599 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2600 because LASSERT on both the data supplied by a client, and the
2601 data on disk is dangerous and incorrect.
2603 Severity : enhancement
2605 Description: Slow truncate/writes to huge files at high offsets.
2606 Details : Directly associate cached pages to lock that protect those pages,
2607 this allows us to quickly find what pages to write and remove
2608 once lock callback is received.
2612 Description: more ldlm soft lockups
2613 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2614 starve other threads from the resource lock for a long time in
2615 case of long waiting queue, so change the debug level from
2616 D_OTHER to the less frequently used D_INFO.
2618 Severity : enhancement
2620 Description: add -gid, -group, -uid, -user options to lfs find
2624 Description: Hung threads in invalidate_inode_pages2_range
2625 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2626 once one is completed. As a result, some RPCs are stuck in the
2627 queue and are never sent.
2631 Description: filter threads hungs on waiting journal commit
2632 Details : Cleanup filter group llog code, then only filter group llog will
2633 be only created in the MDS/OST syncing process.
2637 Description: Procfs and llog threads access destoryed import sometimes.
2638 Details : Sync the import destoryed process with procfs and llog threads by
2639 the import refcount and semaphore.
2641 Severity : enhancement
2643 Description: openlock cache of b1_6 port to HEAD
2648 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2649 Details : If insertion of an extent fails, then discard the inode
2650 preallocation and free data blocks else it can lead to duplicate
2655 Description: don't always update ctime in ext3_xattr_set_handle()
2656 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2657 In some cases the ctime should not be updated, for example for
2658 2.0->1.8 compatibility it is necessary to delete an xattr and it
2659 should not update the ctime.
2663 Bugzilla : 15713/16362
2664 Description: Assertion in iopen_connect_dentry in 1.6.3
2665 Details : looking up an inode via iopen with the wrong generation number can
2666 populate the dcache with a disconneced dentry while the inode
2667 number is in the process of being reallocated. This causes an
2668 assertion failure in iopen since the inode's dentry list contains
2669 both a connected and disconnected dentry.
2673 Description: assertion failure in ldlm_handle2lock()
2674 Details : fix a race between class_handle_unhash() and class_handle2object()
2675 introduced in lustre 1.6.5 by bug 13622.
2680 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2681 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2686 Description: oops in sd_iostats_seq_show()
2687 Details : unloading/reloading the scsi low level driver triggers a kernel
2688 bug when trying to access the sd iostat file.
2693 Description: Kernel panics during QLogic driver reload
2694 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2695 patch, causing memory corruption.
2700 Description: journal_dev option does not work in b1_6
2701 Details : pass mount option during pre-mount.
2703 Severity : enhancement
2705 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2706 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2707 extent information of a file. It can be used to map logical blocks
2708 in a file to physical blocks in the block device.
2712 Description: LDLM soft lockups - improvement
2713 Details : It is be possible to send the lock handle along with each read
2714 or write request because the client is already doing a lock match
2715 itself so there isn't any reason the OST should have to re-do that
2719 Frequency : only on Cray X2
2721 Description: X2 build failures
2722 Details : fix build failures on Cray X2.
2726 Description: xid & resent requests
2727 Details : Initialize RPC XID from clock at startup (randomly if clock is
2730 Severity : enhancement
2732 Description: Add lustre_start utility to start or stop multiple Lustre servers
2737 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2738 Details : In case of memory pressure, list_del() can be called twice on
2739 req->rq_history_list, causing a kernel oops.
2743 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2744 Details : fix stack overflow in the distributed lock manager by defering
2745 export eviction after a failed AST to the elt thread instead of
2746 handling it in the dlm interpret routine.
2750 Description: Convert some comments to new format.
2751 Details : Update documenting comments to match doxygen conventions.
2755 Description: Grammar fixes.
2756 Details : A couple of trivial spelling fixes.
2760 Description: OSD_COUNTERS-mandatory
2761 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2762 to simplify the code.
2766 Description: simplify cmm_device freeing logic.
2767 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2771 Description: Add lockdep support to dt_object_operations locking interface.
2772 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2773 parameter indicating lock ordering. Update mdd code to use new
2778 Description: Introduce failloc constants for lockless IO tests.
2779 Details : Add two new failloc constants to test lockless IO. Only one of
2780 them in implemented---another is checked in yet to be landed
2785 Description: Add lockdep support for inode mutex.
2786 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2787 in the situations where Lustre has to lock more than one inode
2792 Description: Add optional invariants checking support.
2793 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2794 configure switch. This macro is to be used for consistency and
2795 sanity checks that are too expensive to be left in `production'
2800 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2801 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2806 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2807 Details : Use spin_lock_nested() in (the only) situation where more than
2808 one ldlm_lock is locked simultaneously. Also, fix possible
2809 dead-lock in ldlm_lock_change_resource() by enforcing particular
2814 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2815 Details : Instead of specifying each ldlm_lock call-back through separate
2816 parameter, wrap them into struct ldlm_callback_suite.
2820 Description: Kill join_lru obd method and its callers.
2821 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2822 in memory---a requirement implemented through ->o_join_lru() obd
2823 method in HEAD. Remove this method and its users.
2827 Description: Add asynchronous ldlm ENQUEUE completion handler.
2828 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2829 case that stock ldlm_completion_ast() cannot handle as it waits
2830 until lock is granted. Introduce new ldlm_completion_ast_async()
2831 for this. Also comment ldlm_completion_ast().
2835 Description: ldlm_error <-> errno conversion.
2836 Details : Add functions to map (rather arbitrary) between LDLM error codes
2837 and standard errno values. CLIO needs this to prevent LDLM specific
2838 constants from escaping ldlm and osc.
2842 Description: Kill unused ldlm_handle2lock_ns() function.
2843 Details : Kill unused ldlm_handle2lock_ns() function.
2847 Description: Add lu_ref support to ldlm_lock
2848 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2849 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2850 and ldlm_resource. LDLM interface has to be changed, because code that
2851 releases a reference on a lock, has to "know" what reference this is.
2852 In the most frequent case
2854 lock = ldlm_handle2lock(handle);
2856 LDLM_LOCK_PUT(lock);
2858 no changes are required. When any other reference (received _not_ from
2859 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2860 instead of LDLM_LOCK_PUT().
2862 Arguably, changes are pervasive, and interface requires some discipline
2863 for proper use. On the other hand, it was very instrumental in finding
2864 a few leaked lock references.
2868 Description: Add ldlm_lock_addref_try().
2869 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2870 attempts to addref a lock that might be being canceled
2875 Description: Add ldlm_weigh_callback().
2876 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2877 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2878 hard-coded `number of pages' logic.
2882 Description: Add lockdep annotations to llog code.
2883 Details : Use appropriately tagged _nested() locking calls in the places
2884 where llog takes more than one ->lgh_lock lock.
2888 Description: Add loi_kms_set().
2889 Details : Wrap kms updates into a helper function.
2893 Description: Constify instances of struct lsm_operations.
2894 Details : Constify instances of struct lsm_operations.
2898 Description: lu_conf support.
2899 Details : On a server, a file system object is uniquely identified
2900 by a fid, which is sufficient to locate and load all object
2901 state (inode). On a client, on the other hand, more data are
2902 necessary instantiate an object. Change lu_object_find() and
2903 friends to take additional `lu_conf' argument describing object.
2904 Typically this includes layout information.
2908 Description: lu_context fixes.
2909 Details : Introduce new lu_context functions that are needed on the client
2910 side, where some system threads (ptlrpcd) are shared by multiple
2911 modules, and so cannot be stopped during module shutdown.
2915 Description: Add start and stop methods to lu_device_type_operations.
2916 Details : Introduce two new methods in lu_device_type_operations, that are
2917 invoked when first instance of a given type is created and last one
2918 is destroyed respectively. This is need by CLIO.
2922 Description: Add lu_ref support to struct lu_device.
2923 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2924 track leaked references.
2928 Description: Introduce lu_kmem_descr.
2929 Details : lu_kmem_descr and its companion interface allow to create
2930 and destroy a number of kmem caches at once.
2934 Description: Fix lu_object finalization race.
2935 Details : Fix a race between lu_object_find() finding an object and its
2936 concurrent finalization. This race is (most likely) not possible
2937 on the server, but might happen on the client.
2941 Description: Introduce lu_ref interface.
2942 Details : lu_ref is a debugging module allowing to track references to
2943 a given object. It is quite cpu expensive, and has to be
2944 explicitly enabled with --enable-lu_ref. See usage description
2949 Description: Factor lu_site procfs stats into a separate function.
2950 Details : Separate lu_site stats printing code into a separate function
2951 that can be reused on a client.
2955 Description: Constify instances of struct {lu,dt,md}_device_operations.
2956 Details : Constify instances of struct {lu,dt,md}_device_operations.
2960 Description: Introduce struct md_site and move meta-data specific parts of
2961 struct lu_site here.
2962 Details : Move md-specific fields out of struct lu_site into special struct
2963 md_site, so that lu_site can be used on a client.
2967 Description: Kill mdd_lov_destroy().
2968 Details : Remove unused mdd code.
2972 Description: Add st_block checking to multistat.c.
2973 Details : Add st_block checking to multistat.c.
2977 Description: Add lu_ref support to struct obd_device.
2978 Details : Add lu_ref tracking to obd_device.
2982 Description: Kill obd_set_fail_loc().
2983 Details : Remove unused code.
2987 Description: Add special type for ptlrpc_request interpret functions.
2988 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2989 there. Actual usage will be in CLIO.
2993 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2994 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2995 doesn't work with lockdep.
2999 Description: Add rwv.c test program.
3000 Details : New testing program exercising readv(2) and writev(2) (Qian).
3004 Description: Add sendfile.c test program.
3005 Details : New testing program exercising sendfile(2) (Jay).
3009 Description: Ratelimit a message that can be very frequent.
3010 Details : Ratelimit a memory allocation failure message that can
3015 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3016 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3020 Description: Liblustre build fixes.
3021 Details : Liblustre build fixes.
3025 Description: libcfs: add cfs_{need,cond}_resched() interface.
3026 Details : libcfs: add cfs_{need,cond}_resched() definition and
3027 implementations for Linux, NT, and liblustre.
3029 Severity : enhancement
3031 Description: More exported tunables for mballoc
3032 Details : Add support for tunable preallocation window and new tunables for
3033 large/small requests
3037 Description: Detect corruption of block bitmap and checking for preallocations
3038 Details : Checks validity of on-disk block bitmap. Also it does better
3039 checking of number of applied preallocations. When corruption is
3040 found, it turns filesystem readonly to prevent further corruptions.
3044 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3045 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3046 window to 0. If it does hit cache miss, and read-pattern is still
3047 stride-io mode, does not reset the stride window, but also does not
3048 increase the stride window length in this case.
3052 Frequency : only for big-endian servers
3053 Description: Check if system is big-endian while mounting fs with extents feature
3054 Details : Mounting a filesystem with extents feature will fail on big-endian
3055 systems since ext3-based ldiskfs is not supported on big-endian
3056 systems. This can be over-riden with "bigendian_extents" mount option.
3058 Severity : enhancement
3060 Description: The root squash functionality
3061 Details : A security feature, which is to prevent users from being able
3062 to mount lustre on their desktop, run as root, and delete
3063 all of the files in the filesystem. The goal is accomplished by
3064 remapping user id (UID) and group id (GID) of the root user to
3065 a UID and GID specified by the system administartor via Lustre
3066 configuration management server (MGS). The functionality also
3067 allows to specify sets of clients for which the remapping does
3072 Description: Excessive recovery window
3073 Details : With AT enabled, the recovery window can be excessively long (6000+
3074 seconds). To address this problem, we no longer use
3075 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3076 timeout no longer depends on the service time, it is set to
3077 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3078 time via pb_service_time.
3082 Description: Watchdog triggered on MDS failover
3083 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3084 the OSTs know that the MDS "UUID" can be reused for the same export
3085 from a different NID, so we do not need to wait for the export to be
3089 Frequency : rare, only if using MMP with Linux RAID
3091 Description: MMP doesn't work with Linux RAID
3092 Details : While using HA for Lustre servers with Linux RAID, it is possible
3093 that MMP will not detect multiple mounts. To make this work we
3094 need to unplug the device queue in RAID when the MMP block is being
3095 written. Also while reading the MMP block, we should read it from
3096 disk and not the cached one.
3098 Severity : enhancement
3100 Description: open file using fid
3101 Details : A file can be opened using just its fid, like
3102 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3105 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3108 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3109 Details : add proper locking for creation of jbd_4k slab cache
3113 Description: MMP check in ext3_remount() fails without displaying any error
3114 Details : When multiple mount protection fails during remount, proper error
3117 Severity : enhancement
3119 Description: Allow stripe size to be up to 4G-64k
3120 Details : Fix math logic to allow large stripe sizes.
3124 Description: add check for >8TB ldiskfs filesystems
3125 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3126 Don't allow >8TB ldiskfs filesystems to be mounted without
3127 force_over_8tb mount option
3132 Descriptoin: OSS complains "too many blocks" and crashes
3133 Details : In filter_commitrw_write():
3134 1. lquota_chkquota();
3135 2. obdo_from_inode();
3136 3. lquota_pending_commit();
3137 "2." has changed uid and gid in this case, then caused
3138 lquota_chkquota() and lquota_pending_commit() to use
3139 different uid and gid. That is the root of the bug.
3143 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3144 Details : Detect and print out the 'stripe count = -1' case properly.
3148 Description: Out or order replies might be lost on replay
3149 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3150 request with tid smaller than one currently being added, add it
3151 to the start, not end of the list.
3153 --------------------------------------------------------------------------------
3155 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3157 * Support for kernels:
3158 2.6.5-7.283 (SLES 9),
3159 2.6.9-55.EL (RHEL 4),
3160 2.6.16.46-0.14 (SLES 10),
3161 2.6.18.8 vanilla (kernel.org)
3162 * Client support for unpatched kernels:
3163 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3164 2.6.16 - 2.6.22 vanilla (kernel.org)
3165 * Due to recently discovered recovery problems, we do not recommend
3166 using patchless RHEL 4 clients with this or any earlier release.
3167 * Recommended e2fsprogs version: 1.39.cfs8
3168 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3169 * Starting with this release, the ldiskfs backing filesystem required
3170 by Lustre is now in its own package, lustre-ldiskfs. This package
3171 should be installed. It is versioned separately from Lustre and
3172 may be released separately in future.
3176 Description: block reactivating mgc import until all deactivates complete
3177 Details : Fix race when failing back MDT/MGS to itself (testing)
3180 Frequency : at statup only
3182 Description: mds_lov_synchronize race leads to various problems
3183 Details : simultaneous MDT->OST connections at startup can cause the
3184 sync to abort, leaving the OSC in a bad state.
3186 Severity : enhancement
3188 Description: add optional extra BUILD_VERSION info
3189 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3190 to override the lustre version.
3193 Frequency : 2.6.18 servers only
3195 Description: ll_kern_mount() doesn't release the module reference
3196 Details : The ldiskfs module reference count never drops down to 0
3197 because ll_kern_mount() doesn't release the module reference.
3202 Description: server LBUG when using old ost_num_threads parameter
3203 Details : Accept the old ost_num_threads parameter but warn that it
3204 is deprecated, and fix an off-by-one error that caused an LBUG.
3209 Description: Transient SCSI error results in persistent IO issue
3210 Details : iobuf->dr_error is not reinitialized to 0 between two
3214 Frequency : sometimes when underlying device returns I/O errors
3216 Description: OSTs not going read-only during write failures
3217 Details : OSTs are not remounted read-only when the journal commit threads
3218 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3219 instead of the ext3 wrappers.
3223 Description: poor connect scaling with increasing client count
3224 Details : Don't run filter_grant_sanity_check for more than 100 exports
3225 to improve scaling for large numbers of clients.
3228 Frequency : SLES10 only
3230 Description: sanity-quota.sh quotacheck failed: rc = -22
3231 Details : Quotas cannot be enabled on SLES10.
3234 Frequency : liblustre clients only
3236 Description: getdirentries does not give error when run on compute nodes
3237 Details : getdirentries does not fail when the size specified as an argument
3238 is too small to contain at least one entry
3240 Severity : enhancement
3242 Description: Add LNET router traceability for debug purposes
3243 Details : If a checksum failure occurs with a router as part of the
3244 IO path, the NID of the last router that forwarded the bulk data
3245 is printed so it can be identified.
3250 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3251 Details : Due to a race condition, liblustre clients were occasionally
3252 evicted incorrectly.
3254 Severity : enhancement
3256 Description: lfs setstripe use optional parameters instead of postional
3259 Severity : enhancement
3261 Description: Nanosecond timestamp support for ldiskfs
3262 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3263 resolution timestamps. There is not yet support for this at
3264 the Lustre filesystem level.
3267 Frequency : during server recovery
3269 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3270 Details : request with rq_no_resend flag not awake l_wait_event if they get
3274 Frequency : nfs export on patchless client
3276 Description: connectathon hang when test nfs export over patchless client
3277 Details : Disconnected dentry cannot be found with lookup, so we do not need
3278 to unhash it or make it invalid
3281 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3282 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3283 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3288 Description: open req refcounting wrong on reconnect
3289 Details : If reconnect happened between getting open reply from server and
3290 call to mdc_set_replay_data in ll_file_open, we will schedule
3291 replay for unreferenced request that we are about to free.
3292 Subsequent close will crash in variety of ways.
3293 Check that request is still eligible for replay in
3294 mdc_set_replay_data().
3299 Description: disable writes to filesystem when reading health_check file
3300 Details : the default for reading the health_check proc file has changed
3301 to NOT do a journal transaction and write to disk, because this
3302 can cause reads of the /proc file to hang and block HA state
3303 checking on a healthy but otherwise heavily loaded system. It
3304 is possible to return to the previous behaviour during configure
3305 with --enable-health-write.
3307 Severity : enhancement
3309 Description: 64-bit inode version
3310 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3311 to the inode. This will be required for version-based recovery.
3316 Description: MDS fails to start if a duplicate client export is detected
3317 Details : in some rare cases it was possible for a client to connect to
3318 an MDS multiple times. Upon recovery the MDS would detect this
3319 and fail during startup. Handle this more gracefully.
3321 Severity : enhancement
3323 Description: Add -o localflock option to simulate old noflock
3325 Details : This will achieve local-only flock/fcntl locks
3331 Description: log_commit_thread vs filter_destroy race leads to crash
3332 Details : Take import reference before releasing llog record semaphore
3337 Description: Wrong request locking in request set processing
3338 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3339 imp_delayed_list, in this place should be used imp_lock.
3342 Frequency : when reconnection
3344 Description: Grant Leak when osc reconnect to OST
3345 Details : When osc reconnect ost, OST(filter) should check whether it
3346 should grant more space to client by comparing fed_grant and
3347 cl_avail_grant, and return the granted space to client instead
3348 of "new granted" space, because client will call osc_init_grant
3349 to update the client grant space info.
3352 Frequency : when client reconnect to OST
3354 Description: Grant Leak when osc do resend and replay bulk write
3355 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3356 bulk write request, because the grant info while be sync between
3357 OSC and OST when reconnect, and we should ignore the grant info
3358 these of resend/replay write req.
3363 Description: Grant space more than available left space sometimes.
3364 Details : When then OST is about to be full, if two bulk writing from
3365 different clients came to OST. Accord the avaliable space of the
3366 OST, the first req should be permitted, and the second one
3367 should be denied by ENOSPC. But if the seconde arrived before
3368 the first one is committed. The OST might wrongly permit second
3369 writing, which will cause grant space > available space.
3372 Frequency : when client is evicted
3374 Description: Grant might be wrongly erased when osc is evicted by OST
3375 Details : when the import is evicted by server, it will fork another
3376 thread ptlrpc_invalidate_import_thread to invalidate the
3377 import, where the grant will be set to 0. While the original
3378 thread will update the grant it got when connecting. So if
3379 the former happened latter, the grant will be wrongly errased
3380 because of this race.
3385 Description: Checking Stale with correct fid
3386 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3387 to check whether it is stale, because sometimes, we want the
3388 enqueue happened anyway, and op_data.fid2 will not be initialized.
3390 Severity : enhancement
3392 Description: update patchless client
3393 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3396 Frequency : only with 2.4 kernel
3398 Description: random memory corruption
3399 Details : size of struct ll_inode_info is to big for union inode.u and this
3400 can be cause of random memory corruption.
3405 Description: Memory leak in recovery
3406 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3407 It should also check obd_fail before fsfilt_start, otherwise if
3408 fsfilt_start return -EROFS,(failover mds during mds recovery).
3409 then the req will return with repmsg->transno = 0 and rc = EROFS.
3410 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3411 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3412 be freed no matter whether fsfilt_commit success or not.
3415 Frequency : only with huge count clients
3417 Description: Prevents from taking the superblock lock in llap_from_page for
3419 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3420 for llap_from_page call in ll_removepage prevents from taking the
3421 superblock lock for a soon died page.
3426 Description: Not check open intent error before release open handle
3427 Details : in some rare cases, the open intent error is not checked before
3428 release open handle, which may cause
3429 ASSERTION(open_req->rq_transno != 0), because it tries to release
3430 the failed open handle.
3435 Description: Set cat log bitmap only after create log success.
3436 Details : in some rare cases, the cat log bitmap is set too early. and it
3437 should be set only after create log success.
3441 Description: Accessing a block bevice can re-enable I/O when Lustre is
3442 tearing down a device.
3443 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3448 Description: service threads may hog cpus when there are a lot of requests
3449 Details : Insert cond_resched to give other threads a chance to use some CPU
3454 Description: the cat log was not initialized in recovery
3455 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3456 unlink log on mds will not be initialized until mds post
3457 recovery. And also in mds post recovery, the unlink log will
3458 initialization will be done asynchronausly, so there will be race
3459 between add unlink log and unlink log initialization.
3463 Description: brw_stats were being printed incorrectly
3464 Details : brw_stats were being printed as log2 but all of them were not
3465 recorded as log2. Also remove some code duplication arising from
3466 filter_tally_{read,write}.
3470 Frequency : rare, only in recovery.
3471 Description: ASSERTION(req->rq_type != LI_POISON) failed
3472 Details : imp_lock should be held while iterating over imp_sending_list for
3473 prevent destroy request after get timeout in ptlrpc_queue_wait.
3477 Description: replay-single.sh test 52 fails
3478 Details : A lock's skiplist need to be cleanup when it being unlinked
3479 from its resource list.
3483 Description: Short directio read returns full requested size rather than
3485 Details : Direct I/O operations should return actual amount of bytes
3486 transferred rather than requested size.
3488 Severity : enhancement
3490 Description: metadata RPC reduction (e.g. for rm performance)
3491 Details : decrease the amount of synchronous RPC between clients and servers
3492 by canceling conflicing lock before the operation on the client
3493 and packing thier handles into the main operation RPC to server.
3495 Severity : enhancement
3497 Description: Async OSC create to avoid the blocking unnecessarily.
3498 Details : If an OST has no remaining object, system will block on the
3499 creation when it needs to create a new object on this OST. Now,
3500 ways use pre-created objects when available, instead of blocking on
3501 an empty osc while others are not empty. If we must block, we
3502 block for the shortest possible period of time.
3506 Description: improve handling recoverable errors
3507 Details : if request processig with error which can be recoverable on server
3508 request should be resend, otherwise page released from cache and
3511 Severity : enhancement
3513 Description: refine locking for avoid write wrong info into lov_objid file
3514 Details : fix possible races with add new target and write/update data in
3517 --------------------------------------------------------------------------------
3519 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3524 Frequency : on some architectures
3526 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3527 Details : Uninitialized flags sometimes cause configuration commands to
3531 Frequency : patchless clients only
3533 Description: missing __iget() symbol export
3534 Details : The __iget() symbol export is missing. To avoid the need for
3535 this on patchless clients the deathrow inode reaper is turned
3536 off, and we depend on the VM to clean up old inodes. This
3537 dependency was during via the fix for bug 12181.
3539 --------------------------------------------------------------------------------
3541 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3543 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3544 INTEROPERATE with older versions automatically. In many cases a
3545 special upgrade step is needed. Please read the
3546 user documentation before upgrading any part of a live system.
3547 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3548 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3549 * WARNING: Lustre configuration and startup changes are required with
3550 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3554 Severity : enhancement
3556 Description: MountConf
3557 Details : Lustre configuration is now managed via mkfs and mount
3558 commands instead of lmc and lconf. New obd types (MGS, MGC)
3559 are added for dynamic configuration management. See
3560 https://mail.clusterfs.com/wikis/lustre/MountConf for
3563 Severity : enhancement
3565 Description: dynamic OST addition
3566 Details : OSTs can now be added to a live filesystem
3568 Severity : enhancement
3570 Description: startup order invariance
3571 Details : MDTs and OSTs can be started in any order. Clients only
3572 require the MDT to complete startup.
3574 Severity : enhancement
3576 Description: parallel, asynchronous orphan cleanup
3577 Details : orphan cleanup is now performed in separate threads for each
3578 OST, allowing parallel non-blocking operation.
3580 Severity : enhancement
3582 Description: optimized stripe assignment
3583 Details : stripe assignments are now made based on ost space available,
3584 ost previous usage, and OSS previous usage, in order to try
3585 to optimize storage space and networking resources.
3587 Severity : enhancement
3589 Description: Permanently set tunables
3590 Details : All writable /proc/fs/lustre tunables can now be permanently
3591 set on a per-server basis, at mkfs time or on a live system.
3593 Severity : enhancement
3595 Description: Lustre message v2
3596 Details : Add lustre message format v2.
3598 Severity : enhancement
3600 Description: client OST exclusion list
3601 Details : Clients can be started with a list of OSTs that should be
3602 declared "inactive" for known non-responsive OSTs.
3606 Description: SPEC SFS validation failure on NFS v2 over lustre.
3607 Details : Changes the blocksize for regular files to be 2x RPC size,
3608 and not depend on stripe size.
3610 Severity : enhancement
3612 Description: Multiple MD RPCs in flight.
3613 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3614 To avoid overly-overloading MDS, introduce a limit on number of
3615 MDS RPCs in flight for a single client and add /proc controls
3616 to adjust this limit.
3618 Severity : enhancement
3620 Description: client read/write statistics
3621 Details : Add client read/write call usage stats for performance
3622 analysis of user processes.
3623 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3624 file access. extents_stats shows chunk size distribution.
3625 extents_stats_per_process show chunk size distribution per
3628 Severity : enhancement
3630 Description: mds statistics
3631 Details : Add detailed mds operations statistics in
3632 /proc/fs/lustre/mds/*/stats.
3636 Description: Failure of copying files with lustre special EAs.
3637 Details : Client side always return success for setxattr call for lustre
3638 special xattr (currently only "trusted.lov").
3643 Description: Refcount LNET uuids
3644 Details : The global LNET uuid list grew linearly with every startup;
3645 refcount repeated list entries instead of always adding to
3648 Severity : enhancement
3650 Description: Dynamic service threads
3651 Details : Within a small range, start extra service threads
3652 automatically when the request queue builds up.
3655 Frequency : mixed-endian client/server environments
3657 Description: mixed-endian crashes
3658 Details : The new msg_v2 system had some failures in mixed-endian
3661 Severity : enhancement
3663 Description: Easy OST removal
3664 Details : OSTs can be permanently deactivated with e.g. 'lctl
3665 conf_param lustre-OST0001.osc.active=0'
3667 Severity : enhancement
3669 Description: MGS proc entries
3670 Details : Added basic proc entries for the MGS showing what filesystems
3673 Severity : enhancement
3675 Description: provide MGS failover
3676 Details : Added config lock reacquisition after MGS server failover.
3678 Severity : enhancement
3680 Description: add Linux 2.4 support
3681 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3685 Description: a large application tries to do I/O to the same resource and dies
3686 in the middle of it.
3687 Details : Check the req->rq_arrival time after the call to
3688 ost_brw_lock_get(), but before we do anything about
3689 processing it & sending the BULK transfer request. This
3690 should help move old stale pending locks off the queue as
3691 quickly as obd_timeout.
3694 Frequency : when an incorrect nid is specified during startup
3696 Description: ptlrpc connect to non-existant node causes kernel crash
3697 Details : LNET can't be re-entered from an event callback, which
3698 happened when we expire a message after the export has been
3699 cleaned up. Instead, hand the zombie cleanup off to another
3702 Severity : enhancement
3704 Description: plain/inodebits lock performance improvement
3705 Details : Grouping plain/inodebits in granted list by their request modes
3706 and bits policy, thus improving the performance of search through
3710 Frequency : only if OST filesystem is corrupted
3712 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3713 Details : for a short time RPCs with bulk IO are in the replay list,
3714 but replay of bulk IOs is unimplemented. If the OST filesystem
3715 is corrupted due to disk cache incoherency and then replay is
3716 started it is possible to trip an assertion. Avoid putting
3717 committed RPCs into the replay list at all to avoid this issue.
3720 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3722 Description: System hang on startup
3723 Details : This bug allowed the liblustre (e.g. catamount) client to
3724 return to the app before handling all startup RPCs. This
3725 could leave the node unresponsive to lustre network traffic
3726 and manifested as a server ptllnd timeout.
3728 Severity : enhancement
3730 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3731 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3732 boolean to control whether to print peer debug info when a
3733 client's RPC times out.
3736 Frequency : only for kernels with patches from Lustre below 1.4.3
3738 Description: Remove old rdonly API
3739 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3742 Frequency : only for devices with external journals
3744 Description: Set external device read-only also
3745 Details : During a commanded failover stop, we set the disk device
3746 read-only while the server shuts down. We now also set any
3747 external journal device read-only at the same time.
3750 Frequency : when upgrading from 1.4 while trying to change parameters
3752 Description: The wrong (new) MDC name was used when setting parameters for
3753 upgraded MDT's. Also allows changing of OSC (and MDC)
3754 parameters if --writeconf is specified at tunefs upgrade time.
3757 Frequency : when setting specific ost indicies
3759 Description: QOS code breaks on skipped indicies
3760 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3761 created with --index need not be sequential.
3763 Severity : enhancement
3765 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3766 Details : The uninit_groups feature works in conjunction with the kernel
3767 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3768 up the pass1 processing of e2fsck. This is a read-only feature
3769 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3770 mount filesystems that have had this feature enabled.
3772 Severity : enhancement
3774 Description: Improve multi-block allocation algorithm to avoid fragmentation
3775 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3776 allocation locality and avoid filesystem fragmentation.
3778 ------------------------------------------------------------------------------
3780 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3782 * Support for kernels:
3783 2.4.21-47.0.1.EL (RHEL 3)
3784 2.6.5-7.283 (SLES 9)
3785 2.6.9-42.0.10.EL (RHEL 4)
3786 2.6.12.6 vanilla (kernel.org)
3787 2.6.16.27-0.9 (SLES 10)
3788 * Recommended e2fsprogs version: 1.39.cfs5
3790 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3794 Frequency : occasional, depends on client load and configuration
3795 Bugzilla : 12181, 12203
3796 Description: data loss for recently-modified files
3798 Details : In some cases it is possible that recently written or created
3799 files may not be written to disk in a timely manner (this should
3800 normally be within 30s unless client IO load is very high).
3801 The problem appears as zero-length files or files that are a
3802 multiple of 1MB in size after a client crash or client eviction
3803 that are missing data at the end of the file.
3805 This problem is more likely to be hit on clients where files are
3806 repeatedly created and unlinked in the same directory, clients
3807 have a large amount of RAM, have many CPUs, the filesystem has
3808 many OSTs, the clients are rebooted frequently, and/or the files
3809 are not accessed by other nodes after being written.
3811 The presence of the problem can be detected by looking at
3812 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3813 smaller than the second (nr_unused) then dirty files will not
3814 be flushed automatically to disk. "sync; sleep 10" should be
3815 run several times on the node before unmounting it to update
3816 Lustre (this is also safe to run on nodes without this problem).
3818 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3819 that can cause this same problem, so customers using that kernel
3820 also need to update the kernel in addition to Lustre. In order
3821 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3823 It is normal that files written just before a client crash (less
3824 than 30s) may not yet have been flushed to disk, even for local
3828 Frequency : frequent on thin XT3 nodes
3830 Description: UUID collision on thin XT3 Linux nodes
3831 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3832 randomly, since we relied on an insufficiently-seeded PRNG.
3837 Description: OSS hangs after "All ost request buffers busy"
3838 Details : A deadlock between quota and journal operations caused OSS
3839 hangs after printing "All ost request buffers busy."
3842 Frequency : always on liblustre builds
3844 Description: Cleanup compiler warnings on liblustre
3847 Frequency : always on liblustre builds on XT3
3849 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3853 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3854 > 32000 subdirectories
3855 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3856 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3857 subdirectory count crosses 32000. This will aid e2fsck to
3858 correctly handle more than 32000 subdirectories.
3861 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3863 Description: System hang on startup
3864 Details : This bug allowed the liblustre (e.g. catamount) client to
3865 return to the app before handling all startup RPCs. This
3866 could leave the node unresponsive to lustre network traffic
3867 and manifested as a server ptllnd timeout.
3869 Severity : enhancement
3871 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3872 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3873 boolean to control whether to print peer debug info when a
3874 client's RPC times out.
3879 Description: make O_SYNC working on 2.6 kernels
3880 Details : 2.6 kernels use different method for mark pages for write,
3881 so need add a code to lustre for O_SYNC work.
3886 Description: Failure to close file and release space on NFS
3887 Details : Put inode details into lock acquired in ll_intent_file_open.
3888 Use mdc_intent_lock in ll_intent_open to properly
3889 detect all kind of errors unhandled by mdc_enqueue.
3894 Description: proc file read during shutdown sometimes raced obd removal,
3896 Details : Add lock to prevent obd access after proc file removal.
3899 Frequency : Only for files larger than 4GB on 32-bit clients.
3901 Description: improperly doing page alignment of locks
3902 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3903 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3908 Description: RPCs being resent when they shouldn't be
3909 Details : Some RPCs that should not be resent are being resent. This
3910 can cause inconsistencies in the RPC state machine. Do not
3911 resend such requests.
3914 Frequency : rare, only with NFS export
3916 Description: Crash on NFS re-export node
3917 Details : under very unusual load conditions an assertion is hit in
3918 ll_intent_file_open()
3921 Frequency : only if OST filesystem is corrupted
3923 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3924 Details : for a short time RPCs with bulk IO are in the replay list,
3925 but replay of bulk IOs is unimplemented. If the OST filesystem
3926 is corrupted due to disk cache incoherency and then replay is
3927 started it is possible to trip an assertion. Avoid putting
3928 committed RPCs into the replay list at all to avoid this issue.
3933 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3934 Details : Large single O_DIRECT read and write calls can fail to allocate
3935 a sufficiently large buffer to process the request. In case of
3936 allocation failure the allocation is retried with a smaller
3937 buffer and broken into smaller requests.
3939 Severity : enhancement
3941 Description: Add -o localflock option to simulate old noflock behaviour.
3942 Details : This will achieve local-only flock/fcntl locks coherentness.
3947 Description: versioning check is incomplete
3948 Details : Checking the version difference of client vs. server, report
3949 error if the gap is too big.
3954 Description: add support PG_writeback bit
3955 Details : add support for PG_writeback bit for Lustre, for more carefull
3956 work with page cache in 2.6 kernel. This also fix some deadlocks
3957 and remove hack for work O_SYNC with 2.6 kernel.
3959 Severity : enhancement
3961 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3962 Details : The uninit_groups feature works in conjunction with the kernel
3963 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3964 up the pass1 processing of e2fsck. This is a read-only feature
3965 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3966 mount filesystems that have had this feature enabled.
3968 Severity : enhancement
3970 Description: Improve multi-block allocation algorithm to avoid fragmentation
3971 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3972 allocation locality and avoid filesystem fragmentation.
3974 ------------------------------------------------------------------------------
3976 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3978 * Support for kernels:
3979 2.6.9-42.0.3.EL (RHEL 4)
3980 2.6.5-7.276 (SLES 9)
3981 2.4.21-47.0.1.EL (RHEL 3)
3982 2.6.12.6 vanilla (kernel.org)
3983 2.6.16.21-0.8 (SLES10)
3984 * Recommended e2fsprogs version: 1.39.cfs2-0
3986 * The backwards-compatible /proc/sys/portals symlink has been removed
3987 in this release. Before upgrading, please ensure that you change
3988 any configuration scripts or /etc/sysctl.conf files that access
3989 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3990 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3991 in advance of the upgrade on any system running Lustre 1.4.6 or
3992 newer, since /proc/sys/lnet was added in that version.
3993 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3997 Frequency : only when quota is used
3999 Description: avoid scanning export list for quota master
4000 Details : Change the algorithms to avoid scanning export list in order
4001 to improve the efficiency.
4004 Frequency : MDS failover only, very rarely
4006 Description: "went back in time" messages on mds failover
4007 Details : The greatest transno may be lost when the current operation
4008 finishes with an error (transno==0) and the client's last_rcvd
4009 record is over-written. Save the greatest transno in the
4010 mds_last_transno for this case.
4013 Frequency : always for specific kernels and striping counts
4015 Description: client may get "Matching packet too big" without ACL support
4016 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4017 when trying to access files in certain configurations. The
4018 clients should in fact be denied when mounting because they do
4019 not understand ACLs.
4022 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4024 Description: many clients connecting with IO in progress causes connect timeouts
4025 Details : Avoid synchronous journal commits to avoid delays caused by many
4026 clients connecting/disconnecting when bulk IO is in progress.
4027 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4028 OST_IO_PORTAL to avoid delays behind potentially many pending
4032 Frequency : occasionally with multiple writers to a single file
4034 Description: shared writes to file may result in wrong size reported by stat()
4035 Details : Allow growing of kms when extent lock is cancelled
4038 Frequency : always with random mmap IO to multi-striped file
4040 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4041 Details : Only if the hole is at the end of OST object so that kms is too
4042 small. Fix is to increase kms accordingly in ll_nopage.
4045 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4047 Description: writes to a missing object would leak memory on the OST
4048 Details : If there is an inconsistency between the MDS and OST filesystems,
4049 such that the MDS references an object that doesn't exist, writes
4050 to that object will leak memory due to incorrect cleanup in the
4051 error handling path, eventually running out of memory on the OST.
4056 Description: Creating too long symlink causes lustre errors
4057 Details : Check symlink and name lengths before sending requests to MDS.
4060 Frequency : only if flock is enabled (not on by default)
4062 Description: posix locks not released on fd closure on 2.6.9+
4063 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4064 kernels, this caused such locks not to be released on fd close and
4065 then assertions on fs unmount about still used locks.
4068 Frequency : MDS failover only, very rarely
4070 Description: clients may get ASSERTION(granted_lock != NULL)
4071 Details : When request was taking a long time, and a client was resending
4072 a getattr by name lock request. The were multiple lock requests
4073 with the same client lock handle and
4074 mds_getattr_name->fixup_handle_for_resent_request found one of the
4075 lock handles but later failed with ASSERTION(granted_lock != NULL).
4080 Description: handle->h_buffer_credits > 0, assertion failure
4081 Details : h_buffer_credits is zero after truncate, causing assertion
4082 failure. This patch extends the transaction or creates a new
4086 Frequency : NFS re-export or patchless client
4087 Bugzilla : 11179, 10796
4088 Description: Crash on NFS re-export node (__d_move)
4089 Details : We do not want to hash the dentry if we don't have a lock.
4090 But if this dentry is later used in d_move, we'd hit uninitialised
4091 list head d_hash, so we just do this to init d_hash field but
4092 leave dentry unhashed.
4095 Frequency : NFS re-export or patchless client
4097 Description: NFS exports has problem with symbolic link
4098 Details : lustre client didn't properly install dentry when re-exported
4099 to NFS or running patchless client.
4102 Frequency : NFS re-export or patchless client
4104 Description: Various nfs/patchless fixes.
4105 Details : fixes reuse disconected alias for lookup process - this fixes
4106 warning "find_exported_dentry: npd != pd",
4107 fix permission error with open files at nfs.
4108 fix apply umask when do revalidate.
4111 Frequency : occasional
4113 Description: Crash on NFS re-export node
4114 Details : calling clear_page() on the wrong pointer triggered oops in
4115 generic_mapping_read().
4118 Frequency : rarely, using O_DIRECT IO
4120 Description: unaligned directio crashes client with LASSERT
4121 Details : check for unaligned buffers before trying any requests.
4124 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4126 Description: stale data returned from RAID cache
4127 Details : If only a small amount of IO is done to the RAID device before
4128 reading it again it is possible to get stale data from the RAID
4129 cache instead of reading it from disk.
4132 Frequency : always for sles10 kernel
4134 Description: sles10 support
4135 Details : ll_follow_link: compile fixes and using of nd_set_link
4136 under newer kernels.
4139 Frequency : depends on arch, kernel and compiler version, always on sles10
4142 Description: recursive or deep enough symlinks cause stack overflow
4143 Details : getting rid of large stack-allocated variable in
4147 Frequency : depends on hardware
4149 Description: lustre write performance loss in the SLES10 kernel
4150 Details : the performance loss is caused by using of write barriers in the
4151 ext3 code. The SLES10 kernel turns barrier support on by
4152 default. The fix is to undo that change for ldiskfs.
4154 ------------------------------------------------------------------------------
4156 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4158 * Support for kernels:
4159 2.6.9-42.0.3EL (RHEL 4)
4160 2.6.5-7.276 (SLES 9)
4161 2.4.21-47.0.1.EL (RHEL 3)
4162 2.6.12.6 vanilla (kernel.org)
4166 Frequency : quota enabled and large files being deleted
4168 Description: releasing more than 4GB of quota at once hangs OST
4169 Details : If a user deletes more than 4GB of files on a single OST it
4170 will cause the OST to spin in an infinite loop. Release
4171 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4176 Description: statfs data retrieved from /proc may be stale or zero
4177 Details : When reading per-device statfs data from /proc, in the
4178 {kbytes,files}_{total,free,avail} files, it may appear
4179 as zero or be out of date.
4182 Frequency : systems with MD RAID1 external journal devices
4184 Description: lconf's call to blkid is confused by RAID1 journal devices
4185 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4186 of returning all block devices that match the journal UUID.
4189 Frequency : always, for aggregate stripe size over 4GB
4191 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4192 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4193 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4196 Frequency : always if "lfs find" used on a local file/directory
4198 Description: "lfs find" segfaults if used on a local file/directory
4199 Details : The case where a directory component was not specified wasn't
4200 handled correctly. Handle this properly.
4203 Frequency : always on ppc64
4205 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4206 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4207 perform endianness conversion.
4210 Frequency : rarely (truncate to non-zero file size after write under load)
4211 Bugzilla : 10730, 10687
4212 Description: Files padded with zeros to next 4K multiple
4213 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4214 it is possible that files that are truncated to a non-zero size
4215 immediately after being written are filled with zero bytes beyond
4216 the truncated size. No file data is lost.
4218 Severity : enhancement
4220 Description: Allow recovery/failover for liblustre clients.
4221 Details : liblustre clients were unaware of failover configurations until
4224 Severity : enhancement
4226 Description: user file locks should fail when not mounting with flock option
4227 Details : Set up an error-returning stub in ll_file_operations.lock field
4228 to prevent incorrect behaviour when client is mounted without
4229 flock option. Also, set up properly f_op->flock field for
4233 Frequency : always on ia64
4235 Description: "lfs df" loops on printing out MDS statfs information
4236 Details : The obd_ioctl_data was not initialized and in some systems
4237 this caused a failure during the ioctl that did not return
4238 an error. Initialize the struct and return an error on failure.
4241 Frequency : SLES 9 only
4243 Description: Error of copying files with lustre special EAs as root
4244 Details : Client side always return success for setxattr call for lustre
4245 special xattr (currently only "trusted.lov").
4248 Frequency : rarely on clusters with both ia64+i386 clients
4250 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4251 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4252 clients are doing concurrent writes to the same file it is
4253 possible that the ia64 clients may LASSERT because the OST
4254 extent locks are not PAGE_SIZE aligned. Ensure that grown
4255 locks are always aligned on the request boundary.
4258 Frequency : specific use, occasional
4260 Description: Overwriting in use executable truncates on-disk binary image
4261 Details : If one node attempts to overwrite an executable in use by
4262 another node, we now correctly return ETXTBSY instead of
4263 truncating the file.
4265 Severity : enhancement
4267 Description: Async OSC create to avoid the blocking unnecessarily.
4268 Details : If an OST has no remaining object, system will block on the creating
4269 when need to create a new object on this OST. Now, ways use
4270 pre-created objects when available, instead of blocking on an
4271 empty osc while others are not empty. If we must block, we block
4272 for the shortest possible period of time.
4277 Description: chmod on Lustre root is propagated to other clients
4278 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4279 invalid by the follow_mount time.
4282 Frequency : liblustre clients only
4284 Description: Race in 'instant cancel' lock handling could lead to such locks
4285 never to be granted in case of SMP MDS
4286 Details : Do not destroy not yet granted but cbpending locks in
4290 Frequency : replay/resend of open
4292 Description: non null lock assetion failure in mds_intent_policy
4293 Details : Trying to replay/resend lockless open requests resulted in
4294 mds_open() returning 0 with no lock. Now it sets a flag if
4295 a lock is going to be returned.
4297 Severity : enhancement
4299 Description: Checksum enhancements
4300 Details : New checksum enhancements allow for resending RPCs that failed
4303 Severity : enhancement
4305 Description: Tunables on number of dirty pages in cacche
4306 Details : Allow to set limit on number of dirty pages cached.
4311 Description: client crash on unmount - lock still has references
4312 Details : In some error handling cases it was possible to leak a lock
4313 reference on a client while accessing a file. This was not
4314 harmful to the client during operation, but would cause the
4315 client to crash when the filesystem is unmounted.
4318 Frequency : specific case, rare
4320 Description: ETXTBSY on mds though file not in use by client
4321 Details : ETXTBSY is no longer incorrectly returned when attempting to
4322 chmod or chown a directory that the user previously tried to
4323 execute or a currently-executing binary.
4326 Frequency : extremely rare except on liblustre-based clients
4328 Description: Lustre space not freed when files are deleted
4329 Details : Clean up open-unlinked files after client eviction. Previously
4330 the unlink was skipped and the files remained as orphans.
4335 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4336 Details : In some cases it was possible to send a blocking callback to
4337 a client doing a glimpse, even though that client didn't get
4338 a lock granted. When the glimpse lock is cancelled on the OST
4339 the freed lock is left on the waiting list and corrupted the list.
4342 Frequency : all core dumps
4344 Description: Broke core dumps to lustre
4345 Details : Negative dentry may be unhashed if parent does not have UPDATE
4346 lock, but some callers, e.g. do_coredump, expect dentry to be
4347 hashed after successful create, hash it in ll_create_it.
4349 ------------------------------------------------------------------------------
4351 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4353 * Support for kernels:
4354 2.6.9-42.0.2.EL (RHEL 4)
4355 2.6.5-7.276 (SLES 9)
4356 2.4.21-40.EL (RHEL 3)
4357 2.6.12.6 vanilla (kernel.org)
4361 Frequency : always on RHEL 3
4363 Description: Number of open files grows over time
4364 Details : The number of open files grows over time, whether or not
4365 Lustre is started. This was due to a filp leak introduced
4366 by one of our kernel patches.
4368 ------------------------------------------------------------------------------
4370 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4372 * Support for kernels:
4373 2.6.9-42.EL (RHEL 4)
4374 2.6.5-7.276 (SLES 9)
4375 2.4.21-40.EL (RHEL 3)
4376 2.6.12.6 vanilla (kernel.org)
4381 Bugzilla : 5719, 9635, 9792, 9684
4382 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4383 Details : If a server is under heavy load and cannot reply to new
4384 connection requests before the client resends the (re)connect,
4385 the connection handling code can behave badly if two service
4386 threads are concurrently handing separate (re)connections from
4387 the same client. Add better locking to the connection handling
4388 code, and ensure that only a single connection will be processed
4389 for a given client UUID, even if the lock is dropped.
4391 Severity : enhancement
4393 Description: add TCP zero-copy support to kernel
4394 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4395 sends if the hardware supports scatter-gather and checksumming.
4396 This allows socklnd to do client-write and server-read more
4397 efficiently and reduce CPU utilization from skbuf copying.
4400 Frequency : only if NFS exporting from client
4402 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4403 Details : If setting attributes on a file created under NFS that had
4404 never been opened it would be possible to oops the client
4405 if the file had no objects.
4408 Frequency : always for liblustre
4410 Description: liblustre client does MDS+OSTs setattr RPC for each write
4411 Details : When doing a write from a liblustre client, the client
4412 incorrectly issued an RPC to the MDS and each OST the file was
4413 striped over in order to update the timestamps. When writing
4414 with small chunks and many clients this could overwhelm the MDS
4415 with RPCs. In all cases it would slow down the write because
4416 these RPCs are unnecessary.
4418 Severity : enhancement
4420 Description: allow number of MDS service threads to be changed at module load
4421 Details : It is now possible to change the number of MDS service threads
4422 running. Adding "options mds mds_num_threads={N}" to the MDS's
4423 /etc/modprobe.conf will set the number of threads for the next
4424 time Lustre is restarted (assuming the "mds" module is also
4425 reloaded at that time). The default number of threads will
4426 stay the same, 32 for most systems.
4431 Description: OST crash if filesystem is unformatted or corrupt
4432 Details : If an OST is started on a device that has never been formatted
4433 or if the filesystem is corrupt and cannot even mount then the
4434 error handling cleanup routines would dereference a NULL pointer.
4439 Description: NULL pointer deref in llap_from_page.
4440 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4441 incorrect) mapping if the page was truncated/reclaimed while it was
4442 searched for. Check for this condition and skip such pages when
4443 doing readahead. Introduce extra check to llap_from_page() to
4444 verify page->mapping->host is non-NULL (so page is not anonymous).
4447 Frequency : Sometimes when using sys_sendfile
4449 Description: "page not covered by a lock" warnings from ll_readpage
4450 Details : sendfile called ll_readpage without right page locks present.
4451 Now we introduced ll_file_sendfile that does necessary locking
4452 around call to generic_file_sendfile() much like we do in
4456 Frequency : with certain MDS communication failures at client mount time
4458 Description: NULL pointer deref after failed client mount
4459 Details : a client connection request may delayed by the network layer
4460 and not be sent until after the PTLRPC layer has timed out the
4461 request. If the client fails the mount immediately it will try
4462 to clean up before the network times out the request. Add a
4463 reference from the request import to the obd device and delay
4464 the cleanup until the network drops the request.
4467 Frequency : occasionally during client (re)connect
4469 Description: assertion failure during client (re)connect
4470 Details : processing a client connection request may be delayed by the
4471 client or server longer than the client connect timeout. This
4472 causes the client to resend the connection request. If the
4473 original connection request is replied in this interval, the
4474 client may trip an assertion failure in ptlrpc_connect_interpret()
4475 which thought it would be the only running connect process.
4478 Frequency : only with obd_echo servers and clients that are rebooted
4480 Description: kernel BUG accessing uninitialized data structure
4481 Details : When running an obd_echo server it did not start the ping_evictor
4482 thread, and when a client was evicted an uninitialized data
4483 structure was accessed. Start the ping_evictor in the RPC
4484 service startup instead of the OBD startup.
4486 Severity : enhancement
4487 Bugzilla : 10193 (patchless)
4488 Description: Remove dependency on various unexported kernel interfaces.
4489 Details : No longer need reparent_to_init, exit_mm, exit_files,
4490 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4493 Frequency : rare (only users of deprecated and unsupported LDAP config)
4495 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4496 Details : LDAP apparently contains 'lustreName' attributes instead of
4497 'name'. A simple remapping of the name is sufficient.
4500 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4502 Description: waiting_locks_callback trips kernel BUG if client is evicted
4503 Details : Running with the dump_on_timeout debug flag turned on makes
4504 it possible that the waiting_locks_callback() can try to dump
4505 the Lustre kernel debug logs from an interrupt handler. Defer
4506 this log dumping to the expired_lock_main() thread.
4508 Severity : enhancement
4510 Description: Support NFS exporting on 2.6 kernels.
4511 Details : Implement non-rawops metadata methods for NFS server to use without
4512 changing NFS server code.
4515 Frequency : very rare (synthetic metadata workload only)
4517 Description: two racing renames might cause an MDS thread to deadlock
4518 Details : Running the "racer" program may cause one MDS thread to rename
4519 a file from being the source of a rename to being the target of
4520 a rename at exactly the same time that another thread is doing
4521 so, and the second thread has already enqueued these locks after
4522 doing a lookup of the target and is trying to relock them in
4523 order. Ensure that we don't try to re-lock the same resource.
4526 Frequency : only very large systems with liblustre clients
4528 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4529 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4530 all OSTs in parallel. This allows the eviction of stale liblustre
4531 clients to proceed much faster than if they were done in series,
4532 and also offers similar improvements for other set_info RPCs.
4537 Description: excessive CPU usage during initial read phase on client
4538 Details : During the initial read phase on a client, it would agressively
4539 retry readahead on the file, consuming too much CPU and impacting
4540 performance (since 1.4.5.8). Improve the readahead algorithm
4541 to avoid this, and also improve some other common cases (read
4542 of small files in particular, where "small" is files smaller than
4543 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4548 Description: MDS crash when receiving packet with unknown intent.
4549 Details : Do not LBUG in unknown intent case, just return -EFAULT
4551 Severity : enhancement
4552 Bugzilla : 9293, 9385
4553 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4554 Details : Do not serialize getattr (non-intent version) and statfs.
4557 Frequency : occasional, when OST network is overloaded/intermittent
4559 Description: client evicted by OST after bulk IO timeout
4560 Details : If a client sends a bulk IO request (read or write) the OST
4561 may evict the client if it is unresposive to its data GET/PUT
4562 request. This is incorrect if the network is overloaded (takes
4563 too long to transfer the RPC data) or dropped the OST GET/PUT
4564 request. There is no need to evict the client at all, since
4565 the pinger and/or lock callbacks will handle this, and the
4566 client can restart the bulk request.
4569 Frequency : Always when mmapping file with no objects
4571 Description: client crashes when mmapping file with no objects
4572 Details : Check that we actually have objects in a file before doing any
4573 operations on objects in ll_vm_open, ll_vm_close and
4579 Description: Request leak when working with deleted CWD
4580 Details : Introduce advanced request refcount tracking for requests
4581 referenced from lustre intent.
4583 Severity : Enhancement
4585 Description: Cache open file handles on client.
4586 Details : MDS now will return special lock along with openhandle, if
4587 requested and client is allowed to hold openhandle, even if unused,
4588 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4589 closing files for every read/write openration.
4591 Severity : Enhancement
4593 Description: Cache open negative dentries on client when possible.
4594 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4595 negative dentries on lock revocation.
4600 Description: Remounting a client read-only wasn't possible with a zconf mount
4601 Details : It wasn't possible to remount a client read-only with llmount.
4603 Severity : enhancement
4604 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4605 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4606 MPICH is included to add Lustre-specific ADIO interfaces.
4607 This is based closely on the UFS ADIO layer and only differs
4608 in file creation, in order to allow the OST striping to be set.
4609 This is user-contributed code and not supported by CFS.
4614 Description: extended inode attributes (immutable, append-only) work improperly
4615 when 2.4 and 2.6 kernels are used on client/server or vice versa
4616 Details : Introduce kernel-independent values for these flags.
4618 Severity : enhancement
4621 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4622 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4623 granularity. Now, Fractional MB granularity is be supported,
4624 this is very useful for low memory system.
4626 Severity : enhancement
4628 Description: Getattr by fid
4629 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4630 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4631 have locks and avoids getattr rpc requests for it on every stat.
4634 Frequency : Always, for filesystems larger than 2TB
4636 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4637 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4638 filesystems larger than 2TB to be mounted with mballoc enabled.
4641 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4643 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4644 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4645 filesystems larger than 2TB is mounted then the kernel will
4646 silently corrupt the start of the filesystem. CONFIG_LBD is
4647 enabled for all CFS-supported kernels, but the possibility of
4648 this happening with a modified kernel config exists.
4650 Severity : enhancement
4652 Description: add client O_DIRECT support for 2.6 kernels
4653 Details : It is now possible to do O_DIRECT reads and writes to files
4654 in the Lustre client mountpoint on 2.6 kernel clients.
4656 Severity : enhancement
4658 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4659 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4660 parallel, not waiting for response from every OST before sending
4661 a rpc to the next OST.
4666 Description: setattr vs write race when updating file timestamps
4667 Details : Client processes that update a file timestamp into the past
4668 right after writing to the file (e.g. tar) it is possible that
4669 the updated file modification time can be reset to the current
4670 time due to a race between processing the setattr and write RPC.
4672 Severity : enhancement
4674 Description: Bring 'lfs find' closer in line with regular Linux find.
4675 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4676 -print0 options and obtains all the needed info through the lustre
4679 Severity : enhancement
4681 Description: support up to 1024 configured devices on one node
4682 Details : change obd_dev array from statically allocated to dynamically
4683 allocated structs as they are first used to reduce memory usage
4688 Description: Flush dirty partially truncated pages during truncate
4689 Details : Immediatelly flush partially truncated pages in filter_setattr,
4690 this way we completely avoid having any pages in page cache on OST
4691 and can retire ugly workarounds during writes to flush such pages.
4696 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4697 Details : protect inode from truncation within vfs_unlink() context
4698 just take a reference before calling vfs_unlink() and release it
4699 when parent's i_sem is free.
4704 Description: last_id value checked outside lock on OST caused LASSERT failure
4705 Details : If there were multiple MDS->OST object precreate requests in
4706 flight, it was possible that the OST's last object id was checked
4707 outside a lock and incorrectly tripped an assertion. Move checks
4708 inside locks, and discard old precreate requests.
4711 Frequency : always, if extents are used on OSTs
4713 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4714 Details : OSTs using the extents format would not zero the high 16 bits of
4715 the index physical block number. This is not a problem for any
4716 OST filesystems smaller than 16TB, and no kernels support ext3
4717 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4718 new/modified files) and can be fixed for existing filesystems
4719 with e2fsprogs-1.39-cfs1.
4724 Description: import connection selection may be incorrect if timer wraps
4725 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4726 import connections to be ignored if the 32-bit jiffies counter
4727 wraps. Use a 64-bit jiffies counter.
4730 Frequency : very large clusters immediately after boot
4732 Description: LNET request buffers exhausted under heavy short-term load
4733 Details : If a large number of client requests are generated on a service
4734 that has previously never seen so many requests it is possible
4735 that the request buffer growth cannot keep up with the spike in
4736 demand. Instead of dropping incoming requests, they are held in
4737 the LND until the RPC service can accept more requests.
4740 Frequency : Sometimes during replay
4742 Description: Assertion failure in ll_local_open after replay.
4743 Details : If replay happened on an open request reply before we were able
4744 to set replay handler, reply will become not swabbed tripping the
4745 assertion in ll_local_open. Now we set the handler right after
4746 recognising of open request
4749 Frequency : very rare
4751 Description: kernel reports "badness in vsnprintf"
4752 Details : Reading from the "recovery_status" /proc file in small chunks
4753 may cause a negative length in lprocfs_obd_rd_recovery_status()
4754 call to vsnprintf() (which is otherwise harmless). Exit early
4755 if there is no more space in the output buffer.
4757 Severity : enhancement
4759 Description: clear OBD RPC statistics by writing to them
4760 Details : It is now possible to clear the OBD RPC statistics by writing
4761 to the "stats" file.
4766 Description: Client mtime is not the same on different clients after utimes
4767 Details : In some cases, the client was using the utimes() syscall on
4768 a file cached on another node. The clients now validate the
4769 ctime from the MDS + OSTs to determine which one is right.
4774 Description: Inability to activate failout mode
4775 Details : lconf script incorrectly assumed that in pythong string's numeric
4776 value is used in comparisons.
4779 Frequency : always with multiple stripes per file
4781 Description: Inefficient object allocation for mutli-stripe files
4782 Details : When selecting which OSTs to stripe files over, for files with
4783 a stripe count that divides evenly into the number of OSTs,
4784 the MDS is always picking the same starting OST for each file.
4785 Return the OST selection heuristic to the original design.
4790 Description: mount failures may take full timeout to return an error
4791 Details : Under some heavy load conditions it is possible that a
4792 failed mount can wait for the full obd_timeout interval,
4793 possibly several minutes, before reporting an error.
4794 Instead return an error as soon as the status is known.
4796 Frequency : quota enabled and large files being deleted
4798 Description: releasing more than 4GB of quota at once hangs OST
4799 Details : If a user deletes more than 4GB of files on a single OST it
4800 will cause the OST to spin in an infinite loop. Release
4801 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4806 Description: statfs data retrieved from /proc may be stale or zero
4807 Details : When reading per-device statfs data from /proc, in the
4808 {kbytes,files}_{total,free,avail} files, it may appear
4809 as zero or be out of date.
4812 Frequency : systems with MD RAID1 external journal devices
4814 Description: lconf's call to blkid is confused by RAID1 journal devices
4815 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4816 of returning all block devices that match the journal UUID.
4819 Frequency : always, for aggregate stripe size over 4GB
4821 Description: assertion fails when trying to use 4GB stripe size
4822 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4823 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4826 Frequency : always on ppc64
4828 Description: the first write on an ext3 filesystem with mballoc got stuck
4829 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4830 perform endianness conversion.
4832 ------------------------------------------------------------------------------
4834 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4836 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4837 INTEROPERATE with older versions automatically. Please read the
4838 user documentation before upgrading any part of a live system.
4839 * WARNING: Lustre networking configuration changes are required with
4840 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4843 * Support for kernels:
4844 2.6.9-22.0.2.EL (RHEL 4)
4845 2.6.5-7.244 (SLES 9)
4846 2.6.12.6 vanilla (kernel.org)
4849 Severity : enhancement
4850 Bugzilla : 7981/8208
4851 Description: Introduced Lustre Networking (LNET)
4852 Details : LNET is new networking infrastructure for Lustre, it includes
4853 a reorganized network configuration mode (see the user
4854 documentation for full details) as well as support for routing
4855 between different network fabrics. Lustre Networking Devices
4856 (LNDs) for the supported network fabrics have also been
4857 created for this new infrastructure.
4859 Severity : enhancement
4860 Description: Introduced Access control lists
4861 Details : clients can set ACLs on files and directories in order to have
4862 more fine-grained permissions than the standard Unix UGO+RWX.
4863 The MDS must be started with the "-o acl" mount option.
4865 Severity : enhancement
4866 Description: Introduced filesystem quotas
4867 Details : Administrators may now establish per-user quotas on the
4870 Severity : enhancement
4872 Description: Configuration change for the XT3
4873 The PTLLND is now used to run Lustre over Portals on the XT3
4874 The configure option(s) --with-cray-portals are no longer used.
4875 Rather --with-portals=<path-to-portals-includes> is used to
4876 enable building on the XT3. In addition to enable XT3 specific
4877 features the option --enable-cray-xt3 must be used.
4882 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4883 Details : A race between allocating a new llcd and re-getting the llcd_lock
4884 allowed another thread to grab newly-allocated llcd.
4886 Severity : enhancement
4888 Description: 2.6 OST async journal commit and locking fix to improve performance
4889 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4890 2.6 kernels are now async as they already were in 2.4 kernels so
4891 that they can commit concurrently with the network bulk transfer.
4892 For block-allocated files the filter allocation semaphore is held
4893 to avoid filesystem fragmentation during allocation. BKL lock
4894 removed for 2.6 xattr operations where it is no longer needed.
4899 Description: lconf incorrectly determined whether two IP networks could talk
4900 Details : In some more complicated routing and multiple-network
4901 configurations, lconf will avoid trying to make a network
4902 connection to a disjoint part of the IP space. It was doing the
4903 math incorrectly for one set of cases.
4908 Description: Fix for potential infinite loop processing records in an llog.
4909 Details : If an llog record is corrupted/zeroed, it is possible to loop
4910 forever in llog_process(). Validate the llog record length
4911 and skip the remainder of the block on error.
4914 Frequency : occasional (liblustre only)
4916 Description: liblustre could not open files whose last component is a symlink
4917 Details : sysio_path_walk() would incorrectly pass the open intent to
4918 intermediate path components.
4921 Frequency : rare (liblustre only with non-standard tuning)
4922 Bugzilla : 7201 (7350)
4923 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4924 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4925 after it has been released, only do this for referenced locks
4927 Severity : enhancement
4929 Description: specifying an (invalid) directory default stripe_size of -1
4930 would reset the directory default striping
4931 Details : stripe_size -1 was used internally to signal directory stripe
4932 removal, now use "all default" to signal dir stripe removal
4933 as a directory striping of "all default" is not useful
4936 Frequency : common for large clusters running liblustre clients
4938 Description: doing an ls when liblustre clients are running is slow
4939 Details : sending a glimpse AST to a liblustre client waits for every AST
4940 to time out, as liblustre clients will not respond. Since they
4941 cannot cache data we refresh the OST lock LVB from disk instead.
4943 Severity : enhancement
4945 Description: doing an ls at the same time as file IO can be slow
4946 Details : enqueue and other "small" requests can be blocked behind many
4947 large IO requests. Create a new OST IO portal for non-IO
4948 requests so they can be processed faster.
4951 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4953 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4954 created file with mode 000
4955 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4956 allow an ftruncate() as a non-root user to a file with mode 000.
4957 Lustre now respects this flag to disable mode checks when
4958 truncating a file owned by the user
4961 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4964 Description: Revoking locks from clients that went dead or catatonic might take
4966 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4967 cancellation of such locks instant on servers without waiting for
4968 any reply from clients. Clients drops these locks when cancel
4969 notification from server is received without replying.
4972 Frequency : liblustre-only, when liblustre client dies or becomes busy
4974 Description: Doing ls on Linux clients can take a long time with active
4976 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4977 granting such locks to it in the first place if possible. Locks
4978 are taken by proxy on the OST during the read or write and
4979 dropped immediately afterward. Add connect flags handling, do
4980 not grant locks to liblustre clients for glimpse ASTs.
4982 Severity : enhancement
4984 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4985 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4986 sequential reads and behaves badly with applications doing random
4987 reads. Improve it by reading ahead at least read region, and
4988 avoiding excessive large RPC for small reads.
4990 Severity : enhancement
4992 Description: Creating more than 1000 files for a single job may cause a load
4993 imbalance on the OSTs if there are also a large number of OSTs.
4994 Details : qos_prep_create() uses an OST index reseed value that is an
4995 even multiple of the number of available OSTs so that if the
4996 reseed happens in the middle of the object allocation it will
4997 still utilize the OSTs as uniformly as possible.
5002 Description: OST or MDS may oops in ping_evictor_main()
5003 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5004 but doesn't restart at beginning of obd_exports_timed list
5007 Severity : enhancement
5009 Description: improve by-nid export eviction on the MDS and OST
5010 Details : allow multiple exports with the same NID to be evicted at one
5011 time without re-searching the exports list.
5014 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5016 Description: MDS may oops in groups_free()
5017 Details : in rare race conditions a newly allocated group_info struct is
5018 freed again, and this can be NULL. The 2.4 compatibility code
5019 for groups_free() checked for a NULL pointer, but 2.6 did not.
5022 Frequency : common for liblustre clients doing little filesystem IO
5023 Bugzilla : 9352, 7313
5024 Description: server may evict liblustre clients accessing contended locks
5025 Details : if a client is granted a lock or receives a completion AST
5026 with a blocking AST already set it would not reply to the AST
5027 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5030 Frequency : lfs setstripe, only systems with more than 160 OSTs
5032 Description: unable to set striping with a starting offset beyond OST 160
5033 Details : llapi_create_file() incorrectly limited the starting stripe
5034 index to the maximum single-file stripe count.
5037 Frequency : LDAP users only
5039 Description: lconf did not handle in-kernel recovery with LDAP properly
5040 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5042 Severity : enhancement
5044 Description: bind OST threads to NUMA nodes to improve performance
5045 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5046 node and do their allocations there to localize memory access
5048 Severity : enhancement
5050 Description: llmount can determine client NID directly from Myrinet (GM)
5051 Details : the client NID code from gmnalnid was moved directly into
5052 llmount, removing the need to use this or specifying the
5053 client NID explicitly when mounting GM clients with zeroconf
5056 Frequency : if client is started with down MDS
5058 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5059 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5060 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5061 ptlrpc_import_delayed_req() because we didn't check if the
5062 request was interrupted, and we also didn't break out of the
5063 event loop if there was no timeout
5068 Description: data loss during non-page-aligned writes to a single file from
5069 both multiple nodes and multiple threads on one node at same time
5070 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5071 inconsistency led to false short-reads, that were cached and later
5072 used by ->prepare_write() to fill in partially written page,
5073 leading to data loss.
5076 Frequency : always, if lconf --abort_recovery used
5078 Description: lconf --abort_recovery fails with 'Operation not supported'
5079 Details : lconf was attempting to abort recovery on the MDT device and not
5082 Severity : enhancement
5084 Description: remove cleanup logs
5085 Details : replace lconf-generated cleanup logs with lustre internal
5086 cleanup routines. Eliminates the need for client-cleanup and
5089 Severity : enhancement
5091 Description: add support for EAs (user and system) on lustre filesystems
5092 Details : it is now possible to store extended attributes in the Lustre
5093 client filesystem, and with the user_xattr mount option it
5094 is possible to allow users to store EAs on their files also
5096 Severity : enhancement
5098 Description: Add possibility (config option) to show minimal available OST free
5100 Details : When compiled with --enable-mindf configure option, statfs(2)
5101 (and so, df) will return least minimal free space available from
5102 all OSTs as amount of free space on FS, instead of summary of
5103 free spaces of all OSTs.
5105 Severity : enhancement
5107 Description: do not expand extent locks acquired on OST-side
5108 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5109 by server: they are not cached anyway.
5112 Frequency : when mmap is used/binaries executed from Lustre
5114 Description: Unmmap pages before throwing them away from read cache.
5115 Details : llap_shrink cache now attempts to unmap pages before discarding
5116 them (if unmapping failed - do not discard). SLES9 kernel has
5117 extra checks that trigger if this unmapping is not done first.
5122 Description: lconf didn't resolve symlinks before checking to see whether a
5123 given mountpoint was already in use
5126 Frequency : when migrating failover services
5127 Bugzilla : 6395, 9514
5128 Description: When migrating a subset of services from a node (e.g. failback
5129 from a failover service node) the remaining services would
5130 time out and evict clients.
5131 Details : lconf --force (implied by --failover) sets the global obd_timeout
5132 to 5 seconds in order to quickly disconnect, but this caused
5133 other RPCs to time out too quickly. Do not change the global
5134 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5136 Severity : enhancement
5137 Frequency : if MDS is started with down OST
5138 Bugzilla : 9439,5706
5139 Description: Allow startup/shutdown of an MDS without depending on the
5140 availability of the OSTs.
5141 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5142 Add appropriate locking and lov-osc refcounts for safe
5143 cleaning. Add osc abort_inflight calls in case the
5144 synchronize never started.
5147 Frequency : occasional (Cray XT3 only)
5149 Description: root not authorized to access files in CRAY_PORTALS environment
5150 Details : The client process capabilities were not honoured on the MDS in
5151 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5152 been accessed by an authorized user then root was able to access
5153 the file on the local client also. The root user capabilities
5154 are now allowed on the MDS, as this environment has secure UID.
5157 Frequency : occasional
5159 Description: ldiskfs "too long searching" message happens too often
5160 Details : A debugging message (otherwise harmless) prints too often on
5161 the OST console. This has been reduced to only happen when
5162 there are fragmentation problems on the filesystem.
5167 Description: Division by zero in statfs when all OSCs are inactive
5168 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5169 lov_statfs divides by value returned by lov_get_stripecnt().
5173 Bugzilla : 9489, 3273
5174 Description: First write from each client to each OST was only 4kB in size,
5175 to initialize client writeback cache, which caused sub-optimal
5176 RPCs and poor layout on disk for the first written file.
5177 Details : Clients now request an initial cache grant at (re)connect time
5178 and so that they can start streaming writes to the cache right
5179 away and always do full-sized RPCs if there is enough data.
5180 If the OST is rebooted the client also re-establishes its grant
5181 so that client cached writes will be honoured under the grant.
5186 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5187 Details : Now I/O RPCs go to different portal number and (presumably) fast
5188 lock requests (and glimses) and other RPCs get their own service
5189 threads pool that should be able to service those RPCs
5192 Severity : enhancement
5194 Description: Ability to exchange lustre version between client and servers and
5195 issue warnings at client side if client is too old. Also for
5196 liblustre clients there is ability to refuse connection of too old
5198 Details : New 'version' field is added to connect data structure that is
5199 filled with version info. That info is later checked by server and
5203 Frequency : rare, liblustre only.
5204 Bugzilla : 9296, 9581
5205 Description: Two simultaneous writes from liblustre at offset within same page
5206 might proceed at the same time overwriting eachother with stale
5208 Details : I/O lock withing llu_file_prwv was released too early, before data
5209 actually was hitting the wire. Extended lock-holding time until
5210 server acknowledges receiving data.
5213 Frequency : extremely rare. Never observed in practice.
5215 Description: avoid generating lustre_handle cookie of 0.
5216 Details : class_handle_hash() generates handle cookies by incrementing
5217 global counter, and can hit 0 occasionaly (this is unlikely, but
5218 not impossible, because initial value of cookie counter is
5219 selected randonly). Value of 0 is used as a sentinel meaning
5220 "unassigned handle" --- avoid it. Also coalesce two critical
5221 sections in this function into one.
5223 Severity : enhancement
5225 Description: allow liblustre clients to delegate truncate locking to OST
5226 Details : To avoid overhead of locking, liblustre client instructs OST to
5227 take extent lock in ost_punch() on client's behalf. New connection
5228 flag is added to handle backward compatibility.
5230 Severity : enhancement
5231 Bugzilla : 4928, 7341, 9758
5232 Description: allow number of OST service threads to be specified
5233 Details : a module parameter allows the number of OST service threads
5234 to be specified via "options ost ost_num_threads={N}" in the
5235 OSS's /etc/modules.conf or /etc/modprobe.conf.
5239 Bugzilla : 6146, 9635, 9895
5240 Description: servers crash with bad pointer in target_handle_connect()
5241 Details : In rare cases when a client is reconnecting it was possible that
5242 the connection request was the last reference for that export.
5243 We would temporarily drop the export reference and get a new
5244 one, but this may have been the last reference and the export
5245 was just destroyed. Get new reference before dropping old one.
5247 Severity : enhancement
5248 Frequency : if client is started with failover MDS
5250 Description: Allow multiple MDS hostnames in the mount command
5251 Details : Try to read the configuration from all specified MDS
5252 hostnames during a client mount in case the "primary"
5255 Severity : enhancement
5257 Description: Stop sending data to evicted clients as soon as possible.
5258 Details : Check if the client we are about to send or are sending data to
5259 was evicted already. (Check is done every second of waiting,
5260 for which l_wait_event interface was extended to allow checking
5261 of exit condition at specified intervals).
5264 Frequency : rare, normally only when NFS exporting is done from client
5266 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5268 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5271 Frequency : always, if srand() is called before liblustre initialization
5273 Description: Liblustre uses system PRNG disturbing its usage by user application
5274 Details : Introduce internal to lustre fast and high-quality PRNG for
5275 lustre usage and make liblustre and some other places in generic
5276 lustre code to use it.
5278 Severity : enhancement
5279 Bugzilla : 9477, 9557, 9870
5280 Description: Verify that the MDS configuration logs are updated when xml is
5281 Details : Check if the .xml configuration logs are newer than the config
5282 logs stored on the MDS and report an error if this is the case.
5283 Request --write-conf, or allow starting with --old_conf.
5285 Severity : enhancement
5287 Description: Handle symlinks in the path when checking if Lustre is mounted.
5288 Details : Resolve intermediate symlinks when checking if a client has
5289 mounted a filesystem to avoid duplicate client mounts.
5294 Description: lconf can hit an error exception but still return success.
5295 Details : The lconf command catches the Command error exception at the top
5296 level script context and will exit with the associated exit
5297 status, but doesn't ensure that this exit status is non-zero.
5302 Description: failure of ptlrpc thread startup can cause oops
5303 Details : Starting a ptlrpc service thread can fail if there are a large
5304 number of threads or the server memory is very fragmented.
5305 Handle this without oopsing.
5308 Frequency : always, only if liblustre and non-default acceptor port was used
5310 Description: liblustre cannot connect to servers with non-default acceptor port
5311 Details : tcpnal_set_default_params() was not called and was therefore
5312 ignoring the environment varaible TCPNAL_PORT, as well as other
5313 TCPNAL_ environment variables
5318 Description: two objects could be created on the same OST for a single file
5319 Details : If an OST is down, in some cases it was possible to create two
5320 objects on a single OST for a single file. No problems other
5321 than potential performance impact and spurious error messages.
5325 Bugzilla : 5681, 9562
5326 Description: Client may oops in ll_unhash_aliases
5327 Details : Client dcache may become inconsistent in race condition.
5328 In some cases "getcwd" can fail if the current directory is
5334 Description: Inode refcounting problems in NFS export code
5335 Details : link_raw functions used to call d_instantiate without obtaining
5336 extra inode reference first.
5340 Bugzilla : 9942, 9903
5341 Description: Referencing freed requests leading to crash, memleaks with NFS.
5342 Details : We used to require that call to ll_revalidate_it was always
5343 followed by ll_lookup_it. Also with revalidate_special() it is
5344 possible to call ll_revalidate_it() twice for the same dentry
5345 even if first occurence returned success. This fix changes semantic
5346 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5347 reference on a request referred from the intent.
5348 ll_intent_release() then releases such a request.
5351 Frequency : rare, normally benchmark loads only
5353 Description: unlinked inodes were kept in memory on the client
5354 Details : If a client is repeatedly creating and unlinking files it
5355 can accumulate a lot of stale inodes in the inode slab cache.
5356 If there is no other client load running this can cause the
5357 client node to run out of memory. Instead flush old inodes
5358 from client cache that have the same inode number as a new inode.
5361 Frequency : SLES9 2.6.5 kernel and long filenames only
5362 Bugzilla : 9969, 10379
5363 Description: utime reports stale NFS file handle
5364 Details : SLES9 uses out-of-dentry names in some cases, which confused
5365 the lustre dentry revalidation. Change it to always use the
5369 Frequency : rare, unless heavy write-truncate concurrency is continuous
5370 Bugzilla : 4180, 6984, 7171, 9963, 9331
5371 Description: OST becomes very slow and/or deadlocked during object unlink
5372 Details : filter_destroy() was holding onto the parent directory lock
5373 while truncating+unlinking objects. For very large objects this
5374 may block other threads for a long time and slow overall OST
5375 responsiveness. It may also be possible to get a lock ordering
5376 deadlock in this case, or run out of journal credits because of
5377 the combined truncate+unlink. Solution is to do object truncate
5378 first in one transaction without parent lock, and then do the
5379 final unlink in a new transaction with the parent lock. This
5380 reduces the lock hold time dramatically.
5383 Frequency : rare, 2.4 kernels only
5385 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5386 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5387 before unmounting their filesystem in order to avoid deadlock.
5388 The kernel_locked() function in 2.4 kernels only checks whether
5389 the kernel lock is held, not whether it is this process that is
5390 holding it as 2.6 kernels do.
5395 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5396 Details : The client ptlrpc code may be trying to reconnect to a down
5397 server before a previous connection attempt has timed out.
5398 Increase the reconnect interval to be longer than the connection
5399 timeout interval to avoid sending duplicate connections to
5403 Frequency : echo_client brw_test command
5405 Description: fix echo_client to work with OST preallocated code
5406 Details : OST preallocation code (5137) didn't take echo_client IO path
5407 into account: echo_client calls filter methods outside of any
5408 OST thread and, hence, there is no per-thread preallocated
5409 pages and buffers to use. Solution: hijack pga pages for IO. As
5410 a byproduct, this avoids unnecessary data copying.
5414 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5415 Description: Client can oops in mdc_commit_close() after open replay
5416 Details : It was possible for the MDS to return an open request with no
5417 transaction number in mds_finish_transno() if the client was
5418 evicted, but without actually returning an error. Clients
5419 would later try to replay that open and may trip an assertion
5420 Simplify the client close codepath, and always return an error
5421 from the MDS in case the open is not successful.
5424 Frequency : rare, 2.6 OSTs only
5426 Description: OST may deadlock under high load on fragmented files
5427 Details : If there was a heavy load and highly-fragmented OST filesystems
5428 it was possible to have all the OST threads deadlock waiting on
5429 allocation of biovecs, because the biovecs were not released
5430 until the entire RPC IO was completed. Instead, release biovecs
5431 as soon as they are complete to ensure forward IO progress.
5433 Severity : enhancement
5435 Description: Support for specifying external journal device at mount
5436 Details : If an OST or MDS device is formatted with an external journal
5437 device, this device major/minor is stored in the ext3 superblock
5438 and may not be valid for failover. Allow detecting and
5439 specifying the external journal at mount time.
5444 Description: Mounting an MDS with pending unlinked files may cause oops
5445 Details : target_finish_recovery() calls mds_postrecov() which returned
5446 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5447 considers this an error and immediately begins cleaning up the
5448 lov, just after starting the mds_lov process
5450 Severity : enhancement
5452 Description: Implement 'lfs df' to report actual free space on per-OST basis
5453 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5454 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5455 usage of objects; '-h' to report in human readable format.
5457 ------------------------------------------------------------------------------
5459 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5466 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5467 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5468 at startup during orphan unlinking. Instead, skip buddy update
5469 before setup, as it will be regenerated after recovery is complete.
5474 Description: If an OST is inactive, its locks might reference stale inodes.
5475 Details : lov_change_cbdata() must iterate over all namespaces, even if
5476 they are inactive to clear inode references from the lock.
5478 Severity : enhancement
5479 Frequency : occasional, if non-standard max_dirty_mb used
5481 Description: Client will block write RPCs if not enough grant
5482 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5483 then the client will block writes while waiting for another RPC
5484 to complete instead of consuming its dirty limit. With change
5485 we get improved performance when max_dirty_mb is small.
5487 Severity : enhancement
5488 Bugzilla : 3389, 6253
5489 Description: Add support for supplementary groups on the MDS.
5490 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5491 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5492 lookups for user supplementary groups into a cache.
5496 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5497 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5498 for liblustre O_EXCL is enforced by MDS.
5502 Description: readdir never returns NULL in liblustre.
5503 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5504 dentry in struct dirent.
5508 Frequency : liblustre only, depends on application IO pattern
5509 Description: liblustre clients evicted if not contacting servers
5510 Details : Don't put liblustre clients into the ping_evictor list, so
5511 they will not be evicted by the pinger ever.
5513 Severity : enhancement
5515 Description: Add ability to evict clients by NID from MDS.
5516 Details : By echoing "nid:$NID" string into
5517 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5518 $NID would be instantly evicted from this MDS and from all active
5519 OSTs connected to it.
5523 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5524 Details : lookup_it_finish() used to query file size from OSTs that was not
5529 Description: service threads change working directory to that of init
5530 Details : Starting lustre service threads may pin the working directory
5531 of the parent thread, making that filesystem busy. Threads
5532 now change to the working directory of init to avoid this.
5536 Frequency : during shutdown only
5537 Description: shutdown with a failed MDS or OST can cause unmount to hang
5538 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5543 Frequency : 2.6 only
5544 Description: chmod/chown may include an extra supplementary group
5545 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5546 supplementary group and if none is specified this is used.
5549 Bugzilla : 5479 (6816)
5551 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5552 Details : If lookup is in progress on a file that is unlinked we might try
5553 to revalidate the inode and fail in revalidate after lookup is
5554 complete and ll_file_open() enqueues the open again but
5555 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5558 Frequency : always, if lconf --abort_recovery used
5560 Description: lconf --abort_recovery fails with 'Operation not supported'
5561 Details : lconf was attempting to abort recovery on the MDT device and not
5564 ------------------------------------------------------------------------------
5566 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5571 Frequency : rare (only unsupported configurations with a node running as an
5573 Bugzilla : 6514, 5137
5574 Description: Mounting a Lustre file system on a node running as an OST could
5576 Details : OSTs now preallocates memory needed to write out data at
5577 startup, instead of when needed, to avoid having to
5578 allocate memory in possibly low memory situations.
5579 Specifically, if the file system is mounted on on OST,
5580 memory pressure could force it to try to write out data,
5581 which it needed to allocate memory to do. Due to the low
5582 memory, it would be unable to do so and the node would
5583 become unresponsive.
5585 Severity : enhancement
5587 Description: Addition of lconf --service command line option
5588 Details : lconf now accepts a '--service <arg>' option, which is
5589 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5591 Severity : enhancement
5593 Description: Failover mode is now the default for OSTs.
5594 Details : By default, OSTs will now run in failover mode. To return to
5595 the old behaviour, add '--failout' to the lmc line for OSTs.
5597 Severity : enhancement
5599 Description: Health checks are now provided for MDS and OSTs
5600 Details : Additional detailed health check information on MSD and OSTs
5601 is now provided through the procfs health_check value.
5604 Frequency : occasional, depends on IO load
5606 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5607 after numerous create/delete cycles
5608 Details : The ext3 inode allocation policy would not allocate new inodes
5609 very well on the OSTs because there are no new directories
5610 being created. Instead we look for groups with free space if
5611 the parent directories are nearly full.
5616 Description: Network or server problems during mount may cause partially
5617 mounted clients instead of returning an error.
5618 Details : The config llog parsing code may overwrite the error return
5619 code during mount error handling, returning success instead
5625 Description: MDS can fail to allocate large reply buffers
5626 Details : After long uptimes the MDS can fail to allocate large reply
5627 buffers (e.g. zconf client mount config records) due to memory
5628 fragmentation or consumption by the buffer cache. Preallocate
5629 some large reply buffers so that these replies can be sent even
5630 under memory pressure.
5634 Frequency : rare (liblustre)
5635 Description: fsx running with liblustre complained that using truncate() to
5636 extend the file doesn't work. This patch corrects that issue.
5637 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5638 ATTR_SIZE and lsm use in llu_setattr_raw.
5642 Frequency : rare, only 2.6 kernels
5643 Description: Unusual file access patterns on the MDS may result in inode
5644 data being lost in very rare circumstances.
5645 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5646 code in the 2.6 kernel caused the inode and/or EA data not to
5647 be read from disk, causing single-file corruption.
5651 Frequency : rare, only 2.6 filesystems using extents
5652 Description: Heavy concurrent write and delete load may cause data corruption.
5653 Details : It was possible under high-load situations to have an extent
5654 metadata block in the block device cache from a just-unlinked
5655 file overwrite a newly-allocated data block. We now unmap any
5656 metadata buffers that alias just-allocated data blocks.
5660 Frequency : filesystems with default stripe_count larger than 77
5661 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5662 Details : lconf specifies an inode size of 4096 bytes when the default
5663 stripe_count is larger than 77. This conflicts with the default
5664 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5665 this case to avoid pinning too much memory for large EAs.
5667 ------------------------------------------------------------------------------
5669 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5674 Frequency : rare (extremely heavy IO load with hundreds of clients)
5676 Description: Client is evicted, gets IO error writing to file
5677 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5678 caused clients to be evicted for AST timeouts. The fixes in
5679 bug 5192 mean we no longer need to have such short AST timeouts
5680 so ldlm_timeout has been increased.
5683 Frequency : occasional during --force or --failover shutdown under load
5684 Bugzilla : 5949, 4834
5685 Description: Server oops/LBUG if stopped with --force or --failover under load
5686 Details : a collection of import/export refcount and cleanup ordering
5687 issues fixed for safer force cleanup
5690 Frequency : only filesystems larger than 120 OSTs
5691 Bugzilla : 5990, 6223
5692 Description: lfs getstripe would oops on a very large filesystem
5693 Details : lov_getconfig used kfree on vmalloc'd memory
5696 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5697 Bugzilla : 6242, 6243
5698 Description: reading from files that had been truncated to a non-zero size
5699 but never opened returned no data
5700 Details : ll_file_read() reads zeros from no-object files to EOF
5705 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5706 Details : There's little harm in aborting MDS/OSS recovery and letting it
5707 try again, so I removed the LASSERT and return an error instead.
5709 Severity : enhancement
5711 Description: New debugging infrastructure for tracking down data corruption
5712 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5713 (b) cover more of the client-side code path, and (c) try to narrow
5714 down where problems occurred
5718 Bugzilla : 3819, 4364, 4397, 6313
5719 Description: Racing close and eviction MDS could cause assertion in mds_close
5720 Details : It was possible to get multiple mfd references during close and
5721 client eviction, leading to one thread referencing a freed mfd.
5723 Severity: : enhancement
5724 Bugzilla : 3262, 6359
5725 Description: Attempts to reconnect to servers are now more aggressive.
5726 Details : This builds on the enhanced upcall-less recovery that was added
5727 in 1.4.2. When trying to reconnect to servers, clients will
5728 now try each server in the failover group every 10 seconds. By
5729 default, clients would previously try one server every 25 seconds.
5734 Description: After recovery, certain operations trigger a failed
5735 assertion on a client.
5736 Details : Failing over an mds, using lconf -d --failover, while a
5737 client was doing a readdir() call would cause the client to
5738 LBUG after recovery completed and the readdir() was resent.
5740 Severity : enhancement
5742 Description: Default groups are now added by lconf
5743 Details : You can now run lconf --group <servicename> without having to
5744 manually add groups with lmc.
5747 Frequency : occasional
5749 Description: Nodes with an elan id of 0 trigger a failed assertion
5752 Frequency : always when accessing e.g. tty/console device nodes
5754 Description: tty and some other devices nodes cannot be used on lustre
5755 Details : file's private_data field is used by device data and lustre
5756 values in there got lost. New field was added to struct file to
5757 store fs-specific private data.
5760 Frequency : when exporting Lustre via NFS
5762 Description: NFSD failed occasionally when looking up a path component
5763 Details : NFSD is looking up ".." which was broken in ext3 directories
5764 that had grown large enough to become hashed.
5767 Frequency : Clusters with multiple interfaces not on the same subnet
5769 Description: Nodes will repeatedly try to reconnect to an interface which it
5770 cannot reach and report an error to the log.
5771 Details : Extra peer list entries will be created by lconf with some peers
5772 unreachable. lconf now validates the peer before adding it.
5775 Frequency : Only if a default stripe is set on the filesystem root.
5777 Description: Setting a default stripe on the filesystem root prevented the
5778 filesystem from being remounted.
5779 Details : The client was sending extra request flags in the root getattr
5780 request and did not allocate a reply buffer for the dir EA.
5783 Frequency : occasional, higher if lots of files are accessed by one client
5784 Bugzilla : 6159, 6097
5785 Description: Client trips assertion regarding lsm mismatch/magic
5786 Details : While revalidating inodes the VFS looks up inodes with ifind()
5787 and in rare cases can find an inode that is being freed.
5788 The ll_test_inode() code will free the lsm during ifind()
5789 when it finds an existing inode and then the VFS later attaches
5790 this free lsm to a new inode.
5794 Bugzilla : 6422, 7030
5795 Description: MDS deadlock between mkdir and client eviction
5796 Details : Creating a new file via mkdir or mknod (starting a transaction
5797 and getting the ns lock) can deadlock with client eviction
5798 (gets ns lock and trying to finish a synchronous transaction).
5801 Frequency : occasional
5802 Description: While starting a server, the fsfilt_ext3 module could not be
5804 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5805 kernels. Previously, lconf would still use the ext3 name
5806 when trying to load modules. Now, it will correctly use
5807 ext3 on 2.4 and ldiskfs on 2.6.
5809 Severity : enhancement
5810 Description: The default stripe count has been changed to 1
5811 Details : The interpretation of the default stripe count (0, to lfs
5812 or lmc) has been changed to mean striping across a single
5813 OST, rather than all available. For general usage we have
5814 found a stripe count of 1 or 2 works best.
5816 Severity : enhancement
5817 Description: Add support for compiling against Cray portals.
5818 Details : Conditional compiling for some areas that are different
5822 Frequency : occasional
5823 Bugzilla : 6409, 6834
5824 Description: Creating files with an explicit stripe count may lead to
5825 a failed assertion on the MDS
5826 Details : If some OSTs are full or unavailable, creating files may
5827 trigger a failed assertion on the MDS. Now, Lustre will
5828 try to use other servers or return an error to the
5832 Frequency : occasional
5834 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5835 caused lock timeout during readahead (since 1.4.2).
5836 Details : Processes doing readahead might match a lock that hasn't been
5837 granted yet if there are overlapping and conflicting lock
5838 requests. The readahead process waits on ungranted lock
5839 (original lock is CBPENDING), while OST waits for that process
5840 to cancel CBPENDING read lock and eventually evicts client.
5842 Severity : enhancement
5844 Description: Initial enabling of flock support for clients
5845 Details : Implements fcntl advisory locking and file status functions.
5846 This feature is provided as an optional mount flag (default
5847 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5848 locking are implemented yet, and those that are are not guaranteed
5849 to be completely correct in production environments.
5850 mount -t lustre -o [flock|noflock] ...
5853 Frequency : occasional
5855 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5856 assertion in the ext3 JBD (journaling) layer.
5857 Details : The b_committed_data struct is protected by the big kernel lock
5858 in 2.4 kernels, serializing journal_commit_transaction() and
5859 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5860 access to this struct. In 2.6 kernels there is finer grained
5861 locking to improve SMP performance of the JBD layer.
5865 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5870 Description: merge and cleanup kernel patches.
5873 -----------------------------------------------------------------------------
5875 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5877 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5878 versions of Lustre. Please update all servers and clients to
5879 version 1.4.2 or later at the same time. You must also run
5880 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5881 to update the configuration logs.
5883 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5884 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5885 - swab llog records of type '0' so we get proper header size/idx (5861)
5886 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5887 - fix rename of one directory over another leaking an inode (5953)
5888 - avoid SetPageDirty on 2.6 (5981)
5889 - don't re-add just-being-destroyed locks to the waiting list (5653)
5890 - when creating new directories, inherit the parent's custom
5891 striping settings if present parent (3048)
5892 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5893 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5894 - don't hold client locks on temporary worklist from l_lru (5666)
5895 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5896 - automatically evict dead clients (5921)
5897 - Update file size properly in create+truncate+fstat case (6196)
5898 - Do not unhash mountpoint dentries, do not allow removal of
5900 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5901 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5902 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5903 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5904 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5905 - Correctly handle reads of files with no objects (6243)
5906 - lctl recover will also mark a device active if deactivate used (5933)
5908 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5909 - allow --write-conf on an MDS with different nettype than client (5619)
5910 - don't write config llogs to MDS for mounts not from that MDS (5617)
5911 - lconf should create multiple TCP connections from a client (5201)
5912 - init scripts are now turned off by default; run chkconfig --on
5913 lustre and chkconfig --on lustrefs to use them
5914 - upcalls are no longer needed for clients to recover to failover
5916 - add --abort-recovery option to lconf to abort recovery on device
5918 - add support for an arbitrary number of OSTs (3026)
5919 - Quota support protocol changes.
5920 - forward compatibility changes to wire structs (6007)
5921 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5922 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5923 - improved "lustre status" script
5924 - initialize blocksize for non-regular files (6062)
5925 - added --disable-server and --disable-client configure options (5782)
5926 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5927 - Vanilla 2.4.29 support
5928 - increase maximum number of obd devices to 520 (6242)
5929 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5930 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5931 SLES 9 2.6.5 kernels
5932 - limit stripes per file to 160 (the maximum EA size) (6093)
5934 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5937 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5938 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5939 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5940 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5941 - small ext3 extents cleanups and fixes (5733)
5942 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5943 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5944 - avoid needless client->OST connect, fix handle mismatch (5317)
5945 - fix DLM error path that led to out-of-sync client, long delays (5779)
5946 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5947 - fix several locking issues related to i_size (5492,5624,5654,5672)
5948 - don't move pending lock onto export if it is already evicted (5683)
5949 - fix kernel oops when creating .foo in unlinked directory (5548)
5950 - fix deadlock in obdfilter statistics vs. object create (5811)
5951 - use time_{before,after} to avoid timer jiffies wrap (5882)
5952 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5953 - Do not leak request if server was not able to process it (5154)
5954 - If mds_open unable to find parent dir, make that negative lookup(5154)
5955 - don't create new directories with extent-mapping (5909, 5936)
5957 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5958 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5959 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5960 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5961 - support for RHEL4 kernel on i686 (5773)
5962 - provide error messages when incompatible logs are encountered (5898)
5964 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5965 * version 1.4.0.10 (1.4.1 release candidate 1)
5967 - don't keep a lock reference when lock is not granted (4238)
5968 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5969 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5970 - drop import inflight refcount on signal_completed_replay error (5255)
5971 - unlock page after async write error during send (3677)
5972 - handle missing objects in filter_preprw_read properly (5265)
5973 - no transno return for symlink open, don't save no-trasno open (3440)
5974 - don't try to complete elan receive that already failed (4012)
5975 - free RPC server reply state on error (5406)
5976 - clean up thread from ptlrpc_start_thread() on error (5160)
5977 - readahead could read extra page into cache that wasn't ejected (5388)
5978 - prevent races in class_attach/setup/cleanup/detach (5260)
5979 - don't dereference de->d_inode after l_dput of de (5458)
5980 - use "int" for stripe value returned from lock_to_stripe (5544)
5981 - mballoc allocation and error-checking fixes in 2.6 (5504)
5982 - block device patches to fix I/O request sizes in 2.6 (5482)
5983 - look up hostnames for IB nals (5602)
5984 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5985 - don't start multiple acceptors for the same port (5277)
5986 - fix incorrect LASSERT in mds_getattr_name (5635)
5987 - export a proc file for general "ping" checking (5628)
5988 - fix "lfs check" to not block when the MDS is down (5628)
5990 - service request history (4965)
5991 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5992 - create an "evict_client" /proc entry on OSTs, like the MDS has
5993 - fix mount usage message, return errors per mount(8) (5168)
5994 - change grep [] to grep "[]" in tests so they work in more UMLs
5995 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5996 - remove ancient LOV_MAGIC_V0 EA support (5047)
5997 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5998 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5999 - allow readahead allocations to fail when low on memory (5383)
6000 - mmap locking landed again, after considerable improvement (2828)
6001 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6003 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6006 - send OST transaction number in read/write reply to free req (4966)
6007 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6008 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6009 - the watchdog thread now runs as interruptible (5246)
6010 - flock/lockf fixes (but it's still disabled, pending 5135)
6011 - don't use EXT3 constants in llite code (5094)
6012 - memory shortage at startup could cause assertion (5176)
6014 - reorganization of lov code
6015 - single portals codebase
6017 - add extents/mballoc support (5025)
6018 - direct I/O reads in the obdfilter (4048)
6019 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6021 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6024 - send OST transaction number in read/write reply to free req (4966)
6025 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6026 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6027 - the watchdog thread now runs as interruptible (5246)
6028 - handle missing objects in filter_preprw_read properly (5265)
6029 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6030 - drop import inflight refcount on signal_completed_replay error (5255)
6031 - unlock page after async write error during send (3677)
6032 - return original error code on reconstructed replies (3761)
6033 - no transno return for symlink open, don't save no-trasno open (3440)
6035 - add pid to ldlm debugging output (4922)
6036 - bump the watchdog timeouts -- we can't handle 30sec yet
6037 - extra debugging for orphan dentry/inode bug (5259)
6039 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6042 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6043 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6044 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6045 - fix NULL dereference in /proc/sys/portals/routes (4827)
6046 - allow failed mdc_close() operations to be interrupted (4561)
6047 - stop precreate on OST before MDS would time out on it (4778)
6048 - don't send partial-page writes before EOF from client (4410)
6049 - discard client grant for sub-page writes on large-page clients (4520)
6050 - don't free dentries not owned by NFS code, check generation (4806)
6051 - fix lsm leak if mds_create_objects() fails (4801)
6052 - limit debug_daemon file size, always print CERROR messages (4789)
6053 - use transno after validating reply (3892)
6054 - process timed out requests if import state changes (3754)
6055 - update mtime on OST during writes, return in glimpse (4829)
6056 - add mkfsoptions to LDAP (4679)
6057 - use ->max_readahead method instead of zapping global ra (5039)
6058 - don't interrupt __l_wait_event() during strace
6060 - add software watchdogs to catch hung threads quickly (4941)
6061 - make lustrefs init script start after nfs is mounted
6062 - fix CWARN/ERROR duplication (4930)
6063 - return async write errors to application if possible (2248)
6064 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6065 - print NAL number in %x format (4645)
6066 - update barely-supported suse-2.4.21-171 series (4842)
6067 - support for sles 9 %post scripts
6068 - support for building 2.6 kernel-source packages
6069 - support for sles km_* packages
6071 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6074 - ignore -ENOENT errors in osc_destroy (3639)
6075 - notify osc create thread that OSC is being cleaned up (4600)
6076 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6077 - reconstruct ost_handle() like mds_handle() (4657)
6078 - create a new thread to do import eviction to avoid deadlock (3969)
6079 - let lconf resolve symlinked-to devices (4629)
6080 - don't unlink "objects" from directory with default EA (4554)
6081 - hold socknal file ref over connect in case target is down (4394)
6082 - allow more than 32000 subdirectories in a single directory (3244)
6083 - fix blocks count for O_DIRECT writes (3751)
6084 - OST returns ENOSPC from object create when no space left (4539)
6085 - don't send truncate RPC if file size isn't changing (4410)
6086 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6087 - bind to privileged port in socknal and tcpnal (3689)
6089 - rate limit CERROR/CWARN console message to avoid overload (4519)
6090 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6091 - basic mmap support (3918)
6092 - kernel patch series update from b1_4 (4711)
6094 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6097 - avoid crash during MDS cleanup with OST shut down (2775)
6098 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6099 - don't use bad inodes on the MDS (3744)
6100 - dynamic object preallocation to improve recovery speed (4236)
6101 - don't hold spinlock over lock dumping or change debug flags (4401)
6102 - don't zero obd_dev when it is force cleaned (3651)
6103 - print grants to console if they go negative (4431)
6104 - "lctl deactivate" will stop automatic recovery attempts (3406)
6105 - look for existing locks in ldlm_handle_enqueue() (3764)
6106 - don't resolve lock handle twice in recovery avoiding race (4401)
6107 - revalidate should check working dir is a directory (4134)
6109 - don't always mark "slow" obdfilter messages as errors (4418)
6111 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6114 - don't close LustreDB during write_conf until it is done (3860)
6115 - fix typo in lconf for_each_profile (3821)
6116 - allow dumping logs from multiple threads at one time (3820)
6117 - don't allow multiple threads in OSC recovery (3812)
6118 - fix debug_size parameters (3864)
6119 - fix mds_postrecov to initialize import for llog ctxt (3121)
6120 - replace config semaphore with spinlock (3306)
6121 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6122 - don't allow enqueue to complete on a destroyed export (3822)
6123 - down write_lock before checking llog header bitmap (3825)
6124 - recover from lock replay timeout (3764)
6125 - up llog sem before sending rpc (3652)
6126 - reduce ns lock hold times when setting kms (3267)
6127 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6128 - fix NULL deref and obd_dev leak on setup error (3312)
6129 - replace some LBUG about llog ops with error handling (3841)
6130 - don't match INVALID dentries from d_lookup and spin (3784)
6131 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6132 - fix invalid assertion in ptlrpc_set_wait (3880)
6134 - add libwrap support for the TCP acceptor (3996)
6135 - add /proc/sys/portals/routes for non-root route listing (3994)
6136 - allow setting MDS UUID in .xml (2580)
6137 - print the stack of a process that LBUGs (4228)
6139 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6142 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6143 - make sure to unset replay flag from failed open requests (3440)
6144 - if default stripe count is 0, use OST count for inode size (3636)
6145 - update parent mtime/ctime on client for create/unlink (2611)
6146 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6147 - free recovery state on server during a forced cleanup (3571)
6148 - unregister_reply for resent reqs (3063)
6149 - loop back devices mounting and status check on 2.6 (3563)
6150 - fix resource-creation race that can provoke i_size == 0 (3513)
6151 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6152 - more debugging for page-accounting assertion (3746)
6153 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6154 - avoid deadlock after precreation failure (3758)
6155 - fix race and lock order deadlock in orphan handling (3450, 3750)
6156 - add validity checks when grabbing inodes from l_ast_data (3599)
6158 - add /proc/.../recovery_status to obdfilter (3428)
6159 - lightweight CDEBUG infrastructure, debug daemon (3668)
6160 - change default OSC RPC parameters to be better on small clusters
6161 - turn off OST read cache for files smaller than 32MB
6162 - install man pages and include them in rpms (3100)
6163 - add new init script for (un)mounting lustre filesystems (2593)
6164 - run chkconfig in %post for init scripts (3701)
6165 - drop scimac NAL (unmaintained)
6167 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6170 - clean kiobufs before and after use (3485)
6171 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6172 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6173 - change init script to fail more clearly if not run as root (1528)
6174 - allow clients to reconnect during replay (1742)
6175 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6176 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6177 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6178 - fix cond_resched (3554)
6179 - only evict unfinished clients after recovery (3515)
6180 - allow bulk resend, prevent data loss (3570)
6181 - dynamic ptlrpc request buffer allocation (2102)
6182 - don't allow unlinking open directory if it isn't empty (2904)
6183 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6184 - remove extraneous obd dereference causing LASSERT failure (3334)
6185 - don't use get_cycles() when creating temp. files on the mds (3156)
6186 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6187 - handle EEXIST for set-stripe, set proper directory name (3336)
6189 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6190 - fix ksocknal_fmb_callback() error messages (2918)
6192 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6195 - don't copy lvb into (possibly NULL) reply on error (2983)
6196 - don't deref dentry after dput, don't free lvb on error (2922)
6197 - use the kms to determine writeback rpc length (2947)
6198 - increment oti_logcookies when osc is inactive (2948)
6199 - update client's i_blocks count via lvb messages (2543)
6200 - handle intent open/close of special files properly (1557)
6201 - mount MDS with errors=remount-ro, like obdfilter (2009)
6202 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6203 - don't use cancelling-locks' kms values (2947)
6204 - use highest lock extent for kms, not last one (2925)
6205 - don't dereference ERR_PTR() dentry in error handling path (3107)
6206 - fix thread race in portals_debug_dumplog() (3122)
6207 - create lprocfs device entries at setup instead of at attach (1519)
6208 - common AST error handler, don't evict client on completion race (3145)
6209 - zero nameidata in detach_mnt in 2.6 (3118)
6210 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6211 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6212 - initialize RPC timeout timer earlier for 2.6 (3219)
6213 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6214 - print nal/nid for unknown nid (3258)
6215 - additional checks for oscc recovery before doing precreate (3284)
6216 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6217 - don't crash in mdc_close for bad permissions on open (3285)
6218 - zero i_rdev for non-device files (3147)
6219 - clear page->private before handing to FS, better assertion (3119)
6220 - tune the read pipeline (3236)
6221 - fix incorrect decref of invalidated dentry (2350)
6222 - provide read-ahead stats and refine rpc in flight stats (3328)
6223 - don't hold journal transaction open across create RPC (3313)
6224 - update atime on MDS at close time (3265)
6225 - close LDAP connection when recovering to avoid server load (3315)
6226 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6227 - don't leak open file on MDS after open resend (3325)
6228 - serialize filter_precreate and filter_destroy_precreated (3329)
6229 - loop device shouldn't call sync_dev() for nul device (3092)
6230 - clear page cache after eviction (2766)
6231 - resynchronize MDS->OST in background (2824)
6232 - refuse to mount the same filesystem twice on same mountpoint (3394)
6233 - allow llmount to create routes for mounting behind routers (3320)
6234 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6235 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6236 - fix rare problem with rename on htree directories (3417)
6238 - allow default OST striping configuration per directory (1414)
6239 - fix compilation for qswnal for 2.6 kernels (3125)
6240 - increase maximum number of MDS request buffers for large systems
6241 - change liblustreapi to be useful for external progs like lfsck (3098)
6242 - increase local configuration timeout for slow disks (3353)
6243 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6245 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6248 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6249 - don't overwrite extent policy data in reply if lock was blocked (2901)
6250 - drop filter export grants atomically with removal from device (2663)
6251 - del obd_self_export from work_list in class_disconnect_exports (2908)
6252 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6253 - swab reply message in mdc_close, other PPC fixes (2464)
6254 - fix destroying of named logs (2325)
6255 - overwrite old logs when running lconf --write_conf (2264)
6256 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6257 - fix race in target_handle_connect (2898)
6258 - mds_reint_create() should take same inode create lock (2926)
6259 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6260 - don't close files for self_export to avoid uninitialized obd (2936)
6261 - allow MDS with the same name as client node (2939)
6262 - hold dentry reference for closed log files for unlink (2325)
6263 - reserve space for all logs during transactions (2059)
6264 - don't evict page beyond end of stripe extent (2925)
6265 - don't oops on a deleted current working directory (2399)
6266 - handle hard links to targets without a parent properly (2517)
6267 - don't dereference NULL lock when racing during eviction (2867)
6268 - don't grow lock extents when lots of conflicting locks (2919)
6270 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6273 - account for cache space usage on clients to avoid data loss (974)
6274 - lfsck support in lustre kernel code (2349)
6275 - reduce journal credits needed for BRW writes (2370)
6276 - orphan handling to avoid losing space on client/server crashes
6277 - ptlrpcd can be blocked, stopping ALL progress (2477)
6278 - use lock value blocks to assist in proper KMS, faster stat (1021)
6279 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6280 - recovery for initial connections (2355)
6281 - fixes for mds_cleanup_orphans (1934)
6282 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6283 - block all file creations until orphan recovery completes (1901)
6284 - client remove rq_connection from request struct (2423)
6285 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6286 - recovery timer race (2670)
6287 - mdc_close recovey bug (2532)
6288 - ptlrpc cleanup bug (2710)
6289 - mds timeout on local locks (2588)
6290 - namespace lock held during RPCs (2431)
6291 - handle interrupted sync write properly (2503)
6292 - don't try to handle a message that hasn't been replied to (2699)
6293 - client assert failure during cleanup after abort recovery (2701)
6294 - leak mdc device after failed mount (2712)
6295 - ptlrpc_check_set allows timedout requests to complete (2714)
6296 - wait for inflight reqs when ptlrpcd finishes (2710)
6297 - make sure unregistered services are removed from the srv_list
6298 - reset bulk XID's when resending them (caught by 1138 test)
6299 - unregister_bulk after timeout
6300 - fix lconf error (2694)
6301 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6302 - readahead locks pages, leaves pending causing memory pressure (2673)
6303 - increase OST request buffers to 4096 on large machines (2729)
6304 - fix up permission of existing directories in simple_mkdir (2661)
6305 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6306 - don't assign transno to errored transactions (2742)
6307 - don't delete objects on OST if given a bogus objid from MDS (2751)
6308 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6309 - if rq_no_resend, then timeout request after recovery (2432)
6310 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6311 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6312 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6313 - MDS recovery shouldn't delete objects that it has given out (2730)
6314 - if enqueue arrives after completion, don't clobber LVB (2819)
6315 - don't unlock pages twice when trigger_group_io returns error (2814)
6316 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6317 - don't write pages to disk if there was an error (1450)
6318 - don't ping imports that have recovery disabled (2676)
6319 - take buffered bytes into account when balancing socknal conn (2817)
6320 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6321 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6322 - remove little-endian swabbing of llog records (1987)
6323 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6324 - retry reposting request buffers if they fail (1191)
6325 - grow extent at grant time to avoid granting a revoked lock (2809)
6326 - lock revoke doesn't evict page if covered by a second lock (2765)
6327 - disable VM readahead to avoid reading outside lock extents (2805)
6329 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6330 - updated kernel patches for hp-2.4.20 kernel (2681)
6332 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6335 - fix truncated write corruption (2366)
6336 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6338 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6339 - ensure len(uuid) < 37 in lmc (1171)
6340 - fix ia64 OOPS in llog_test (2255)
6341 - zero end of page at obdfilter for partial page writes (2648)
6342 - don't leave stale dentries around after renames (bug 2428)
6343 - fix timeouts when evicting a client with a single lock held (2642)
6344 - set deadline for the initial HELLO message to drain (2634)
6345 - print out dotted-quad IP addresses in the socknal (2302)
6347 - additional debugging for MDS client eviction problem (2443)
6348 - fix mkfsoptions support for osts (2603, 2604)
6350 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6353 - add series for the vanilla 2.6.0 kernel
6354 - add series for the vanilla 2.4.24 kernel
6355 - add series for a cray x86/64 UL kernel drop
6356 - fix xattr patches for the vanilla 2.4.19 series
6358 - generate true UUIDs in lmc (1171)
6359 - have portals stack dumping break in UML (2466)
6360 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6361 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6362 - wake callback waiting threads on client eviction (2460)
6363 - Add --ptldebug and --subsystem to lmc (1719)
6364 - update assertion to allow safe interrupt allocation
6365 - set rq_no_resend for cancel requests (2432)
6366 - recalculate ptlrpcd timeout after resend (2494)
6367 - call vfs_rmdir when removing pending directories (2368)
6368 - fix renaming a file to itself (2429)
6369 - lmc creates a default one-stripe lov (2454)
6370 - expand procfs space to handle large clusters (2326)
6371 - increase UML stack to avoid overflow
6372 - update lconf's list of debug and subsystem masks
6373 - fix lfs find --obd (2510)
6374 - /proc tunable for disabling filter read caching (2591)
6375 - stop rpm packages from altering slapd.conf (2301)
6376 - disable nagle in the socknal under 0conf (2578)
6377 - choose mds inode size based on stripe count (2572)
6378 - fix kernel-source rpm problems (2516)
6380 - add --disable-doc to avoid pdf generation (2421)
6381 - update documentation, tests, type-os, comments
6382 - avoid format warnings on ia64
6383 - remove the TOE NAL
6384 - tiny code cleanups by removing unused fields
6386 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6389 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6390 - remove the most blatant lies from BUILDING file (2371)
6391 - change default debug level to reasonable production setting
6392 - reduce client side cache size to reduce cache flush time
6393 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6394 - make TCP zerocopy and pinger support enabled by default (2476)
6395 - sync writes completed after process exits caused crashes (2319)
6396 - maintain correct mount count on the MDS (2356)
6397 - backout 1557, because 2316 wasn't really fixed
6398 - better file I/O statistics gathering in /proc
6399 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6400 - another kernel patch to fix zero-copy TCP function export
6401 - don't take duplicate lock when processing re-sent getattr (2420)
6402 - lctl uses obd_self_export instead of creating new conn (2353)
6403 - MDS/OST recovery case which requires object creation asserted (2425)
6404 - move lfs from /usr/sbin to /usr/bin in packages
6405 - fix race between mds_client_add and mds_client_free (2417)
6406 - use kmalloc instead of slabs in portals (2430)
6407 - don't create duplicate records when a failover MDS is present (2442)
6408 - remove unnecessary mount age check (2332)
6409 - don't remove directory inodes from locks prematurely (2451)
6410 - don't break if MDS service name is the same as hostname (2103)
6411 - fix races in client write RPC generation when cache full (2482)
6413 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6416 - remove now-unused request->rq_obd (278)
6417 - if an allocation fails, print out how much memory we've used (1933)
6418 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6419 - add the "configurable stack size" patch to most series files (1256)
6420 - ability to write large log records, for 100+ OST configs (2306)
6421 - fix NULL deref when filter_prep fails (2314)
6422 - fix operator precedence error in filter_sync
6423 - dynamic allocation of socknal TX descriptors (2315)
6424 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6425 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6426 - ensure that CWARN messages in Portals always get to the syslog
6427 - __init/__exit are not for prototype decls (ldlm_init/exit)
6428 - x86-64 compile warning fixes
6429 - fix gateway LMC keyword conflict (2318)
6430 - fix MDS lock inversions in getattr/reint paths (1844)
6431 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6432 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6433 - DLM race condition prevented some lock evictions (2328)
6434 - ENOMEM detection and retry on socknal sends (2230)
6435 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6436 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6437 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6438 - better error messages when a client is rejected during recovery (1505)
6439 - avoid cancelling locks which were never granted, after failure (2330)
6440 - fix i_sem/journal inversion in mds_client_add (2333)
6441 - fix truncate/getattr lock cycle deadlock (2334)
6442 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6443 - fix two rare exit paths which could leak an l_lock() ref (2321)
6444 - fencepost error in MDS/OST orphan recovery (2226)
6445 - make log record alignment 8 bytes (1988)
6446 - lstripe now fails when requested offset > ost_count (2237)
6447 - ensure that all kernel series have a complete list.h (1607)
6448 - fix crashes in special-file operations (2316)
6449 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6451 - allow configurable automake binary, for testing new versions
6452 - small update to the lfs documentation
6454 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6456 * fix negative export reference count in fsfilt_sync (2312)
6458 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6459 * release candidate 0.9.1
6461 - orphans are moved into the PENDING directory for possible recovery
6462 - replayed opens now open by fid for orphan/rename safety (1042)
6463 - last close of an orphan inode generates a transno (683)
6464 - chdir() and mount() now pin the directory entry (1020)
6465 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6466 - discard very old requests without processing them (1502)
6467 - remove some common, well-understood CERRORs (1505)
6468 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6469 - clear "grant" flags in OST replies until OST grant code lands (1644)
6470 - fix read performance by not clobbering i_blksize on client (1598)
6471 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6472 - make LRU size a /proc tunable, clears locks when reduced (707)
6473 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6474 - server locks take references on exports now (1558)
6475 - build fixes for 2.4.20-rh trees (1663)
6476 - return an error from lov_create if all OSCs are inactive (1751)
6477 - fix import levels when a reconnect happens without a timeout (1597)
6478 - exit early from mds_open if we get a lookup error (1749)
6479 - partial page read at EOF wouldn't wait for disk before sending (1642)
6480 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6481 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6482 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6483 - write/truncate lock inversion (1639)
6484 - Don't auto-load obdclass, portals modules during cleanup (1495)
6485 - fix timestamps from jumping to "now" (1763)
6486 - extra journal assertions (1648)
6487 - add an extra multiunlink test (1771)
6488 - fix read_record/write_record API (1776)
6489 - fix leak of offset_extent, possible incorrect i_size later (1772)
6490 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6491 - Debugging for the kqswnal_get_idle_tx problems (1820)
6492 - Allow recovery to be attempted multiple times (1536)
6493 - Write out MDS last_rcvd file after it is first created (1600)
6494 - Fix tx_descriptor leak in failed transmit situations (1827)
6495 - ext3 journaling fixes for assertion failure after IO error (1871)
6496 - class_export_put() on freed export after completion AST error (1896)
6497 - Fix revalidate looping in VFS (1322)
6498 - Don't access a freed export during MDS_REINT timeout (1521)
6499 - Add open-unlink recovery support on the MDS (1673,1764)
6500 - Return an error if no MDS data was read from last_rcvd (1946)
6501 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6502 - Don't setup a disk device that doesn't match exported UUID (317)
6503 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6504 - avoid committing NULL handle in force close
6505 - local.sh is now a one-stripe LOV configuration
6506 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6507 - fix ext3 htree duplicate directory entry corruption (1516)
6508 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6509 - update to new LOV EA format (2097)
6510 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6512 2003-06-15 Phil Schwan <phil@clusterfs.com>
6515 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6516 - per-import/export recovery handling (958, 931, 959)
6517 - multiple last-rcvd slots, for serving multiple FSes (949)
6518 - connections are again shared between multiple imp/exports (963, 964)
6519 - "umount -f" would hang if any requests needed to be sent (393, 978)
6520 - avoid pinning large req buffer by copying for queued messages (989)
6521 - add "uuid" to "lctl device" command to help upcalls (991)
6522 - "open" RPCs with transnos would confuse recovery counters (1037)
6523 - do proper endian conversion of all wire messages (288, 340, 891)
6524 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6525 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6526 - fix ptlrpc_connection leak in target_handle_connect (1174)
6527 - fix import refcounting bug in OST and MDS cleanup (1134)
6528 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6529 - fix very unlikely obd_types race condition (501)
6530 - remove osc_open hack for echo_client (1187)
6531 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6532 - a failure in read_inode2 leads to deadlock (1139)
6533 - cancel ack-locks as soon as transaction is committed (1072)
6534 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6535 - make sure to commitrw after any preprw to avoid deadlock (1162)
6536 - failing to execute a file in a lustre FS would lock inode (1203)
6537 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6538 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6539 - obd_uuid2tgt would walk off the end of the list (1255)
6540 - on IA64 the osc would give portals incorrect bulk size (1258)
6541 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6542 - fix lock inversion caused by new llite matching code (1282)
6543 - limit the number of dirty pages on a client to 10MB (1286)
6544 - timed out locks were not being corrected cancelled (1289)
6545 - fix O_DIRECT above 4GB on IA-32 (1292)
6546 * major user-visible changes
6547 - fail out/fail over policy now controlled by the upcall (993)
6549 - add OBD_PING to check server availability and failure (954)
6550 - lustre messages are now sent in sending host order (288, 340, 891)
6551 - add eadatalen to MDS getattr reply (340)
6552 - OST read replies may contain second buffer, with per-page status (593)
6554 2003-03-11 Phil Schwan <phil@clusterfs.com>
6557 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6558 - failed open()s could cause deadlock; fixed (867, 869)
6559 - stop cancelling OST locks when files are closed (481)
6560 - overlapping XID spaces caused network corruption (851, 853)
6561 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6562 - setattr_raw added, to do single-RPC, server-side setattrs
6563 - lmc/lconf syntax change for OST UUIDs
6564 - fix crashy race condition between ptlrpc_free_req and osc_close
6565 - don't use request in mdc_enqueue if we hit a timeout (889)
6566 - don't set the inode i_size for regular files from the MDS (896)
6567 - handle out of order completion AST (842)
6568 - don't LBUG if a lock request times out after receiving AST (913)
6569 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6570 - if a bad lock AST arrives, send an error instead of dropping entirely
6571 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6572 - fix leak in bulk IO when only partially completed (899, 900, 926)
6573 - fix O_DIRECT for ia64 (55)
6574 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6575 - C-z after timeout could hang a process forever; fixed (977)
6577 - client-side I/O cache (678, 924, 929, 941, 970)
6579 - READPAGE and SETATTRs which don't take server-side locks get
6582 2003-02-11 Phil Schwan <phil@clusterfs.com>
6585 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6586 - Fix fsfilt_extN_readpage() to read a full page of directory
6587 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6588 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6589 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6590 - Most intent operations were removed, in favour of a new RPC mode
6591 that does a single RPC to the server and bypasses most of the VFS
6592 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6593 - Aggressively cancel local locks on DLM servers
6594 - mds_reint_unlink sends EA to the client if it's the last nlink.
6595 client uses that EA to unlink OST objects.
6596 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6597 - recursive symlinks were fixed (439)
6598 - fixed NULL deref in DEBUG_REQ
6599 - filter_update_lastobjid no longer calls sync, which annoyed extN
6600 - fixed multi-client small-writes to a single file problem (445)
6601 - fixed mtime updates during file writes (607)
6602 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6603 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6604 - fixed Linux OST size reporting problem (444, 656)
6605 - OST now updates object mtime with writes or setattr (607, 619)
6606 - client verifies file size before zeroing page past EOF (445)
6607 - OST now writes last allocated objid to disk with allocation (108)
6608 - LOV on echo now works (409)
6610 - mds_reint_unlink sends a new buffer, with the EA included. this
6611 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6612 set if a regular file was being unlinked, and it was the last link
6613 - use PtlGet from the target for bulk writes (315)
6614 - OST now updates object mtime with writes or setattr (607, 619)
6615 - LDLM now has a grant-time callback to revalidate locked items, if
6617 - Many MDS operations were reorganized to combat race conditions
6619 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6620 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6622 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6625 - Fully reactivate OST imports after reconnection (512, others)
6626 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6627 - More graceful error handling for truncating on dead OST (515)
6628 - Don't error out unless we're actually accessing dead stripes (474)
6629 - Fix garbage sizes when stripes are missing (410)
6630 - LRU counters were broken, causing constant lock purge (433, 432)
6631 - garbage on read from stripes with failed OSTs (441)
6632 - mark OSCs as active before reconnecting during recovery (438)
6633 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6634 - lfind did not preserve OST order in output (443)
6635 - symlinks cause hung clients, incorrect data (439)
6636 - stop dereferencing request after dropping refcount (457)
6637 - don't LASSERT(spin_is_locked) on non-SMP (455)
6638 - fixes for many rename() bugs
6639 - fstat didn't correctly synchronize attributes (399)
6640 - server must handle lock cancellation during blocking AST prep (487)
6641 - bulk descriptors were free()d too soon (511)
6642 - fix paths in lconf, which would load incorrect modules (451, 507)
6643 - fix confusing lconf 'host not found' error message (386)
6644 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6645 - fix race condition in mdc_blocking_ast() for inode access (526)
6646 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6647 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6648 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6649 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6650 - fix lstripe in several ways (kernel side) (527)
6651 - fix request leak in ldlm_cli_enqueue (262)
6652 - incorrect OSC was marked inactive after OST failure
6653 - call mds_fs_cleanup before unmounting filesystem (524)
6654 - fix races between taking ns_lock and ldlm_lock_change_resource
6655 - fix races updating LOV export open file list
6656 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6657 - fix recovery NULL deref in ldlm_cli_cancel_unused
6658 - fix some DLM races by using new hash table for lock handles (419)
6659 - permit the client to specify desired inodes, at replay
6660 - duplicate requests when we queue them for replay reintegration
6661 - fix last_rcvd offset calculation
6662 - sync after each recovered transaction, so we always make progress
6663 - never, not always, ERESTART requests without transnos
6664 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6666 - skip replay if the MDS says that the client is already connected
6667 - don't check for a recovery-enabled export to match lctl's UUID
6668 - don't INC_USE_COUNT for phantom exports
6669 - don't crash when cleaning up phantom exports (567)
6670 - don't double-finish or set replay data for errored mdc_open requests
6671 - abort requests when they time out, so we don't get old replies
6672 - send/receive replies for AST messages again
6673 - if the client says that it doesn't have the lock, cancel it on the
6675 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6676 mark it as destroyed, it will all work out when decref is called
6677 - fix module use counts (22, 581)
6679 - ASTs now expect a reply (server cancels lock on error reply)
6681 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6684 - fix many simultaneous client startup (392)
6685 - fix dentry->d_it clobbering
6686 - credentials weren't being shipped for readdir/getattr operations
6687 - remove invalid assertions triggered during some concurrent MD
6689 - proper Lustre versions added (336, 389)
6690 - fix memory leak for create error case (398)
6691 - fix LOV locking bug that would get cli/srv out of sync
6692 - fix echo client over LOV (409)
6693 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6694 - fix double-O_EXCL intent crash (424)
6695 - avoid sending multiple lock CANCELs (352)
6697 - MDS can do multi-client recovery (modulo bugs in new code)
6699 - many updates, edits, cleanups
6701 2002-11-18 Phil Schwan <phil@clusterfs.com>
6704 - fix null d_it dereference (346)
6705 - fix full OST/dbench hang (333)
6706 - fix permission problem with file removal (286)
6707 - fix removal of OSCs from LOV when they fail
6708 - fix NULL deref during bulk timeout (214)
6709 - fix problems related to multiple filesystems on one MDS (241)
6710 - fixed serious subtle metadata locking bugs
6711 - free locks on clients when inodes are removed due to memory
6713 - fix inode pointer in lock data (285)
6714 - partial support for multiple MDS on a single host (241)
6715 - data locks weren't cancelled at clear_inode time (290, 311)
6716 - intent locks could lead to unbounded lock growth (205)
6717 - added a maximum lock count, an LRU list, and a flusher
6718 - fix multiple rename (365)
6719 - properly abstracted the echo client
6720 - OSC locked 1 byte too many; fixed
6721 - rewrote brw callback code:
6722 - fixed recovery bugs related to LOVs (306)
6723 - fixed too-many-pages-in-one-write crash (191)
6724 - fixed (again) crash in sync_io_timeout (214)
6725 - probably fixed callback-related race (385)
6727 - Add capability to MDS protocol
6728 - LDLM cancellations and callbacks on different portals
6730 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6733 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6734 - defer open object destruction to close time (601981, 138)
6735 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6736 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6737 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6738 - don't call obd_setattr in ll_file_release for destroyed objects
6739 * protocol change to lustre_msg: move |version| and add |flags|
6740 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6741 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6742 * added replay of create, unlink, link and rename operations during
6743 MDS failover; recovery should be much more robust now
6744 * remove failed OSCs from LOVs (only lov_create uses this so far)
6745 * the lustre-HOWTO was brought (more) up to date (582544)
6747 2002-10-23 Phil Schwan <phil@clusterfs.com>
6750 - in-use dentries weren't being reused properly (617851)
6751 - prevent multiple LDLM setup (599178)
6752 - fix LOV size calculations for truncate (617853)
6753 - fix client handling of MDS intent errors (POSIX)
6754 - fix permission bug in lovstripe.c test (624321)
6755 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6756 - truncate past end of file could corrupt data
6757 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6758 - a race in recovery could return ETIMEDOUT to apps (623947)
6759 - building outside the source directory was fixed
6760 * the lustre-HOWTO was brought (more) up to date (582544)
6761 * major progress was made on recovery functionality
6763 2002-10-10 Phil Schwan <phil@clusterfs.com>
6766 - recovery deadlock fix
6767 - rm -rf causes LBUG fix (617817)
6768 - file open by multiple tasks fix (618962)
6769 - directory permissions bugs (602707 and 620007)
6770 - journal_stop fixed with locking (611313)
6771 - O_APPEND failures resolved (618273, perhaps 614459)
6772 - lconf PATH fix (619770)
6773 - IA64 build fix (621450)
6774 - RPC buffer sizes scale with amount of memory
6776 2002-10-01 Phil Schwan <phil@clusterfs.com>
6779 - locks would be cancelled without throwing away data pages,
6780 resulting in inconsistent data (605627)
6781 - inode attributes were not always being refreshed (605627, 612449)
6782 - lconf now continues to cleanup after lctl reports an error
6783 - MDS now enforces user permissions (602707)
6784 - lprocfs cleanup fixed, but not yet enabled (614157)
6785 - fixed infinite server hang, should a client not respond to an AST
6786 - avoid going into recovery if user calls readlink() with a buffer
6787 that's too small (613941)
6788 - AST RPCs no longer require replies (614867) -- this may be changed
6789 - don't crash server if client sends an IOV that's too big (611336)
6790 - fixed lock conversion deadlock (611892)
6791 - fixed the following of symlinks (614622)
6792 * recovery: the server can remove locks from a client that dies, other
6793 clients can make progress
6794 * more extN patch fixes
6795 * compile-time configurable ptlrpc buffer allocations
6797 - collaborative read cache document
6798 - Lustre Lite Performance CDR document-in-progress
6800 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6803 - fix typo in patch-2.4.18
6805 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6808 - clear ptlrpc request each time in handle_incoming_request()
6809 - unlink of files now destroys the object on the OST
6811 2002-09-19 Peter Braam <braam@clusterfs.com>
6813 * add hard link support
6814 * change obdfile creation method
6815 * kernel patch changed
6817 2002-09-19 Peter Braam <braam@clusterfs.com>
6820 - stack overflow bug in extN fixed
6822 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6824 * documentation updates
6825 - add man pages for config tools
6826 - update tests/README to describe testing with new config tools
6827 - finish metadata API descriptions
6828 * bug fixes and cleanups
6829 - statfs workaround for 16TB limit
6830 - LOV stripe allocation improved, can stripe on subset of OSTs
6831 - LOV file size/IO offset was wrong for files > 4GB in size
6832 - object EA data was being dropped, caused files to be unreadable
6833 - memory overflow with non-LOV OST caused memory corruption
6834 - fixed regression tests to work with new config tools, obdfilter
6835 - fixed bug when directory size became larger than 1 block
6836 - fixed bug (for single client case) when PWD was deleted
6837 - invalidate local directory pages when doing intent-based ops
6838 - avoid LDLM oops when lock callback contained bad data
6840 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6842 * documentation updates
6843 * bug fixes and cleanups
6844 - configuration tools
6847 - 64-bit compile warnings
6848 - 64-bit internal statfs data
6850 * test_brw on persistent OST devices
6852 * lprocfs (disabled)
6854 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6856 * documentation updates
6857 * bug fixes and cleanups
6858 * configuration tools
6860 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6863 * many small fixes to 0_5_4
6864 * io/network handling
6865 * thinkos in MDS operations
6867 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6870 * crucial basic fixes to 0.5.3
6871 * IOR, Iozone work over Elan
6874 2002-08-07 Phil Schwan <phil@clusterfs.com>
6875 * version 0_5_3, our first alpha
6876 * we use the new Portals iovs
6877 * documentation updates
6878 * bug fixes and cleanups
6879 * small changes in the DLM wire protocol
6881 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6882 * version 0_5_1 with some initial stability,
6883 * locking on MD and file I/O.
6884 * documentation updates
6885 * several bug fixes since 0.5.0
6886 * small changes in wire protocol
6888 2002-07-18 Phil Schwan <phil@clusterfs.com>
6890 * delivered as Lustre Light Alpha
6891 * fixed a crash after handling invalid MDS requests
6892 * fixed directory pages for architectures with non-4k pages sizes
6894 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6895 * release version v0_4_4
6896 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6898 * Fixes a number of interruption problems with OST operations.
6899 * Update documentation for portals header changes
6900 * Move all wire protocol structs/defines to lustre_idl.h
6901 * Fixes symlink length bug.
6902 * Add tcpdump to repository.
6904 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6905 * release version v0_4_3
6906 * Fixes statfs for inodes on extN.
6907 * Fixes bug in runtests which would delete /etc/hosts.
6908 * Use 64-bit object IDs wherever possible (not into VFS though)
6909 Remove ost_get_info, which is unused by lustre, and out of date.
6911 2002-07-03 Peter Braam <braam@clusterfs.com>
6912 * release version v0_4_2 Fixes a lookup error (type not passed)
6913 * move forward to head of Portals
6914 * move forward to latest Lustre kernel
6916 2002-06-25 Peter Braam <braam@clusterfs.com>
6917 * release version v0_4_1. Hopefully stable on single node use.