3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server primary kernels built and tested during release cycle:
6 5.14.0-284.30.1.el9 (RHEL9.2)
7 5.14.0-162.23.1.el9 (RHEL9.1)
8 4.18.0-477.27.1.el8 (RHEL8.8)
9 * Other server kernels known to build and work at some point (others may also work):
10 3.10.0-862.14.4.el7 (RHEL7.5)
11 3.10.0-957.27.2.el7 (RHEL7.6)
12 3.10.0-1062.18.1.el7 (RHEL7.7)
13 3.10.0-1127.19.1.el7 (RHEL7.8)
14 3.10.0-1160.88.1.el7 (RHEL7.9)
15 4.18.0-80.11.2.el8 (RHEL8.0)
16 4.18.0-147.8.1.el8 (RHEL8.1)
17 4.18.0-193.28.1.el8 (RHEL8.2)
18 4.18.0-240.22.1.el8 (RHEL8.3)
19 4.18.0-305.25.1.el8 (RHEL8.4)
20 4.18.0-348.23.1.el8 (RHEL8.5)
21 4.18.0-372.32.1.el8 (RHEL8.6)
22 4.18.0-425.10.1.el8 (RHEL8.7)
23 4.4.120-92.70 (SLES12 SP2)
24 4.4.180-94.100 (SLES12 SP3)
25 4.4.0-131 (Ubuntu 16.04)
26 4.15.0-32 (Ubuntu 18.04)
27 5.4.0-48 (Ubuntu 20.04)
28 vanilla linux 5.4.0 (ZFS + ldiskfs)
29 vanilla linux 5.4.21 (ZFS + ldiskfs)
30 vanilla linux 5.4.136 (ZFS + ldiskfs)
31 5.10.0-60.94.0.118.oe2203 (openEuler 22.03 LTS)
32 5.10.0-136.32.0.108.oe2203sp1 (openEuler 22.03 LTS SP1)
33 5.10.0-153.19.0.95.oe2203sp2 (openEuler 22.03 LTS SP2)
34 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
35 * Client primary kernels built and tested during release cycle:
36 5.14.0-284.30.1.el9 (RHEL9.2)
37 5.14.0-162.23.1.el9 (RHEL9.1)
38 4.18.0-477.27.1.el8 (RHEL8.8)
39 5.4.0-37 (Ubuntu 20.04)
40 5.14.21-150500.55.22 (SLES15 SP5)
41 * Other clients known to build on these kernels at some point (others may also work):
42 3.10.0-862.14.4.el7 (RHEL7.5)
43 3.10.0-957.27.2.el7 (RHEL7.6)
44 4.14.0-49.13.1.el7a (RHEL7.5)
45 4.14.0-115.2.2.el7a (RHEL7.6)
46 3.10.0-1062.18.1.el7 (RHEL7.7)
47 3.10.0-1127.19.1.el7 (RHEL7.8)
48 3.10.0-1160.88.1.el7 (RHEL7.9)
49 4.18.0-80.11.2.el8 (RHEL8.0)
50 4.18.0-147.8.1.el8 (RHEL8.1)
51 4.18.0-193.28.1.el8 (RHEL8.2)
52 4.18.0-240.22.1.el8 (RHEL8.3)
53 4.18.0-305.25.1.el8 (RHEL8.4)
54 4.18.0-348.23.1.el8 (RHEL8.5)
55 4.18.0-372.32.1.el8 (RHEL8.6)
56 4.18.0-425.10.1.el8 (RHEL8.7)
57 5.14.0-70.30.1.el9 (RHEL9.0)
58 4.4.120-92.70 (SLES12 SP2)
59 4.4.180-94.100 (SLES12 SP3)
60 4.12.14-95.48 (SLES12 SP4)
61 4.12.14-122.133 (SLES12 SP5)
62 4.12.14-197.75 (SLES15 SP1)
63 5.3.18-24.96 (SLES15 SP2)
64 5.3.18-150300.59.93 (SLES15 SP3)
65 5.14.21-150400.24.46 (SLES15 SP4)
66 4.4.0-131 (Ubuntu 16.04)
67 4.15.0-48 (Ubuntu 18.04)
68 5.8.0-53 (Ubuntu 20.04.2 HWE)
69 5.11.0-31 (Ubuntu 20.04.3 HWE)
70 5.11.0 (vanilla kernel.org)
71 5.10.0 (openEuler 22.03 LTS)
72 * Recommended e2fsprogs version: 1.47.0-wc1 or newer
73 * Recommended ZFS version: 2.1.11
74 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
75 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
76 stack overflow. For more information, please refer to bugzilla 17630.
77 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
78 the re-exporting nodes to be mounted with "32bitapi" mount option
80 --------------------------------------------------------------------------------
85 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
86 for currently supported client and server kernel versions.
87 * Server primary kernels built and tested during release cycle:
88 4.18.0-348.2.1.el8 (RHEL8.5)
89 * Other server kernels known to build and work at some point (others may also work):
90 3.10.0-862.14.4.el7 (RHEL7.5)
91 3.10.0-957.27.2.el7 (RHEL7.6)
92 3.10.0-1062.18.1.el7 (RHEL7.7)
93 3.10.0-1127.19.1.el7 (RHEL7.8)
94 3.10.0-1160.49.1.el7 (RHEL7.9)
95 4.18.0-80.11.2.el8 (RHEL8.0)
96 4.18.0-147.8.1.el8 (RHEL8.1)
97 4.18.0-193.28.1.el8 (RHEL8.2)
98 4.18.0-240.22.1.el8 (RHEL8.3)
99 4.18.0-305.25.1.el8 (RHEL8.4)
100 4.4.120-92.70 (SLES12 SP2)
101 4.4.180-94.100 (SLES12 SP3)
102 4.4.0-131 (Ubuntu 16.04)
103 4.15.0-32 (Ubuntu 18.04)
104 5.4.0-48 (Ubuntu 20.04)
105 vanilla linux 5.4.0 (ZFS + ldiskfs)
106 vanilla linux 5.4.21 (ZFS + ldiskfs)
107 vanilla linux 5.4.136 (ZFS + ldiskfs)
108 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
109 * Client primary kernels built and tested during release cycle:
110 4.18.0-348.2.1.el8 (RHEL8.5)
111 5.4.0-37 (Ubuntu 20.04)
112 5.3.18-59.27 (SLES15 SP3)
113 * Other clients known to build on these kernels at some point (others may also work):
114 3.10.0-862.14.4.el7 (RHEL7.5)
115 3.10.0-957.27.2.el7 (RHEL7.6)
116 4.14.0-49.13.1.el7a (RHEL7.5)
117 4.14.0-115.2.2.el7a (RHEL7.6)
118 3.10.0-1062.18.1.el7 (RHEL7.7)
119 3.10.0-1127.19.1.el7 (RHEL7.8)
120 3.10.0-1160.49.1.el7 (RHEL7.9)
121 4.18.0-80.11.2.el8 (RHEL8.0)
122 4.18.0-147.8.1.el8 (RHEL8.1)
123 4.18.0-193.28.1.el8 (RHEL8.2)
124 4.18.0-240.22.1.el8 (RHEL8.3)
125 4.18.0-305.25.1.el8 (RHEL8.4)
126 4.4.120-92.70 (SLES12 SP2)
127 4.4.180-94.100 (SLES12 SP3)
128 4.12.14-95.48 (SLES12 SP4)
129 4.12.14-122.91 (SLES12 SP5)
130 4.12.14-197.75 (SLES15 SP1)
131 5.3.18-24.96 (SLES15 SP2)
132 4.4.0-131 (Ubuntu 16.04)
133 4.15.0-48 (Ubuntu 18.04)
134 5.8.0-53 (Ubuntu 20.04.2 HWE)
135 5.11.0-31 (Ubuntu 20.04.3 HWE)
136 5.11.0 (vanilla kernel.org)
137 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
138 * Recommended ZFS version: 2.0.7
139 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
140 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
141 stack overflow. For more information, please refer to bugzilla 17630.
142 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
143 the re-exporting nodes to be mounted with "32bitapi" mount option
145 --------------------------------------------------------------------------------
149 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
150 for currently supported client and server kernel versions.
151 * Server primary kernels built and tested during release cycle:
152 3.10.0-1062.18.1.el7 (RHEL7.7)
153 3.10.0-1127.19.1.el7 (RHEL7.8)
154 3.10.0-1160.6.1.el7 (RHEL7.9)
155 * Other server kernels known to build and work at some point (others may also work):
156 3.10.0-862.14.4.el7 (RHEL7.5)
157 3.10.0-957.27.2.el7 (RHEL7.6)
158 4.18.0-80.11.2.el8 (RHEL8.0)
159 4.18.0-147.8.1.el8 (RHEL8.1)
160 4.18.0-193.28.1.el8 (RHEL8.2)
161 4.18.0-240.1.1.el8 (RHEL8.3)
162 4.4.120-92.70 (SLES12 SP2)
163 4.4.180-94.100 (SLES12 SP3)
164 4.4.0-131 (Ubuntu 16.04)
165 4.15.0-32 (Ubuntu 18.04)
166 5.4.0-48 (Ubuntu 20.04)
167 vanilla linux 5.4.0 (ZFS + ldiskfs)
168 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
169 * Client primary kernels built and tested during release cycle:
170 3.10.0-1062.18.1.el7 (RHEL7.7)
171 3.10.0-1127.19.1.el7 (RHEL7.8)
172 3.10.0-1160.6.1.el7 (RHEL7.9)
173 4.12.14-95.48 (SLES12 SP4)
174 4.12.14-122.20 (SLES12 SP5)
175 4.12.14-197.75 (SLES15 SP1)
176 5.3.18-24.24 (SLES15 SP2)
177 4.15.0-48 (Ubuntu 18.04)
178 * Other clients known to build on these kernels at some point (others may also work):
179 3.10.0-862.14.4.el7 (RHEL7.5)
180 3.10.0-957.27.2.el7 (RHEL7.6)
181 4.14.0-49.13.1.el7a (RHEL7.5)
182 4.14.0-115.2.2.el7a (RHEL7.6)
183 4.18.0-80.11.2.el8 (RHEL8.0)
184 4.18.0-147.8.1.el8 (RHEL8.1)
185 4.18.0-193.28.1.el8 (RHEL8.2)
186 4.18.0-240.1.1.el8 (RHEL8.3)
187 4.4.120-92.70 (SLES12 SP2)
188 4.4.180-94.100 (SLES12 SP3)
189 4.4.0-131 (Ubuntu 16.04)
190 5.4.0-37 (Ubuntu 20.04)
191 5.4.0 (vanilla kernel.org)
192 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
193 * Recommended ZFS version: 2.0.0
194 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
195 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
196 stack overflow. For more information, please refer to bugzilla 17630.
197 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
198 the re-exporting nodes to be mounted with "32bitapi" mount option
200 --------------------------------------------------------------------------------
204 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
205 for currently supported client and server kernel versions.
206 * Server primary kernels built and tested during release cycle:
207 3.10.0-1062.1.1.el7 (RHEL7.7)
208 * Other server kernels known to build and work at some point (others may also work):
209 3.10.0-862.14.4.el7 (RHEL7.5)
210 3.10.0-957.27.2.el7 (RHEL7.6)
211 4.18.0-80.7.1.el8 (RHEL8.0)
212 4.4.120-92.70 (SLES12 SP2)
213 4.4.180-94.100 (SLES12 SP3)
214 4.4.0-131 (Ubuntu 16.04)
215 4.15.0-32 (Ubuntu 18.04)
216 vanilla linux 4.6.7 (ZFS only)
217 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
218 * Client primary unpatched kernels built and tested during release cycle:
219 3.10.0-1062.1.1.el7 (RHEL7.7)
220 4.12.14-95.29 (SLES12 SP4)
221 4.15.0-48 (Ubuntu 18.04)
222 * Other clients known to build on unpatched kernels at some point (others may also work):
223 3.10.0-862.14.4.el7 (RHEL7.5)
224 3.10.0-957.27.2.el7 (RHEL7.6)
225 4.14.0-49.13.1.el7a (RHEL7.5)
226 4.14.0-115.2.2.el7a (RHEL7.6)
227 4.18.0-80.7.1.el8 (RHEL8.0)
228 4.4.120-92.70 (SLES12 SP2)
229 4.4.180-94.100 (SLES12 SP3)
230 4.4.0-131 (Ubuntu 16.04)
231 vanilla linux 4.15.0 (ZFS only)
232 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
233 * Recommended ZFS / SPL version: 0.7.13
234 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
235 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
236 stack overflow. For more information, please refer to bugzilla 17630.
237 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
238 the re-exporting nodes to be mounted with "32bitapi" mount option
239 * Note for Ubuntu 14.04.5:
240 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
241 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
243 --------------------------------------------------------------------------------
247 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
248 for currently supported client and server kernel versions.
249 * Server known to build on the following kernels (others may also work):
250 2.6.32-696.18.7.el6 (RHEL6.9)
251 3.10.0-862.14.4.el7 (RHEL7.5)
252 3.10.0-957.el7 (RHEL7.6)
253 4.4.120-92.70 (SLES12 SP2)
254 4.4.162-94.69 (SLES12 SP3)
255 4.4.0-131 (Ubuntu 16.04)
256 4.15.0-32 (Ubuntu 18.04)
257 vanilla linux 4.6.7 (ZFS only)
258 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
259 * Client known to build on unpatched kernels (others may also work):
260 2.6.32-696.18.7.el6 (RHEL6.9)
261 2.6.32-754.6.3.el6 (RHEL6.10)
262 3.10.0-862.14.4.el7 (RHEL7.5)
263 3.10.0-957.el7 (RHEL7.6)
264 4.14.0-49.13.1.el7a (RHEL7.5)
265 4.14.0-115.2.2.el7a (RHEL7.6)
266 4.4.120-92.70 (SLES12 SP2)
267 4.4.162-94.69 (SLES12 SP3)
268 4.4.0-131 (Ubuntu 16.04)
269 4.15.0-32 (Ubuntu 18.04)
270 vanilla linux 4.15.0 (ZFS only)
271 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
272 * Recommended ZFS / SPL version: 0.7.9
273 * Tested with ZFS / SPL version: 0.7.9
274 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
275 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
276 stack overflow. For more information, please refer to bugzilla 17630.
277 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
278 the re-exporting nodes to be mounted with "32bitapi" mount option
279 * Note for Ubuntu 14.04.5:
280 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
281 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
283 --------------------------------------------------------------------------------
285 2018-04-03 Intel Corporation
287 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
288 for currently supported client and server kernel versions.
289 * Server known to build on patched kernels:
290 2.6.32-431.29.2.el6 (RHEL6.5)
291 2.6.32-504.30.3.el6 (RHEL6.6)
292 2.6.32-573.26.1.el6 (RHEL6.7)
293 2.6.32-642.15.1.el6 (RHEL6.8)
294 2.6.32-696.18.7.el6 (RHEL6.9)
295 3.10.0-693.21.1.el7 (RHEL7.4)
296 3.0.101-0.47.71 (SLES11 SP3)
297 3.0.101-107 (SLES11 SP4)
298 3.12.74-60.64.40 (SLES12 SP1)
299 4.4.114-92.64 (SLES12 SP2)
300 4.4.114-94.11 (SLES12 SP3)
301 3.13.0-101 (Ubuntu 14.04, ZFS only)
302 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
303 4.4.0-116 (Ubuntu 16.04)
304 vanilla linux 4.6.7 (ZFS only)
305 * Client known to build on unpatched kernels:
306 2.6.32-431.29.2.el6 (RHEL6.5)
307 2.6.32-504.30.3.el6 (RHEL6.6)
308 2.6.32-573.26.1.el6 (RHEL6.7)
309 2.6.32-642.15.1.el6 (RHEL6.8)
310 2.6.32-696.18.7.el6 (RHEL6.9)
311 3.10.0-693.21.1.el7 (RHEL7.4)
312 3.0.101-0.47.71 (SLES11 SP3)
313 3.0.101-107 (SLES11 SP4)
314 3.12.74-60.64.40 (SLES12 SP1)
315 4.4.114-92.64 (SLES12 SP2)
316 4.4.114-94.11 (SLES12 SP3)
317 3.13.0-101 (Ubuntu 14.04)
318 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
319 4.4.0-116 (Ubuntu 16.04)
321 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
322 * Recommended ZFS / SPL version: 0.7.6
323 * Tested with ZFS / SPL version: 0.7.6
324 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
325 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
326 stack overflow. For more information, please refer to bugzilla 17630.
327 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
328 the re-exporting nodes to be mounted with "32bitapi" mount option
329 * Note for Ubuntu 14.04.5:
330 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
331 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
333 --------------------------------------------------------------------------------
335 2017-07-05 Intel Corporation
337 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
338 for currently supported client and server kernel versions.
339 * Server known to build on patched kernels:
340 2.6.32-431.29.2.el6 (RHEL6.5)
341 2.6.32-504.30.3.el6 (RHEL6.6)
342 2.6.32-573.26.1.el6 (RHEL6.7)
343 2.6.32-642.15.1.el6 (RHEL6.8)
344 2.6.32-696.3.1.el6 (RHEL6.9)
345 3.10.0-514.21.1.el7 (RHEL7.3)
346 3.0.101-0.47.71 (SLES11 SP3)
347 3.0.101-100 (SLES11 SP4)
348 3.12.74-60.64.40 (SLES12 SP1)
349 4.4.59-92.17 (SLES12 SP2)
350 vanilla linux 4.6.7 (ZFS only)
351 * Client known to build on unpatched kernels:
352 2.6.32-431.29.2.el6 (RHEL6.5)
353 2.6.32-504.30.3.el6 (RHEL6.6)
354 2.6.32-573.26.1.el6 (RHEL6.7)
355 2.6.32-642.15.1.el6 (RHEL6.8)
356 2.6.32-696.3.1.el6 (RHEL6.9)
357 3.10.0-514.21.1.el7 (RHEL7.3)
358 3.0.101-0.47.71 (SLES11 SP3)
359 3.0.101-100 (SLES11 SP4)
360 3.12.74-60.64.40 (SLES12 SP1)
361 4.4.59-92.17 (SLES12 SP2)
363 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
364 * Recommended ZFS / SPL version: 0.6.5.9
365 * Tested with ZFS / SPL version: 0.6.5.9
366 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
367 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
368 stack overflow. For more information, please refer to bugzilla 17630.
369 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
370 the re-exporting nodes to be mounted with "32bitapi" mount option
372 --------------------------------------------------------------------------------
374 12-07-2016 Intel Corporation
376 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
377 for currently supported client and server kernel versions.
378 * Server known to build on patched kernels:
379 2.6.32-431.29.2.el6 (RHEL6.5)
380 2.6.32-504.30.3.el6 (RHEL6.6)
381 2.6.32-573.26.1.el6 (RHEL6.7)
382 2.6.32-642.11.1.el6 (RHEL6.8)
383 3.10.0-514.el7 (RHEL7.3)
384 3.0.101-0.47.71 (SLES11 SP3)
385 3.0.101-80 (SLES11 SP4)
386 3.12.62-60.64.8 (SLES12 SP1)
387 vanilla linux 4.5.7 (ZFS only)
388 * Client known to build on unpatched kernels:
389 2.6.32-431.29.2.el6 (RHEL6.5)
390 2.6.32-504.30.3.el6 (RHEL6.6)
391 2.6.32-573.26.1.el6 (RHEL6.7)
392 2.6.32-642.11.1.el6 (RHEL6.8)
393 3.10.0-514.el7 (RHEL7.3)
394 3.0.101-0.47.71 (SLES11 SP3)
395 3.0.101-80 (SLES11 SP4)
396 3.12.62-60.64.8 (SLES12 SP1)
398 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
399 * Recommended ZFS / SPL version: 0.6.5.8
400 * Tested with ZFS / SPL version: 0.6.5.8
401 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
402 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
403 stack overflow. For more information, please refer to bugzilla 17630.
404 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
405 the re-exporting nodes to be mounted with "32bitapi" mount option
407 --------------------------------------------------------------------------------
409 02-29-2016 Intel Corporation
411 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
412 for currently supported client and server kernel versions.
413 * Server known to build on patched kernels:
414 2.6.32-431.29.2.el6 (RHEL6.5)
415 2.6.32-504.30.3.el6 (RHEL6.6)
416 2.6.32-573.12.1.el6 (RHEL6.7)
417 3.10.0-327.3.1.el7 (RHEL7.2)
418 3.0.101-0.47.71 (SLES11 SP3)
419 3.0.101-68 (SLES11 SP4)
420 vanilla linux 4.2.1 (ZFS only)
421 * Client known to build on unpatched kernels:
422 2.6.32-431.29.2.el6 (RHEL6.5)
423 2.6.32-504.30.3.el6 (RHEL6.6)
424 2.6.32-573.12.1.el6 (RHEL6.7)
425 3.10.0-327.3.1.el7 (RHEL7.2)
426 3.0.101-0.47.71 (SLES11 SP3)
427 3.0.101-68 (SLES11 SP4)
430 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
431 * Recommended ZFS / SPL version: 0.6.4.2
432 * Tested with ZFS / SPL version: 0.6.4.2
433 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
434 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
435 stack overflow. For more information, please refer to bugzilla 17630.
436 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
437 the re-exporting nodes to be mounted with "32bitapi" mount option
439 --------------------------------------------------------------------------------
441 03-10-2015 Intel Corporation
443 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
444 for currently supported client and server kernel versions.
445 * Server known to build on patched kernels:
446 2.6.32-431.29.2.el6 (RHEL6.5)
447 2.6.32-504.8.1.el6 (RHEL6.6)
448 3.0.101-0.46 (SLES11 SP3)
449 * Client known to build on unpatched kernels:
450 2.6.32-431.29.2.el6 (RHEL6.5)
451 2.6.32-504.8.1.el6 (RHEL6.6)
452 3.10.0-123.20.1.el7 (RHEL7)
453 3.0.101-0.46 (SLES11 SP3)
454 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
455 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
456 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
457 stack overflow. For more information, please refer to bugzilla 17630.
458 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
459 the re-exporting nodes to be mounted with "32bitapi" mount option
461 Severity : enhancement
463 Description: control OST-index in IDIF via ROCOMPAT flag.
464 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
465 last_rcvd file. For new formatted OST device, it will be auto set;
466 for the case of upgrading from old OST device, you can enable it
467 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
468 enabled, for new created OST-object, its IDIF-in-LMA will contain
469 the OST-index; for the existing OST-object, the OSD will convert
470 old format IDIF as new format IDIF with OST-index stored in the
471 LMA EA when accessing such OST-object or via OI scrub. Once such
472 flag is enabled, it cannot be reverted back, so the system cannot
473 be downgraded to the orignal incompatible version.
475 --------------------------------------------------------------------------------
477 07-30-2014 Intel Corporation
479 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
480 for currently supported client and server kernel versions.
481 * Server known to build on patched kernels:
482 2.6.32-431.20.3.el6 (RHEL6.5)
483 3.0.101-0.31 (SLES11 SP3)
484 * Client known to build on unpatched kernels:
485 2.6.32-431.20.3.el6 (RHEL6.5)
486 3.0.101-0.31 (SLES11 SP3)
489 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
490 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
491 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
492 stack overflow. For more information, please refer to bugzilla 17630.
493 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
494 the re-exporting nodes to be mounted with "32bitapi" mount option
496 --------------------------------------------------------------------------------
498 10-31-2013 Intel Corporation
500 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
501 * Server builds for kernels:
502 2.6.32-358.18.1.el6 (RHEL6.4)
503 * Client builds for unpatched kernels:
504 2.6.32-358.18.1.el6 (RHEL6.4)
505 3.0.92-0.8 (SLES11 SP3)
507 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
508 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
509 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
510 stack overflow. For more information, please refer to bugzilla 17630.
511 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
512 the re-exporting nodes to be mounted with "32bitapi" mount option
514 --------------------------------------------------------------------------------
516 05-30-2013 Intel Corporation
518 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
519 * Server builds for kernels:
520 2.6.32-358.6.2.el6 (RHEL6.4)
521 * Client builds for unpatched kernels:
522 2.6.32-358.6.2.el6 (RHEL6.4)
523 3.0.74-0.6.6 (SLES11 SP2)
525 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
526 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
527 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
528 stack overflow. For more information, please refer to bugzilla 17630.
529 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
530 the re-exporting nodes to be mounted with "32bitapi" mount option
532 --------------------------------------------------------------------------------
534 9-30-2012 Whamcloud, Inc.
536 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
537 * Server support for kernels:
538 2.6.32-279.5.1.el6 (RHEL6)
539 * Client support for unpatched kernels:
540 2.6.32-279.5.1.el6 (RHEL6)
542 2.6.18-238.19.1.el5 (RHEL5)
543 2.6.16.60-0.69.1 (SLES10)
544 * Recommended e2fsprogs version: 1.41.90.wc4
545 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
546 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
547 stack overflow. For more information, please refer to bugzilla 17630.
548 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
549 the re-exporting nodes to be mounted with "32bitapi" mount option
551 Severity : enhancement
553 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
555 Severity : enhancement
557 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
559 Severity : enhancement
561 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
563 Severity : enhancement
565 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
567 Severity : enhancement
569 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
571 --------------------------------------------------------------------------------
573 04-30-2012 Whamcloud, Inc.
575 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
576 * Server support for kernels:
577 2.6.32-220.4.2.el6 (RHEL6)
578 * Client support for unpatched kernels:
579 2.6.32-220.4.2.el6 (RHEL6)
581 2.6.18-238.19.1.el5 (RHEL5)
582 2.6.16.60-0.69.1 (SLES10)
583 * Recommended e2fsprogs version: 1.41.90.wc4
584 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
585 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
586 stack overflow. For more information, please refer to bugzilla 17630.
587 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
588 the re-exporting nodes to be mounted with "32bitapi" mount option
590 Severity : enhancement
592 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
594 Severity : enhancement
596 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
598 Severity : enhancement
600 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
602 Severity : enhancement
604 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
606 --------------------------------------------------------------------------------
608 09-30-2011 Whamcloud, Inc.
610 * Server support for kernels:
611 2.6.18-238.19.1.el5 (RHEL5)
612 2.6.32-131.6.1.el6 (RHEL6)
613 * Client support for unpatched kernels:
614 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
615 2.6.18-238.19.1.el5 (RHEL5)
616 2.6.32-131.6.1.el6 (RHEL6)
617 2.6.29.4-167.fc11 (FC11)
618 2.6.16.60-0.69.1 (SLES10)
620 * Recommended e2fsprogs version: 1.41.90.wc3
621 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
622 Lustre filesystem with 4K stack may cause a stack overflow. For more
623 information, please refer to bugzilla 17630.
624 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
625 mounted with "32bitapi" mount option
627 Severity : enhancement
628 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
630 Severity : enhancement
631 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
633 Severity : enhancement
634 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
637 Frequency : only with SLES10
639 Description: Use OFED "KMP" provided by Novell
640 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
641 Lustre is now built against this package. That means you need to
642 install the ofed-kmp package from Novell for the patchless client.
643 Note that the ofed-kmp that Novell ships may not exactly match the
644 kernel version but should still be compatible.
646 Severity : enhancement
648 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
649 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
651 --------------------------------------------------------------------------------
653 2010-07-15 Oracle, Inc.
655 * Server support for kernels:
656 2.6.18-164.11.1.el5 (RHEL 5)
657 2.6.18-164.11.1.0.1.el5 (OEL 5)
658 * Client support for unpatched kernels:
659 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
660 2.6.18-164.11.1.el5 (RHEL 5),
661 2.6.18-164.11.1.0.1.el5 (OEL 5)
662 2.6.16.60-0.42.8 (SLES 10),
664 2.6.29.4-167.fc11 (FC11)
665 * Recommended e2fsprogs version: 1.41.10-sun2
667 Severity : enhancement
669 Description: Distribute lfs_migrate script with lustre
670 Details : The lfs_migrate mentioned in the manual is now included in the
671 lustre tree in lustre/scripts and will distributed in lustre RPMs.
673 Severity : enhancement
675 Description: Failover nid list added to import proc file.
677 Severity : enhancement
679 Description: "som_preview" mount option required to use SOM
680 Details : The clients will need to use som_preview mount option
684 Description: Lustre does not do 1MB IOs to HW RAID
685 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
687 Severity : enhancement
689 Description: lfs quota output is not very convenient for awk/sed-parsing
690 Details : Some positions in lfs quota output table could be empty or
691 non-empty which made it hard to parse it with scripts, now
692 a dash is put instead of space where there is not supposed
697 Description: Update to OFED 1.5.1
698 Details : OFED 1.5.1 will be used in environments that don't use
699 vendor supplied OFED.
703 Description: Remove lnet. prefix from lctl params display
704 Details : Also renamed lnet.memused to lnet_memused to
709 Description: Use conf_param -d to remove permanant settings.
713 Description: Add list_param -R
714 Details : lctl list_param now supports recursive mode.
717 Bugzilla : 21829, 22312
718 Description: Fix llobdstat.
719 Details : 1. Now updating the write_bytes counter
720 2. Port script fixes from 1.8.
724 Description: decrease the usage of memory on clients.
725 Details : 1. On clients, recycle dentries and inodes unused.
726 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
731 Description: llobdstat fix and enhancement
732 Details : add a counter to set a limit to how many samples will be returned
733 fix a wildcard in the path to limit to obdfilter stats only
737 Description: lfs check only allowed for root
738 Details : Code cleanups and added a regresion test of non-root use of
743 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
744 Details : Fix hr->hr_index overload.
748 Description: Fix FIEMAP functionality.
752 Description: FC11 patchless client support added
756 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
757 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
761 Description: New acc-sm script and various test-framework enhancments.
765 Description: Previously landed patch was reverted.
769 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
773 Description: Add ldiskfs maxdirsize mount option.
777 Description: Fix mkfs.lustre support for 16TB LUNs.
781 Description: Print mmp_check_interval
782 Details : print mmp_check_interval and make it possible to abort
783 mount operation in case it takes too long.
787 Description: Add cascading_rw to lustre mpi tests.
791 Description: Store SOM data in EA.
792 Details : Also several SOM related cleanups.
796 Description: Add refcount to md_open_data
800 Description: Use refcount on lov_request_set to prevent memory corruption.
804 Description: Statahead: use parent dir's i_mutex for synchronization.
808 Description: loadgen improvements.
812 Description: Enable peer health detection in ksocklnd and ko2iblnd.
816 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
820 Description: Fix race on obd_devcie:md_stats
824 Description: Fix SLES 11 client builds.
825 Details : Resolved uninitalized pointer gcc warnings.
829 Description: Allow non-root access for "lfs check".
830 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
834 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
836 Severity : enhancement
838 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
839 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
843 Description: Sanity test_27p fails.
844 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
845 Sanity tests 27 were cleaned up and improved.
849 Description: Unify req format on client/servers
850 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
851 buffers and swabbers approach..
856 Description: drop join file stuff.
858 Severity : enhancement
860 Description: Add 2.6.27 and SLES11 (patchless client) support
864 Description: Fix in ptlrpc_expire_one_request() to print the signed time
867 Severity : enhancement
869 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
872 Severity : enhancement
874 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
876 Severity : enhancement
878 Description: correctly handle big reply message.
879 Details : send LNet event if reply is bigger than buffer and adjust this buffer
884 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
886 Severity : enhancement
888 Description: Add support for OEL5.
892 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
894 Severity : enhancement
896 Description: "weak-modules" support
897 Details : Implement "weak-modules" support which enables kernel modules
898 to be used with any kernel that implements the same kABI. In
899 order to achieve this modules are now installed in
900 /lib/modules/$(uname -r)/updates/kernel on all distributions.
904 Description: client could not reconnect to OST because of an active request.
905 Details : The client now retries to reconnect to the same server, if a
906 connect request failed with EBUSY or -EAGAIN. Test case 35b
907 conf-sanity.sh has been created.
911 Description: Don't allow make backward step on assignin osc next id.
912 Details : race between allocation next id and ll_sync thread can be cause
913 of set wrong osc next id and can be kill valid ost objects.
917 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
921 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
925 Description: Error handling in osc_statfs_interpret() has been improved.
926 Details : Check in osc_statfs_interpret() for EBADR.
930 Description: Do not send statfs() requests to OSTs disabled by administrator.
931 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
933 Severity : enhancement
935 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
937 Severity : enhancement
939 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
941 Severity : enhancement
943 Description: Enhancement for lfs(1) command to use numeric uid/gid.
947 Description: Improved cleanup procedure for conf-sanity test_46a.
951 Description: Do not start bulk transfer, if client tried to reconnect.
952 Details : Do not start bulk transfer, if client tried to reconnect,
953 but failed due to active rpcs. Check for exp_abort_active_req
954 has been implemented.
958 Description: Prevent inconsistences between linux and lustre mount structures.
959 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
960 Make the sleep interruptible.
962 Severity : enhancement
964 Description: provide server to client comms path
965 Details : server to client communications path via new LDLM_SET_INFO rpc
969 Description: truncate starts GFP_FS allocation under transaction causing deadlock
970 Details : ldiskfs_truncate calls grab_cache_page which may start page
971 allocation under an open transaction. This may lead to
972 calling prune_icache with consequent lustre reentrance.
976 Description: send correctly lsm on open replay
977 Details : MDS is trust to LSM size on replay open, but client can set wrong size
982 Description: enable client interface failover
983 Details : When a child reconnects from another NID, properly update export
984 nid hash position and ldlm reverse import.
986 Severity : enhancement
988 Description: Communicate OST degraded/readonly state via statfs to MDS
989 Details : Flags in the statfs returned from OSTs indicate whether the
990 OST is in a degraded RAID state, or if the filesystem has
991 turned read-only after a filesystem error is detected.
996 Description: don't panic if EPROTO was hit when reading symlink
997 Details : correctly handling request reference in error cases.
999 Severity : enhancement
1001 Description: Add LustreNetLink, a kernel-userspace communcation path.
1003 Severity : enhancement
1005 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1010 Description: access to llog context before init.
1011 Details : move handling CATALOGS file at osc layer and forbid access to llog
1012 context before init.
1015 Frequency : rare, only when client is writing to OST on local node
1017 Description: Avoid deadlock for local client writes
1018 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1019 memory freeing context. This allows OST threads to set the
1020 PF_MEMALLOC flag on task structures in order to allocate memory
1021 from reserved pools and complete IO.
1022 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1023 so that the OST threads generate memory pressure and allow
1024 inactive pages to be reclaimed.
1026 Severity : enhancement
1028 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1033 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1034 Details : move ldlm namespace creation in setup phase to avoid grab
1035 _lprocfs_lock with cli_sem held.
1039 Description: Temporarily disable grant shrink.
1040 Details : Disable the feature for debugging.
1043 Frequency : only when formatting test filesystems
1045 Description: Unable to run several mkfs.lustre on loop devices at the same time
1046 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1047 formatting. The solution is to proper handle the error.
1049 Severity : enhancement
1051 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1053 Severity : enhancement
1055 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1060 Description: resolve race between obd_disconnect and class_disconnect_exports
1061 Details : if obd_disconnect will be called to already disconnected export he
1062 forget release one reference and osc module can't unloaded.
1064 Severity : enhancement
1066 Description: Allow tuning service thread via /proc
1067 Details : For each service a new
1068 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1069 created that can be used to set min/max thread counts, and get the
1070 current number of running threads.
1075 Descriptoin: don't return error if have partially created objects for file.
1076 Details : lov_update_create_set uses set->set_success as index for created
1077 objects, so if some requests will be failed, they will have hole
1078 at end of array and qos_shrink_lsm for allocate correct lsm.
1080 Severity : enhancement
1082 Description: Update OFED support to 1.4.1
1087 Description: Reduce small size read RPC
1088 Details : Set read-ahead limit for every file and only do read-ahead when
1089 available read-ahead pages are bigger than 1M to avoid small size
1092 Severity : enhancement
1094 Description: move AT tunable parameters for more consistent usage
1095 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1097 Severity : enhancement
1099 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1100 Details : allow skip disconnected ost for send statfs request and hide error
1106 Description: fix racy locking of mballoc block bitmaps causing BUG
1107 Details : The locking of the mballoc buddy bitmap and the in-memory
1108 block bitmap was using two different spin locks in some
1109 cases. This made it possible to incorrectly access the
1110 mballoc bitmap while another process was modifying it,
1111 causing a sanity assertion to fail. While no on-disk corruption
1112 was reported, there was some risk of this happening.
1115 Frequency : rare, on llog test 6
1117 Descriptoin: don't allow connect to already connected import
1118 Details : allowing connect to already connected import is hide connecting problem.
1123 Descriptoin: allow kill process which wait statahead result
1124 Details : for some reasons 'ls' can stick in waiting result from statahead,
1125 in this case need way for kill this process.
1127 Severity : enhancement
1129 Description: Add state history info file, enhance import info file
1130 Details : Track import connection state changes in a new osc/mdc proc file;
1131 add overview-type data to the osc/mdc import proc file.
1133 Severity : enhancement
1135 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1137 Severity : enhancement
1139 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1144 Descriptoin: OST grant too much space to client even there are not enough space.
1145 Details : Client will shrink its grant cache to OST if there are no write
1146 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1147 this grant cache if there are already not enough avaible space
1148 (left_space < total_clients * 32M).
1151 Frequency : start MDS on uncleanly shutdowned MDS device
1153 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1154 Details : stay in waiting mds<>ost recovery finished produce random bugs
1155 due race between two ll_sync thread for one lov target. send
1156 ACTIVATE event only if connect realy finished and import have
1160 Frequency : rare, connect and disconnect target at same time
1162 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1163 Details : don't call obd_disconnect under lov_lock. this long time
1164 operation and can block ptlrpcd which answer to connect request.
1169 Descriptoin: don't lose wakeup for imp_recovery_waitq
1170 Details : recover_import_no_retry or invalidate_import and import_close can
1171 both sleep on imp_recovery_waitq, but we was send only one wakeup
1175 Frequency : always with long access acl
1177 Descriptoin: mds can't pack reply with long acl.
1178 Details : mds don't control size of acl but they limited by reint/getattr
1181 Severity : enhancement
1183 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1185 Severity : enhancement
1187 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1190 Frequency : start MDS on uncleanly shutdowned MDS device
1192 Descriptoin: aborting recovery hang on MDS
1193 Details : don't throttle destroy RPCs for the MDT.
1196 Frequency : on remount
1198 Description: external journal device not working after the remount
1199 Details : clear dev_rdonly flag for external journal devices in
1205 Description: shutdown vs evict race
1206 Details : client_disconnect_export vs connect request race.
1207 if client will evicted at this time - we start invalidate
1208 thread without referece to import and import can be freed
1212 Frequency : rare, need acl's on inode.
1214 Description: client can't handle ost additional correctly
1215 Details : if ost was added after client connected to mds client can have
1216 hit lnet_try_match_md ... to big messages to wide striped files.
1217 in this case need teach client to handle config events about add
1218 lov target and update client max ea size at that event.
1220 Severity : enhancement
1222 Description: Changelogs
1223 Details : Changelogs are a lightweight mechanism to track filesystem
1224 metadata and namespace changes. The changelog is recorded
1225 permanently on the MDTs, and is periodically "consumed" / purged
1226 when records are no longer needed.
1228 Severity : enhancement
1230 Description: compact fld format with extents
1231 Details : Store range of seq rather than every seq in FLD. Seq
1232 controller update FLD rather than clients. In Case of CMD, mdt0
1233 has FLD, all other metadata server act as non persistent proxy
1234 for FLD queries and cache fld entries in fld cache.
1239 Description: don't skip ost target if they assigned to file
1240 Details : Drop slow OSCs if we can, but not for requested start idx.
1241 This means "if OSC is slow and it is not the requested
1242 start OST, then it can be skipped, otherwise skip it only
1243 if it is inactive/recovering/out-of-space.
1247 Description: more cleanup in mds_lov
1248 Details : not send LOV EA under replay, we can't know about they size at this
1249 time. Don't allow client connect to mds before any ost connected,
1250 for avoid problems with LOV EA size and returning EIO to client.
1252 Severity : enhancement
1254 Description: Interoperability at server side (Disk interoperability)
1256 Severity : enhancement
1258 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1260 Severity : enhancement
1262 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1264 Severity : enhancement
1266 Description: New client IO stack (CLIO).
1268 Severity : enhancement
1270 Description: Commit on sharing. Eliminate inter-client dependencies between
1271 uncommitted transactions by doing transaction commits.
1272 Thereby clients may recovery independently.
1275 Frequency : Create a symlink file with a very long name
1277 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1278 Details : If there is no extra space in the request for early cancels,
1279 ldlm_req_handles_avail() returns 0 instead of a negative value.
1281 Severity : enhancement
1283 Description: Add /proc entry for import status
1284 Details : The mdc, osc, and mgc import directories now have
1285 an import directory that contains useful import data for debugging
1286 connection problems.
1288 Severity : enhancement
1290 Description: Re-disable certain /proc logging
1291 Details : Enable and disable client's offset_stats, extents_stats and
1292 extents_stats_per_process stats logging on the fly.
1295 Frequency : Only on FC kernels 2.6.22+
1297 Description: oops in statahead
1298 Details : Do not drop reference count for the dentry from VFS when lookup,
1299 VFS will do that by itself.
1301 Severity : enhancement
1303 Description: Generic /proc file permissions
1304 Details : Set /Proc file permissions in a more generic way to enable non-
1305 root users operate on some /proc files.
1309 Description: Hitting mdc_commit_close() ASSERTION
1310 Details : Properly handle request reference release in
1311 ll_release_openhandle().
1315 Description: quota recovery deadlock during mds failover
1316 Details : This patch includes att18982, att18236, att18237 in bz14840.
1318 1. fix osts hang when mds does failover with quotaon
1319 2. prevent watchdog storm when osts threads wait for the
1324 Frequency : only patchless client
1325 Description: add workaround for race between add/remove dentry from hash
1327 Severity : enhancement
1329 Description: Allow OST glimpses to return PW locks
1333 Description: LBUG when llog conf file is full
1334 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1338 Description: Prevent import from entering FULL state when server in recovery
1342 Description: service mount cannot take device name with ":"
1343 Details : Only when device name contains ":/" will mount treat it as
1349 Description: replace ptlrpcd with the statahead thread to interpret the async
1350 statahead RPC callback
1354 Frequency : on recovery
1355 Description: I/O failures after umount during fail back
1356 Details : if client reconnected to restarted server we need join to recovery
1357 instead of find server handler is changed and process self
1358 eviction with cancel all locks.
1360 Severity : enhancement
1362 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1364 Severity : enhancement
1366 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1368 Severity : enhancement
1370 Description: Upcall on Lustre log has been dumped
1371 Details : Allow for a user mode script to be called once a Lustre log has
1372 been dumped. It passes the filename of the dumped log to the
1373 script, the location of the script can be specified via
1374 /proc/sys/lnet/debug_log_upcall.
1379 Description: avoid idr_remove called for id which is not allocated.
1380 Details : Move assigment s_dev for clustered nfs to end of initialization,
1381 to avoid problem with error handling.
1386 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1387 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1388 into CONN_USED_HASH and this prodice warning when put connection
1389 again in unused hash.
1392 Severity : enhancement
1394 Description: File striping can now be set to use an arbitrary pool of OSTs.
1396 Severity : enhancement
1398 Description: Export bytes_read/bytes_write count on OSC/OST.
1402 Description: Early reply size mismatch, MGC loses connection
1403 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1404 the connect flags are properly negotiated.
1409 Description: quotas are not honored with O_DIRECT
1410 Details : all writes with the flag O_DIRECT will use grants which leads to
1411 this problem. Now using OBD_BRW_SYNC to guard this.
1415 Description: add quota statistics
1416 Details : 1. sort out quota proc entries and proc code.
1417 2. add quota statistics
1419 Severity : enhancement
1421 Description: enable quota support for HEAD.
1425 Description: Properly propagate oinfo flags from lov to osc for statfs
1426 Details : restore missing copy oi_flags to lov requests.
1428 Severity : enhancement
1430 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1431 : l_getgroups(8), lst(8), routerstat(8)
1432 Details : included man pages for llobdstat(8), llstat(8),
1433 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1435 Severity : enhancement
1437 Description: configure's --enable-quota should check the
1438 : kernel .config for CONFIG_QUOTA
1439 Details : configure is terminated if --enable-quota is passed but
1440 : no quota support is in kernel
1444 Description: Remove portals compatibility
1445 Details : Remove portals compatibility, not interoperable with releases
1450 Description: Resolve device initialization race
1451 Details : Prevent proc handler from accessing devices added to the
1452 obd_devs array but yet be intialized.
1454 Severity : enhancement
1456 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1458 Severity : enhancement
1460 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1465 Description: Using wrong pointer in osc_brw_prep_request
1466 Details : Access to array[-1] can produce panic if kernel compiled with
1467 CONFIG_PAGE_ALLOC enabled
1471 Description: Client runs out of low memory
1472 Details : Consider only lowmem when counting initial number of llap pages
1476 Description: *optional* service tags registration
1477 Details : if the "service tags" package is installed on a Lustre node
1478 When the filesystem is mounted, a local-node service tag will
1479 be created. See http://inventory.sun.com/ for more information
1480 about the Service Tags asset management system.
1484 Description: Kernel BUG tries to release flock
1485 Details : Lustre does not destroy flock lock before last reference goes
1486 away. So always drop flock locks when client is evicted and
1487 perform unlock regardless of successfulness of speaking to MDS.
1491 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1495 Description: sanity test 65a fails if stripecount of -1 is set
1496 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1501 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1502 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1505 Severity : enhancement
1507 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1511 Description: do not process already freed flock
1512 Details : flock can possibly be freed by another thread before it reaches
1513 to ldlm_flock_completion_ast.
1517 Description: LBUG during stress test
1518 Details : Need properly lock accesses the flock deadlock detection list.
1522 Description: oops in page fault handler
1523 Details : kernel page fault handler can return two special 'pages' in error
1524 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1528 Description: timeout with invalidate import.
1529 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1530 should be handled by ptlrpcd. This produce long age waiting and
1531 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1533 Severity : enhancement
1535 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1539 Description: enable MGS and MDT services start separately
1540 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1541 a co-located MGS without starting the MGS, which is a complement
1542 to 'nosvc' mount option.
1546 Frequency : after recovery
1547 Description: precreate to many object's after del orphan.
1548 Details : del orphan st in oscc last_id == next_id and this triger growing
1549 count of precreated objects. Set flag LOW to skip increase count
1550 of precreated objects.
1554 Frequency : rare, on clear nid stats
1555 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1556 Details : when clean nid stats sometimes try destroy live entry,
1557 and this produce panic in free.
1561 Description: Stack overflow during MDS log replay
1562 ease stack pressure by using a thread dealing llog_process.
1566 Description: wait until IO finished before start new when do lock cancel.
1567 Details : VM protocol want old IO finished before start new, in this case
1568 need wait until PG_writeback is cleared until check dirty flag and
1569 call writepages in lock cancel callback.
1571 Severity : enhancement
1573 Description: using special macro for print time and cleanup in includes.
1577 Description: mds_mfd_close() ASSERTION(rc == 0)
1578 Details : In mds_mfd_close(), we need protect inode's writecount change
1579 within its orphan write semaphore to prevent possible races.
1583 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1587 Frequency : rare, on shutdown ost
1588 Description: don't hit live lock with umount ost.
1589 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1590 use shrink_dcache_sb instead.
1594 Description: don't panic with use echo client
1595 Details : echo client pass NULL as client nid pointer and this produce null
1596 pointer dereference.
1600 Description: fix build on ppc32
1601 Details : compile code with -m64 flag produce wrong object file for ppc32.
1605 Description: add message levels for liblustreapi
1609 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1610 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1611 will result in an -EAGAIN and in the exit call path we call
1612 alloc_rr() which will with increasing aggressiveness attempt to
1613 aquire precreated objects on the minimum number of required OSCs.
1617 Description: Use old size assignment to avoid deadlock
1618 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1619 the scheduling while holding a spinlock. We do not need locking
1620 for size in ll_update_inode() because size is only updated from
1621 the MDS for directories or files without objects, so there is no
1622 other place to do the update, and concurrent access to such inodes
1623 are protected by the inode lock.
1627 Description: resolve "_IOWR redefined" build error on SLES10
1631 Description: dump the memory debugging after all modules are unloaded to
1632 suppress false negative in conf_sanity test 39
1634 Severity : enhancement
1636 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1641 Description: fix SLES kernel versioning
1642 Details : the kernel version for our SLES 10 kernel did not include a "-"
1643 before the "smp" at the end. while this was not a problem in
1644 general, it did mean that software trying to use the kernel
1645 version to try to detect a vendor specific kernel would fail.
1646 this was most evident by the OFED build scripts.
1650 Description: Don't update lov_desc members until making sure they are valid
1651 Details : When updating lov_desc members via proc fs, need fix their
1652 validities before doing the real update.
1656 Description: don't put request into delay list while invalidate in flight.
1657 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1658 import in flight. this produce timeout for invalidate and sometimes
1659 can cause stale data.
1663 Frequency : on ppc only
1664 Description: not convert ost objects for directory because it's not exist.
1665 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1669 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1671 Severity : enhancement
1673 Description: Update to support 2.6.22.14 vanilla kernel.
1677 Frequency : rare, on recovery
1678 Description: read procfs can produce deadlock in some situation
1679 Details : Holding lprocfs lock which send rpc can produce block for destroy
1680 obd objects and this also block reconnect with -EALREADY. This
1681 isn't fix all lprocfs bugs - but make it rare.
1683 Severity : enhancement
1685 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1688 Frequency : frequent on X2 node
1690 Description: mdc_set_open_replay_data LBUG
1691 Details : Set replay data for requests that are eligible for replay.
1695 Description: lustre_mgs: operation 101 on unconnected MGS
1696 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1697 MGC, and late on MGC cannot successfully connect to MGS and a lot
1698 of the error messages complaining that MGS is not connected.
1702 Frequency : on network error
1703 Description: panic with double free request if network error
1704 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1705 but it's true only for synchronus enqueue, for async enqueue
1706 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1709 Severity : enhancement
1711 Description: client-side metadata stat-ahead during readdir(directory readahead)
1712 Details : perform client-side metadata stat-ahead when the client detects
1713 readdir and sequential stat of dir entries therein
1716 Frequency : on start mds
1718 Description: Implement get_info(last_id) in obdfilter.
1721 Frequency : occasional
1723 Description: Correctly check stale fid, not start epoch if ost not support SOM
1724 Details : open with flag O_CREATE need set old fid in op_fid3 because
1725 op_fid2 was overwritten with new generated fid, but mds can answer
1726 with one of these two fids and both is not stale. Setattr
1727 incorrectly started an epoch and assume will be called
1728 done_writing, but without SOM done_writing ever being called.
1731 Frequency : rare, depends on device drivers and load
1733 Description: MDS or OSS nodes crash due to stack overflow
1734 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1735 In some cases, in conjunction with device drivers that use a lot
1736 of stack the MDS (or possibly OSS) service threads could overflow
1737 the stack. One change which was identified to consume additional
1738 stack has been reworked to avoid the extra stack usage.
1741 Frequency : occasional
1743 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1744 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1745 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1746 oscc_has_objects and see OSC has no objests and call
1747 oscc_internal_create to resend create request.
1749 Severity : enhancement
1751 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1753 Severity : enhancement
1755 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1758 Frequency : very rare
1760 Description: Fix replay if there is an un-replied request and open
1761 Details : In some cases, older replay request will revert the
1762 mcd->mcd_last_xid on MDS which is used to record the client's
1763 latest sent request.
1765 Severity : enhancement
1767 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1769 Severity : enhancement
1771 Description: Add rhel5 support to HEAD.
1773 Serverity : enhancement
1775 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1780 Description: Don't allow skipping OSTs if index has been specified.
1781 Details : Don't allow skipping OSTs if index has been specified, make
1782 locking in internal create lots better.
1786 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1787 Details : access to bitfield in structure is always rounded to long
1788 and this produce problem with not atomic change any bit.
1792 Description: Lustre make rpms failed.
1793 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1794 builds Lustre rpms from tarball.
1797 Frequency : rare on shutdown ost
1799 Description: If llog cancel was not send before clean_exports phase, this can
1800 produce deadlock in llog code.
1801 Details : If llog thread has last reference to obd and call class_import_put
1802 this produce deadlock because llog_cleanup_commit_master wait when
1803 last llog_commit_thread exited, but this never success because was
1804 called from llog_commit_thread.
1808 Description: allow userland application know is lost one of stripes.
1809 Details : fill lvb_blocks with error code on ost and return it to
1810 application if error flag found.
1814 Description: NULL lov_tgts causing MDS oops
1815 Details : more safe checks for NULL lov_tgts for avoid oops.
1817 Severity : enhancement
1819 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1823 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1824 Details : If an OST is down the MDS will hang indefinitely in
1825 obd_statfs_rqset() waiting for the statfs data. While for
1826 MDS QOS usage of statfs, it should not stuck in waiting.
1828 Severity : enhancement
1830 Description: remote_acl support
1831 Details : Support ACL-based permission check for remote user.
1832 Support setfacl/getfacl for remote user with the utils
1833 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1834 the system "{s,g}etfacl" utils.
1836 Severity : enhancement
1838 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1840 Severity : enhancement
1842 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1846 Description: make mgs_setparam() handle fsname containing dash
1847 Details : fsname containing a dash does not work with lctl conf_param
1849 Severity : enhancement
1851 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1853 Severity : enhancement
1855 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1859 Frequency : rare, at shutdown
1860 Description: access already free / zero obd_namespace.
1861 Details : if client_disconnect_export was called without force flag set,
1862 and exist connect request in flight, this can produce access to
1863 NULL pointer (or already free pointer) when connect_interpret
1864 store ocd flags in obd_namespace.
1868 Frequency : only at startup
1869 Description: not alloc memory with spinlock held.
1870 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1871 if any spinlock held.
1873 Severity : enhancement
1875 Description: make lustre randomly fail allocating memory
1876 Details : Make lustre randomly failed allocating memory for testing purpose.
1878 Severity : enhancement
1880 Description: lost problems with lov objid file
1881 Details : Fixes some scability and access to not inited memory problems
1882 in work with lov objdid file.
1887 Description: lfs find does not continue on file error
1888 Details : Continue other files processing when a file/dir is absent.
1892 Description: Inconsistent usage of lustre_pack_reply()
1893 Details : Standardize the usage of lustre_pack_reply() such that it
1894 always generate a CERROR on failure.
1897 Frequency : occasional
1899 Description: MDS hang or stay in waiting lock
1900 Details : If client receive lock with CBPENDING flag ldlm need send lock
1901 cancel as separate rpc, to avoid situation when cancel request
1902 can't processed due all i/o threads stay in wait lock.
1906 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1910 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1914 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1918 Description: Update to RHEL4 latest kernel.
1920 Severity : enhancement
1922 Description: Build SLES10 patchless client fails
1923 Details : The configure was broken by run ./configure with
1924 --with-linux-obj=.... argument for patchless client. When the
1925 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1926 can't search header adequately. Use absolute path such as
1927 -I($LINUX)/include instead.
1931 Description: interrupt oig_wait produce painc on resend.
1932 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1933 private set, and this produce situation when rq_ptlrpcd_data not
1934 copyed to new allocated request and triggered LBUG on assert
1935 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1936 ptlrpcd set if request is changed and can be safety changed to use
1941 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1942 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1943 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1944 these macros and replace them with their difinitions instead.
1948 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1949 Details : Update which_patch & target file for SLES10 latest kernel.
1952 Description: Remove client patches from SLES 10 kernel.
1953 Details : This causes SLES 10 clients to behave as patchless clients
1954 even on a Lustre-patched (server) kernel.
1956 Severity : enhancement
1958 Description: self-adjustable client's lru lists
1959 Details : use adaptive algorithm for managing client cached locks lru
1960 lists according to current server load, other client's work
1961 pattern, memory activities, etc. Both, server and client
1962 side namespaces provide number of proc tunables for controlling
1965 Severity : enhancement
1967 Description: light-weight GSS support
1968 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1969 and reduce performance overhead.
1971 Severity : enhancement
1973 Description: Linux keyring support
1974 Details : Support using service of Linux keyring for Lustre GSS internal
1975 context refresh/cache mechanism.
1979 Description: Fix errors in lfs documentation
1980 Details : Fixes man pages
1984 Description: don't use GFP_* in generic Lustre code.
1985 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1989 Description: obdclass is limited by single OBD_ALLOC(idarray)
1990 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1994 Description: warnings with build patchless client with vanila 2.6.19 and up
1995 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1996 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2000 Frequency : Only for SLES
2001 Description: sanity_quota fail test_1
2002 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2003 file, which makes TSTID[2] inunique.
2007 Frequency : Only for SLES9
2008 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2009 Details : Update target/ChangeLog/which_patch .
2013 Frequency : Only for test_10 in sanity-quota.sh
2014 Description: a bug in quota test code
2015 Details : checking a test flag in a wrong place
2019 Frequency : Only for test_13 in sanity-quota.sh
2020 Description: a bug in quota test script
2021 Details : shouldn't check overfull allocated quota
2025 Description: jbd statistics
2026 Details : Port older jbd statistics patch for sles10
2028 Severity : enhancement
2030 Description: Add printing inode info into message about error in writepage.
2035 Description: reply_lock_interpret crash due to race with it and lock cancel.
2036 Details : Do not replay locks that are being cancelled. Do not reference
2037 locks by their address during replay, just by their handle.
2041 Frequency : When flocks are used.
2042 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2043 Details : Flock locks might destroy just granted lock if it could be merged
2044 with another existing flock, this is done in completion handler,
2045 so teach ldlm_cli_enquque_fini that this is a valid case for
2051 Description: Oops in read and write path when failing to allocate lock.
2052 Details : Check if lock allocation failed and return error back.
2056 Description: lstripe command fails for valid OST index
2057 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2058 instead of lov->desc.ld_active_tgt_count.
2062 Description: sanity.sh failed test 103
2063 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2068 Description: df doesn't work properly if diskfs blocksize != 4K
2069 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2073 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2074 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2076 Severity : enhancement
2078 Description: Adaptive timeouts
2079 Details : RPC timeouts adapt to changing server load and network
2080 conditions to reduce resend attempts and improve recovery time.
2084 Description: llapi_file_create() does not allow some changes
2085 Details : add llapi_file_open() that allows specifying the mode and
2086 open flags, and also returns an open file handle.
2090 Description: merge and cleanup kernel patches.
2091 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2095 Description: Add journal checksum support.(Kernel part)
2096 Details : The journal checksum feature adds two new flags i.e
2097 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2098 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2099 indicates that the commit block contains the checksum for
2100 the blocks described by the descriptor blocks. Now commit
2101 record can be sent to disk without waiting for descriptor
2102 blocks to be written to disk. This behavior is controlled
2103 using JBD2_FEATURE_ASYNC_COMMIT flag.
2107 Description: OSS needs mutliple precreate threads
2108 Details : Add ability to start more than one create thread per OSS.
2112 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2113 Details : execute lfs setstripe on client
2117 Description: mds_obd_create error creating tmp object
2118 Details : When the user sets quota on root, llog will be affected and can't
2119 create files and write files.
2122 Frequency : Always on ia64 patchless client, and possibly others.
2124 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2125 Details : This allows the patchless client to be loaded on architectures
2126 without this export.
2130 Description: RedHat Update kernel for RHEL5
2131 Details : Add the kernel config file for RHEL5.
2135 Description: RedHat Update kernel for RHEL5
2136 Details : Modify the kernel config file more closer RHEL5.
2140 Description: Build failure against Centos5 (RHEL5)
2141 Details : Define PAGE_SIZE when it isn't present.
2145 Description: test_7 on sanity-quota.sh can't be used on separate machines
2146 Details : add facet to handle it
2150 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2151 Details : with NFS, the anon dentry's parent was set to itself in
2152 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2153 corresponding dentry other than use rec->ur_name.
2155 Severity : enhancement
2157 Description: lfs setstripe enhancement
2158 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2162 Description: enable data checksumming by default
2163 Details : enable checksum by default, allow --disable-checksum
2164 configure option and "-o nochecksum" mount option. Checksums
2165 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2169 Description: lustre support for RHEL5
2170 Details : Add support for RHEL5.
2174 Description: Client eviction due to failover config
2175 Details : after a connection loss, the lustre client should attempt to
2176 reconnect to the last active server first before trying the
2177 other potential connections.
2181 Description: when mds and osts use different quota unit(32bit and 64bit),
2182 quota will be released repeatly.
2183 Details : void sending multiple quota reqs to mds, which will keep the
2184 status between the reqs.
2188 Description: osts not allocated evenly to files
2189 Details : change the condition to increase offset_idx
2192 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2194 Description: Data corruption for OSTs that are formatted larger than 2TB
2196 Details : When generating the bio request for lustre file writes the
2197 sector number would overflow a temporary variable before being
2198 used for the IO. The data reads correctly from Lustre (which
2199 will overflow in a similar manner) but other file data or
2200 filesystem metadata may be corrupted in some cases.
2204 Description: Tune the kernel for good SCSI performance.
2205 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2206 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2211 Description: rewrite ext2-derived code in obdclass/uuid.c
2212 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2213 from scratch preserving functionality.
2217 Description: "lfs find -obd UUID" prints directories
2218 Details : "lfs find -obd UUID" will return all directory names instead
2219 of just file names. It is incorrect because the directories
2220 do not reside on the OSTs.
2224 Description: use i_size_read and i_size_write in 2.6 port
2225 Details : replace inode->i_size access with i_size_read/write()
2228 Frequency : when removing large files
2230 Description: scheduling issue during removal of large Lustre files
2231 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2232 It causes scheduling issues when removing large files (17TB in the
2236 Frequency : only with liblustre clients on XT3
2238 Description: evictions taking too long
2239 Details : allow llrd to evict clients directly on OSTs
2242 Frequency : only on ppc
2244 Description: /proc/fs/lustre/devices broken on ppc
2245 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2246 We should make sure the type of 'index' is loff_t and then cast
2247 explicitly as needed below. Do not assign an explicitly cast
2251 Frequency : only for rhel5
2253 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2254 Details : Modify the target file & which_kernel.
2256 Severity : enhancement
2258 Description: omit set fsid for export NFS
2259 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2260 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2264 Frequency : Always, for kernels after 2.6.16
2265 Description: Fix warning idr_remove called for id=.. which is not allocated.
2266 Details : Last kernels save old s_dev before kill super and not allow
2267 to restore from callback - restore it before call kill_anon_super.
2271 Description: buffer overruns could theoretically occur
2272 Details : llapi_semantic_traverse() modifies the "path" argument by
2273 appending values to the end of the origin string, and a
2274 overrun may occur. Adding buffer overrun check in liblustreapi.
2278 Description: Fix error on 'ls .' at the top of the Lustre mount.
2279 Details : Don't revalidate dentry if it is a root dentry.
2283 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2284 Details : Modify vm-tunables-rhel4.patch.
2288 Description: Kernel config for 2.6.18-vanilla.
2289 Details : Modify targets/2.6-vanilla.target.in.
2290 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2291 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2292 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2293 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2297 Description: improve handling recoverable errors
2298 Details : if request processig with error which can be recoverable on server
2299 request should be resend, otherwise page released from cache and
2304 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2305 Details : Modify target file & which_patch.
2306 A flaw was found in the IA32 system call emulation provided
2307 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2308 value could be stored in the %RAX register, which could trigger an
2309 out-of-bounds system call table access. An untrusted local user
2310 could exploit this flaw to run code in the kernel
2311 (ie a root privilege escalation). (CVE-2007-4573).
2315 Description: O_DIRECT bypasses client statistics.
2316 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2317 read_ahead_stats were updated. copied the stats section from
2318 osc_send_oap_rpc() into async_internal().
2322 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2326 Description: change order of libsysio includes
2327 Details : '#include sysio.h' should always come before '#include xtio.h'
2331 Description: handle "serious error: objid * already exists" more gracefully
2332 Details : If LAST_ID value on disk is smaller than the objects existing in
2333 the O/0/d* directories, it indicates disk corruption and causes an
2334 LBUG(). If the object is 0-length, then we should use the existing
2335 object. This will help to avoid a full fsck in most cases.
2337 Severity : enhancement
2339 Description: adapt the lustre_config script to support the upgrade case
2340 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2341 server targets to 1.6 in parallel.
2345 Description: To avoid grant space > avaible space when the disk is almost
2346 full. Without this patch you might see the error "grant XXXX >
2347 available" or some LBUG about grant, when the disk is almost
2349 Details : In filter_check_grant, for non_grant cache write, we should
2350 check the left space by if (*left > ungranted + bytes), instead
2351 of (*left > ungranted), because only we are sure the left space
2352 is enough for another "bytes", then the ungrant space should be
2353 increase. In client, we should update cl_avail_grant only there
2354 is OBD_MD_FLGRANT in the reply.
2358 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2359 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2360 described in CVE-2007-4573. This problem could allow untrusted
2361 local users to gain root access.
2364 Frequency : when using O_DIRECT and quotas
2366 Description: Incorrect file ownership on O_DIRECT output files
2367 Details : block usage reported by 'lfs quota' does not take into account
2368 files that have been written with O_DIRECT.
2373 Description: touch file failed when fs is not full
2374 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2375 otherwise we can get ENOSP while fs is not full.
2379 Description: parallel lock callbacks
2380 Details : Instead of sending blocking and completion callbacks as separated
2381 requests, adding them to a set and sending in parallel.
2384 Frequency : only for Cray XT3
2385 Bugzilla : 12829/13455
2386 Description: Changing primary group doesn't change the group lustre assigns to
2388 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2389 overridden with the primary group provided by the group upcall,
2390 whereas the supplied fsgid can be trusted if it is in the list of
2391 supplementary groups returned by the group upcall.
2393 Severity : enhancement
2395 Description: Allow masking D_WARNING, D_ERROR messages from console
2396 Details : Console messages can now be disabled via lnet.printk.
2400 Description: User code with malformed file open parameter crashes client node
2401 Details : Before packing join_file req, all the related reference should be
2402 checked carefully in case some malformed flags cause fake
2403 join_file req on client.
2407 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2408 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2409 mode is equal to the granted mode. However, it should also check
2410 that LDLM_FL_CP_REQD is not set.
2414 Description: Heavy nfs access might result in deadlocks
2415 Details : After ELC code landed, it is now improper to enqueue any mds
2416 locks under och_sem, because enqueue might want to decide to
2417 cancel open locks for same inode we are holding och_sem for.
2421 Description: Client eviction while running blogbench
2422 Details : A lot of unlink operations with concurrent I/O can lead to a
2423 deadlock causing evictions. To address the problem, the number of
2424 oustanding OST_DESTROY requests is now throttled to
2425 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2426 callbacks are processed in priority.
2430 Description: enable ACLs on MDS by default
2431 Details : ACLs must be enabled on MDS by default.
2434 Frequency : PPC/PPC64 only
2436 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2437 Details : fix duplicated definitions between asm-ppc64/types.h and
2438 lustre_types.h on PPC.
2441 Frequency : PPC/PPC64 only
2443 Description: asm-ppc/segment.h does not exist
2444 Details : fix compile issue on PPC.
2448 Description: better handle error messages in extents code
2451 Frequency : RHEL4 only
2453 Description: mkfs is very slow on IA64/RHEL4
2454 Details : A performance regression has been discovered in the MPT Fusion
2455 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2456 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2457 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2459 Severity : enhancement
2461 Description: SNMP support enhancement
2462 Details : Adding total number of sampled request for an MDS node in snmp
2465 Severity : enhancement
2467 Description: Optimize ldlm waiting list processing for PR extent locks
2468 Details : When processing waiting list for read extent lock and meeting read
2469 lock that is same or wider to it that is not contended, skip
2470 processing rest of the list and immediatelly return current
2471 status of conflictness, since we are guaranteed there are no
2472 conflicting locks in the rest of the list.
2476 Description: Time out and refuse to reconnect
2477 Details : When the failover node is the primary node, it is possible
2478 to have two identical connections in imp_conn_list. We must
2479 compare not conn's pointers but NIDs, otherwise we can defeat
2480 connection throttling.
2484 Description: port llog fixes from b1_6 into HEAD
2485 Details : Port llog reference couting and some llog cleanups from b1_6
2486 (bug 10800) into HEAD, for protect from panic and access to already
2487 free llog structures.
2491 Description: Detect stride IO mode in read-ahead
2492 Details : When a client does stride read, read-ahead should detect that and
2493 read-ahead pages according to the detected stride pattern.
2497 Description: data checksumming impacts single node performance
2498 Details : add support for several checksum algorithm. Currently, only CRC32
2499 and Adler-32 are supported. The checksum type can be changed on
2500 the fly via /proc/fs/lustre/osc/*/checksum_type.
2504 Description: use adler32 for page checksums
2505 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2510 Description: build for x2 fails
2511 Details : fix compile issue on Cray systems.
2515 Description: Properly match for duplicate locks
2516 Details : Due to different lock order from skiplists code, we need to
2517 traverse entire list for now
2520 Frequency : only on PPC/SLES10
2522 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2523 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2524 be defined. Add a hack in mkfs_lustre.c to work around this compile
2529 Description: LASSERT on MDS when client holding flock lock dies
2530 Details : ldlm pool logic depends on number of granted locks equal to
2531 number of released locks which is not true for flock locks, so
2532 just exclude such locks from consideration.
2536 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2537 Details : Use fsfilt_sync() for both the whole filesystem sync and
2538 individual file sync to eliminate dangerous inode locking
2539 with I_LOCK that can lead to a deadlock.
2543 Description: performance in 1.6.3
2544 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2545 sure that 1MB requests are not fragmented by the block layer.
2547 Severity : enhancement
2549 Description: organize the server-side client stats on per-nid basis
2550 Details : Change the structure of stats under obdfilter and mds to
2558 The "uuid"s file would list the uuids of _active_ exports.
2559 And the clear entry is to clear all stats and stale nids.
2561 Severity : enhancement
2563 Description: eliminate client locks in face of contention
2564 Details : file contention detection and lockless i/o implementation
2565 for contended files.
2569 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2572 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2574 Description: MDS may lose file striping (and hence file data) in some cases
2575 Details : If there are additional extended attributes stored on the MDS,
2576 in particular ACLs, SELinux, or user attributes (if user_xattr
2577 is specified for the client mount options) then there is a risk
2578 of attribute loss. Additionally, the Lustre file striping
2579 needs to be larger than default (e.g. striped over all OSTs),
2580 and an additional attribute must be stored initially in the
2581 inode and then increase in size enough to be moved to the
2582 external attribute block (e.g. ACL growing in size) for file
2587 Description: skiplist implementation simplification
2588 Details : skiplists are used to group compatible locks on granted list
2589 that was implemented as tracking first and last lock of each
2590 lock group the patch changes that to using doubly linked lists
2594 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2595 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2596 because LASSERT on both the data supplied by a client, and the
2597 data on disk is dangerous and incorrect.
2599 Severity : enhancement
2601 Description: Slow truncate/writes to huge files at high offsets.
2602 Details : Directly associate cached pages to lock that protect those pages,
2603 this allows us to quickly find what pages to write and remove
2604 once lock callback is received.
2608 Description: more ldlm soft lockups
2609 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2610 starve other threads from the resource lock for a long time in
2611 case of long waiting queue, so change the debug level from
2612 D_OTHER to the less frequently used D_INFO.
2614 Severity : enhancement
2616 Description: add -gid, -group, -uid, -user options to lfs find
2620 Description: Hung threads in invalidate_inode_pages2_range
2621 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2622 once one is completed. As a result, some RPCs are stuck in the
2623 queue and are never sent.
2627 Description: filter threads hungs on waiting journal commit
2628 Details : Cleanup filter group llog code, then only filter group llog will
2629 be only created in the MDS/OST syncing process.
2633 Description: Procfs and llog threads access destoryed import sometimes.
2634 Details : Sync the import destoryed process with procfs and llog threads by
2635 the import refcount and semaphore.
2637 Severity : enhancement
2639 Description: openlock cache of b1_6 port to HEAD
2644 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2645 Details : If insertion of an extent fails, then discard the inode
2646 preallocation and free data blocks else it can lead to duplicate
2651 Description: don't always update ctime in ext3_xattr_set_handle()
2652 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2653 In some cases the ctime should not be updated, for example for
2654 2.0->1.8 compatibility it is necessary to delete an xattr and it
2655 should not update the ctime.
2659 Bugzilla : 15713/16362
2660 Description: Assertion in iopen_connect_dentry in 1.6.3
2661 Details : looking up an inode via iopen with the wrong generation number can
2662 populate the dcache with a disconneced dentry while the inode
2663 number is in the process of being reallocated. This causes an
2664 assertion failure in iopen since the inode's dentry list contains
2665 both a connected and disconnected dentry.
2669 Description: assertion failure in ldlm_handle2lock()
2670 Details : fix a race between class_handle_unhash() and class_handle2object()
2671 introduced in lustre 1.6.5 by bug 13622.
2676 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2677 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2682 Description: oops in sd_iostats_seq_show()
2683 Details : unloading/reloading the scsi low level driver triggers a kernel
2684 bug when trying to access the sd iostat file.
2689 Description: Kernel panics during QLogic driver reload
2690 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2691 patch, causing memory corruption.
2696 Description: journal_dev option does not work in b1_6
2697 Details : pass mount option during pre-mount.
2699 Severity : enhancement
2701 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2702 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2703 extent information of a file. It can be used to map logical blocks
2704 in a file to physical blocks in the block device.
2708 Description: LDLM soft lockups - improvement
2709 Details : It is be possible to send the lock handle along with each read
2710 or write request because the client is already doing a lock match
2711 itself so there isn't any reason the OST should have to re-do that
2715 Frequency : only on Cray X2
2717 Description: X2 build failures
2718 Details : fix build failures on Cray X2.
2722 Description: xid & resent requests
2723 Details : Initialize RPC XID from clock at startup (randomly if clock is
2726 Severity : enhancement
2728 Description: Add lustre_start utility to start or stop multiple Lustre servers
2733 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2734 Details : In case of memory pressure, list_del() can be called twice on
2735 req->rq_history_list, causing a kernel oops.
2739 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2740 Details : fix stack overflow in the distributed lock manager by defering
2741 export eviction after a failed AST to the elt thread instead of
2742 handling it in the dlm interpret routine.
2746 Description: Convert some comments to new format.
2747 Details : Update documenting comments to match doxygen conventions.
2751 Description: Grammar fixes.
2752 Details : A couple of trivial spelling fixes.
2756 Description: OSD_COUNTERS-mandatory
2757 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2758 to simplify the code.
2762 Description: simplify cmm_device freeing logic.
2763 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2767 Description: Add lockdep support to dt_object_operations locking interface.
2768 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2769 parameter indicating lock ordering. Update mdd code to use new
2774 Description: Introduce failloc constants for lockless IO tests.
2775 Details : Add two new failloc constants to test lockless IO. Only one of
2776 them in implemented---another is checked in yet to be landed
2781 Description: Add lockdep support for inode mutex.
2782 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2783 in the situations where Lustre has to lock more than one inode
2788 Description: Add optional invariants checking support.
2789 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2790 configure switch. This macro is to be used for consistency and
2791 sanity checks that are too expensive to be left in `production'
2796 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2797 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2802 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2803 Details : Use spin_lock_nested() in (the only) situation where more than
2804 one ldlm_lock is locked simultaneously. Also, fix possible
2805 dead-lock in ldlm_lock_change_resource() by enforcing particular
2810 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2811 Details : Instead of specifying each ldlm_lock call-back through separate
2812 parameter, wrap them into struct ldlm_callback_suite.
2816 Description: Kill join_lru obd method and its callers.
2817 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2818 in memory---a requirement implemented through ->o_join_lru() obd
2819 method in HEAD. Remove this method and its users.
2823 Description: Add asynchronous ldlm ENQUEUE completion handler.
2824 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2825 case that stock ldlm_completion_ast() cannot handle as it waits
2826 until lock is granted. Introduce new ldlm_completion_ast_async()
2827 for this. Also comment ldlm_completion_ast().
2831 Description: ldlm_error <-> errno conversion.
2832 Details : Add functions to map (rather arbitrary) between LDLM error codes
2833 and standard errno values. CLIO needs this to prevent LDLM specific
2834 constants from escaping ldlm and osc.
2838 Description: Kill unused ldlm_handle2lock_ns() function.
2839 Details : Kill unused ldlm_handle2lock_ns() function.
2843 Description: Add lu_ref support to ldlm_lock
2844 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2845 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2846 and ldlm_resource. LDLM interface has to be changed, because code that
2847 releases a reference on a lock, has to "know" what reference this is.
2848 In the most frequent case
2850 lock = ldlm_handle2lock(handle);
2852 LDLM_LOCK_PUT(lock);
2854 no changes are required. When any other reference (received _not_ from
2855 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2856 instead of LDLM_LOCK_PUT().
2858 Arguably, changes are pervasive, and interface requires some discipline
2859 for proper use. On the other hand, it was very instrumental in finding
2860 a few leaked lock references.
2864 Description: Add ldlm_lock_addref_try().
2865 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2866 attempts to addref a lock that might be being canceled
2871 Description: Add ldlm_weigh_callback().
2872 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2873 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2874 hard-coded `number of pages' logic.
2878 Description: Add lockdep annotations to llog code.
2879 Details : Use appropriately tagged _nested() locking calls in the places
2880 where llog takes more than one ->lgh_lock lock.
2884 Description: Add loi_kms_set().
2885 Details : Wrap kms updates into a helper function.
2889 Description: Constify instances of struct lsm_operations.
2890 Details : Constify instances of struct lsm_operations.
2894 Description: lu_conf support.
2895 Details : On a server, a file system object is uniquely identified
2896 by a fid, which is sufficient to locate and load all object
2897 state (inode). On a client, on the other hand, more data are
2898 necessary instantiate an object. Change lu_object_find() and
2899 friends to take additional `lu_conf' argument describing object.
2900 Typically this includes layout information.
2904 Description: lu_context fixes.
2905 Details : Introduce new lu_context functions that are needed on the client
2906 side, where some system threads (ptlrpcd) are shared by multiple
2907 modules, and so cannot be stopped during module shutdown.
2911 Description: Add start and stop methods to lu_device_type_operations.
2912 Details : Introduce two new methods in lu_device_type_operations, that are
2913 invoked when first instance of a given type is created and last one
2914 is destroyed respectively. This is need by CLIO.
2918 Description: Add lu_ref support to struct lu_device.
2919 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2920 track leaked references.
2924 Description: Introduce lu_kmem_descr.
2925 Details : lu_kmem_descr and its companion interface allow to create
2926 and destroy a number of kmem caches at once.
2930 Description: Fix lu_object finalization race.
2931 Details : Fix a race between lu_object_find() finding an object and its
2932 concurrent finalization. This race is (most likely) not possible
2933 on the server, but might happen on the client.
2937 Description: Introduce lu_ref interface.
2938 Details : lu_ref is a debugging module allowing to track references to
2939 a given object. It is quite cpu expensive, and has to be
2940 explicitly enabled with --enable-lu_ref. See usage description
2945 Description: Factor lu_site procfs stats into a separate function.
2946 Details : Separate lu_site stats printing code into a separate function
2947 that can be reused on a client.
2951 Description: Constify instances of struct {lu,dt,md}_device_operations.
2952 Details : Constify instances of struct {lu,dt,md}_device_operations.
2956 Description: Introduce struct md_site and move meta-data specific parts of
2957 struct lu_site here.
2958 Details : Move md-specific fields out of struct lu_site into special struct
2959 md_site, so that lu_site can be used on a client.
2963 Description: Kill mdd_lov_destroy().
2964 Details : Remove unused mdd code.
2968 Description: Add st_block checking to multistat.c.
2969 Details : Add st_block checking to multistat.c.
2973 Description: Add lu_ref support to struct obd_device.
2974 Details : Add lu_ref tracking to obd_device.
2978 Description: Kill obd_set_fail_loc().
2979 Details : Remove unused code.
2983 Description: Add special type for ptlrpc_request interpret functions.
2984 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2985 there. Actual usage will be in CLIO.
2989 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2990 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2991 doesn't work with lockdep.
2995 Description: Add rwv.c test program.
2996 Details : New testing program exercising readv(2) and writev(2) (Qian).
3000 Description: Add sendfile.c test program.
3001 Details : New testing program exercising sendfile(2) (Jay).
3005 Description: Ratelimit a message that can be very frequent.
3006 Details : Ratelimit a memory allocation failure message that can
3011 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3012 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3016 Description: Liblustre build fixes.
3017 Details : Liblustre build fixes.
3021 Description: libcfs: add cfs_{need,cond}_resched() interface.
3022 Details : libcfs: add cfs_{need,cond}_resched() definition and
3023 implementations for Linux, NT, and liblustre.
3025 Severity : enhancement
3027 Description: More exported tunables for mballoc
3028 Details : Add support for tunable preallocation window and new tunables for
3029 large/small requests
3033 Description: Detect corruption of block bitmap and checking for preallocations
3034 Details : Checks validity of on-disk block bitmap. Also it does better
3035 checking of number of applied preallocations. When corruption is
3036 found, it turns filesystem readonly to prevent further corruptions.
3040 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3041 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3042 window to 0. If it does hit cache miss, and read-pattern is still
3043 stride-io mode, does not reset the stride window, but also does not
3044 increase the stride window length in this case.
3048 Frequency : only for big-endian servers
3049 Description: Check if system is big-endian while mounting fs with extents feature
3050 Details : Mounting a filesystem with extents feature will fail on big-endian
3051 systems since ext3-based ldiskfs is not supported on big-endian
3052 systems. This can be over-riden with "bigendian_extents" mount option.
3054 Severity : enhancement
3056 Description: The root squash functionality
3057 Details : A security feature, which is to prevent users from being able
3058 to mount lustre on their desktop, run as root, and delete
3059 all of the files in the filesystem. The goal is accomplished by
3060 remapping user id (UID) and group id (GID) of the root user to
3061 a UID and GID specified by the system administartor via Lustre
3062 configuration management server (MGS). The functionality also
3063 allows to specify sets of clients for which the remapping does
3068 Description: Excessive recovery window
3069 Details : With AT enabled, the recovery window can be excessively long (6000+
3070 seconds). To address this problem, we no longer use
3071 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3072 timeout no longer depends on the service time, it is set to
3073 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3074 time via pb_service_time.
3078 Description: Watchdog triggered on MDS failover
3079 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3080 the OSTs know that the MDS "UUID" can be reused for the same export
3081 from a different NID, so we do not need to wait for the export to be
3085 Frequency : rare, only if using MMP with Linux RAID
3087 Description: MMP doesn't work with Linux RAID
3088 Details : While using HA for Lustre servers with Linux RAID, it is possible
3089 that MMP will not detect multiple mounts. To make this work we
3090 need to unplug the device queue in RAID when the MMP block is being
3091 written. Also while reading the MMP block, we should read it from
3092 disk and not the cached one.
3094 Severity : enhancement
3096 Description: open file using fid
3097 Details : A file can be opened using just its fid, like
3098 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3101 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3104 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3105 Details : add proper locking for creation of jbd_4k slab cache
3109 Description: MMP check in ext3_remount() fails without displaying any error
3110 Details : When multiple mount protection fails during remount, proper error
3113 Severity : enhancement
3115 Description: Allow stripe size to be up to 4G-64k
3116 Details : Fix math logic to allow large stripe sizes.
3120 Description: add check for >8TB ldiskfs filesystems
3121 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3122 Don't allow >8TB ldiskfs filesystems to be mounted without
3123 force_over_8tb mount option
3128 Descriptoin: OSS complains "too many blocks" and crashes
3129 Details : In filter_commitrw_write():
3130 1. lquota_chkquota();
3131 2. obdo_from_inode();
3132 3. lquota_pending_commit();
3133 "2." has changed uid and gid in this case, then caused
3134 lquota_chkquota() and lquota_pending_commit() to use
3135 different uid and gid. That is the root of the bug.
3139 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3140 Details : Detect and print out the 'stripe count = -1' case properly.
3144 Description: Out or order replies might be lost on replay
3145 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3146 request with tid smaller than one currently being added, add it
3147 to the start, not end of the list.
3149 --------------------------------------------------------------------------------
3151 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3153 * Support for kernels:
3154 2.6.5-7.283 (SLES 9),
3155 2.6.9-55.EL (RHEL 4),
3156 2.6.16.46-0.14 (SLES 10),
3157 2.6.18.8 vanilla (kernel.org)
3158 * Client support for unpatched kernels:
3159 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3160 2.6.16 - 2.6.22 vanilla (kernel.org)
3161 * Due to recently discovered recovery problems, we do not recommend
3162 using patchless RHEL 4 clients with this or any earlier release.
3163 * Recommended e2fsprogs version: 1.39.cfs8
3164 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3165 * Starting with this release, the ldiskfs backing filesystem required
3166 by Lustre is now in its own package, lustre-ldiskfs. This package
3167 should be installed. It is versioned separately from Lustre and
3168 may be released separately in future.
3172 Description: block reactivating mgc import until all deactivates complete
3173 Details : Fix race when failing back MDT/MGS to itself (testing)
3176 Frequency : at statup only
3178 Description: mds_lov_synchronize race leads to various problems
3179 Details : simultaneous MDT->OST connections at startup can cause the
3180 sync to abort, leaving the OSC in a bad state.
3182 Severity : enhancement
3184 Description: add optional extra BUILD_VERSION info
3185 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3186 to override the lustre version.
3189 Frequency : 2.6.18 servers only
3191 Description: ll_kern_mount() doesn't release the module reference
3192 Details : The ldiskfs module reference count never drops down to 0
3193 because ll_kern_mount() doesn't release the module reference.
3198 Description: server LBUG when using old ost_num_threads parameter
3199 Details : Accept the old ost_num_threads parameter but warn that it
3200 is deprecated, and fix an off-by-one error that caused an LBUG.
3205 Description: Transient SCSI error results in persistent IO issue
3206 Details : iobuf->dr_error is not reinitialized to 0 between two
3210 Frequency : sometimes when underlying device returns I/O errors
3212 Description: OSTs not going read-only during write failures
3213 Details : OSTs are not remounted read-only when the journal commit threads
3214 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3215 instead of the ext3 wrappers.
3219 Description: poor connect scaling with increasing client count
3220 Details : Don't run filter_grant_sanity_check for more than 100 exports
3221 to improve scaling for large numbers of clients.
3224 Frequency : SLES10 only
3226 Description: sanity-quota.sh quotacheck failed: rc = -22
3227 Details : Quotas cannot be enabled on SLES10.
3230 Frequency : liblustre clients only
3232 Description: getdirentries does not give error when run on compute nodes
3233 Details : getdirentries does not fail when the size specified as an argument
3234 is too small to contain at least one entry
3236 Severity : enhancement
3238 Description: Add LNET router traceability for debug purposes
3239 Details : If a checksum failure occurs with a router as part of the
3240 IO path, the NID of the last router that forwarded the bulk data
3241 is printed so it can be identified.
3246 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3247 Details : Due to a race condition, liblustre clients were occasionally
3248 evicted incorrectly.
3250 Severity : enhancement
3252 Description: lfs setstripe use optional parameters instead of postional
3255 Severity : enhancement
3257 Description: Nanosecond timestamp support for ldiskfs
3258 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3259 resolution timestamps. There is not yet support for this at
3260 the Lustre filesystem level.
3263 Frequency : during server recovery
3265 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3266 Details : request with rq_no_resend flag not awake l_wait_event if they get
3270 Frequency : nfs export on patchless client
3272 Description: connectathon hang when test nfs export over patchless client
3273 Details : Disconnected dentry cannot be found with lookup, so we do not need
3274 to unhash it or make it invalid
3277 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3278 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3279 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3284 Description: open req refcounting wrong on reconnect
3285 Details : If reconnect happened between getting open reply from server and
3286 call to mdc_set_replay_data in ll_file_open, we will schedule
3287 replay for unreferenced request that we are about to free.
3288 Subsequent close will crash in variety of ways.
3289 Check that request is still eligible for replay in
3290 mdc_set_replay_data().
3295 Description: disable writes to filesystem when reading health_check file
3296 Details : the default for reading the health_check proc file has changed
3297 to NOT do a journal transaction and write to disk, because this
3298 can cause reads of the /proc file to hang and block HA state
3299 checking on a healthy but otherwise heavily loaded system. It
3300 is possible to return to the previous behaviour during configure
3301 with --enable-health-write.
3303 Severity : enhancement
3305 Description: 64-bit inode version
3306 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3307 to the inode. This will be required for version-based recovery.
3312 Description: MDS fails to start if a duplicate client export is detected
3313 Details : in some rare cases it was possible for a client to connect to
3314 an MDS multiple times. Upon recovery the MDS would detect this
3315 and fail during startup. Handle this more gracefully.
3317 Severity : enhancement
3319 Description: Add -o localflock option to simulate old noflock
3321 Details : This will achieve local-only flock/fcntl locks
3327 Description: log_commit_thread vs filter_destroy race leads to crash
3328 Details : Take import reference before releasing llog record semaphore
3333 Description: Wrong request locking in request set processing
3334 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3335 imp_delayed_list, in this place should be used imp_lock.
3338 Frequency : when reconnection
3340 Description: Grant Leak when osc reconnect to OST
3341 Details : When osc reconnect ost, OST(filter) should check whether it
3342 should grant more space to client by comparing fed_grant and
3343 cl_avail_grant, and return the granted space to client instead
3344 of "new granted" space, because client will call osc_init_grant
3345 to update the client grant space info.
3348 Frequency : when client reconnect to OST
3350 Description: Grant Leak when osc do resend and replay bulk write
3351 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3352 bulk write request, because the grant info while be sync between
3353 OSC and OST when reconnect, and we should ignore the grant info
3354 these of resend/replay write req.
3359 Description: Grant space more than available left space sometimes.
3360 Details : When then OST is about to be full, if two bulk writing from
3361 different clients came to OST. Accord the avaliable space of the
3362 OST, the first req should be permitted, and the second one
3363 should be denied by ENOSPC. But if the seconde arrived before
3364 the first one is committed. The OST might wrongly permit second
3365 writing, which will cause grant space > available space.
3368 Frequency : when client is evicted
3370 Description: Grant might be wrongly erased when osc is evicted by OST
3371 Details : when the import is evicted by server, it will fork another
3372 thread ptlrpc_invalidate_import_thread to invalidate the
3373 import, where the grant will be set to 0. While the original
3374 thread will update the grant it got when connecting. So if
3375 the former happened latter, the grant will be wrongly errased
3376 because of this race.
3381 Description: Checking Stale with correct fid
3382 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3383 to check whether it is stale, because sometimes, we want the
3384 enqueue happened anyway, and op_data.fid2 will not be initialized.
3386 Severity : enhancement
3388 Description: update patchless client
3389 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3392 Frequency : only with 2.4 kernel
3394 Description: random memory corruption
3395 Details : size of struct ll_inode_info is to big for union inode.u and this
3396 can be cause of random memory corruption.
3401 Description: Memory leak in recovery
3402 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3403 It should also check obd_fail before fsfilt_start, otherwise if
3404 fsfilt_start return -EROFS,(failover mds during mds recovery).
3405 then the req will return with repmsg->transno = 0 and rc = EROFS.
3406 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3407 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3408 be freed no matter whether fsfilt_commit success or not.
3411 Frequency : only with huge count clients
3413 Description: Prevents from taking the superblock lock in llap_from_page for
3415 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3416 for llap_from_page call in ll_removepage prevents from taking the
3417 superblock lock for a soon died page.
3422 Description: Not check open intent error before release open handle
3423 Details : in some rare cases, the open intent error is not checked before
3424 release open handle, which may cause
3425 ASSERTION(open_req->rq_transno != 0), because it tries to release
3426 the failed open handle.
3431 Description: Set cat log bitmap only after create log success.
3432 Details : in some rare cases, the cat log bitmap is set too early. and it
3433 should be set only after create log success.
3437 Description: Accessing a block bevice can re-enable I/O when Lustre is
3438 tearing down a device.
3439 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3444 Description: service threads may hog cpus when there are a lot of requests
3445 Details : Insert cond_resched to give other threads a chance to use some CPU
3450 Description: the cat log was not initialized in recovery
3451 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3452 unlink log on mds will not be initialized until mds post
3453 recovery. And also in mds post recovery, the unlink log will
3454 initialization will be done asynchronausly, so there will be race
3455 between add unlink log and unlink log initialization.
3459 Description: brw_stats were being printed incorrectly
3460 Details : brw_stats were being printed as log2 but all of them were not
3461 recorded as log2. Also remove some code duplication arising from
3462 filter_tally_{read,write}.
3466 Frequency : rare, only in recovery.
3467 Description: ASSERTION(req->rq_type != LI_POISON) failed
3468 Details : imp_lock should be held while iterating over imp_sending_list for
3469 prevent destroy request after get timeout in ptlrpc_queue_wait.
3473 Description: replay-single.sh test 52 fails
3474 Details : A lock's skiplist need to be cleanup when it being unlinked
3475 from its resource list.
3479 Description: Short directio read returns full requested size rather than
3481 Details : Direct I/O operations should return actual amount of bytes
3482 transferred rather than requested size.
3484 Severity : enhancement
3486 Description: metadata RPC reduction (e.g. for rm performance)
3487 Details : decrease the amount of synchronous RPC between clients and servers
3488 by canceling conflicing lock before the operation on the client
3489 and packing thier handles into the main operation RPC to server.
3491 Severity : enhancement
3493 Description: Async OSC create to avoid the blocking unnecessarily.
3494 Details : If an OST has no remaining object, system will block on the
3495 creation when it needs to create a new object on this OST. Now,
3496 ways use pre-created objects when available, instead of blocking on
3497 an empty osc while others are not empty. If we must block, we
3498 block for the shortest possible period of time.
3502 Description: improve handling recoverable errors
3503 Details : if request processig with error which can be recoverable on server
3504 request should be resend, otherwise page released from cache and
3507 Severity : enhancement
3509 Description: refine locking for avoid write wrong info into lov_objid file
3510 Details : fix possible races with add new target and write/update data in
3513 --------------------------------------------------------------------------------
3515 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3520 Frequency : on some architectures
3522 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3523 Details : Uninitialized flags sometimes cause configuration commands to
3527 Frequency : patchless clients only
3529 Description: missing __iget() symbol export
3530 Details : The __iget() symbol export is missing. To avoid the need for
3531 this on patchless clients the deathrow inode reaper is turned
3532 off, and we depend on the VM to clean up old inodes. This
3533 dependency was during via the fix for bug 12181.
3535 --------------------------------------------------------------------------------
3537 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3539 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3540 INTEROPERATE with older versions automatically. In many cases a
3541 special upgrade step is needed. Please read the
3542 user documentation before upgrading any part of a live system.
3543 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3544 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3545 * WARNING: Lustre configuration and startup changes are required with
3546 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3550 Severity : enhancement
3552 Description: MountConf
3553 Details : Lustre configuration is now managed via mkfs and mount
3554 commands instead of lmc and lconf. New obd types (MGS, MGC)
3555 are added for dynamic configuration management. See
3556 https://mail.clusterfs.com/wikis/lustre/MountConf for
3559 Severity : enhancement
3561 Description: dynamic OST addition
3562 Details : OSTs can now be added to a live filesystem
3564 Severity : enhancement
3566 Description: startup order invariance
3567 Details : MDTs and OSTs can be started in any order. Clients only
3568 require the MDT to complete startup.
3570 Severity : enhancement
3572 Description: parallel, asynchronous orphan cleanup
3573 Details : orphan cleanup is now performed in separate threads for each
3574 OST, allowing parallel non-blocking operation.
3576 Severity : enhancement
3578 Description: optimized stripe assignment
3579 Details : stripe assignments are now made based on ost space available,
3580 ost previous usage, and OSS previous usage, in order to try
3581 to optimize storage space and networking resources.
3583 Severity : enhancement
3585 Description: Permanently set tunables
3586 Details : All writable /proc/fs/lustre tunables can now be permanently
3587 set on a per-server basis, at mkfs time or on a live system.
3589 Severity : enhancement
3591 Description: Lustre message v2
3592 Details : Add lustre message format v2.
3594 Severity : enhancement
3596 Description: client OST exclusion list
3597 Details : Clients can be started with a list of OSTs that should be
3598 declared "inactive" for known non-responsive OSTs.
3602 Description: SPEC SFS validation failure on NFS v2 over lustre.
3603 Details : Changes the blocksize for regular files to be 2x RPC size,
3604 and not depend on stripe size.
3606 Severity : enhancement
3608 Description: Multiple MD RPCs in flight.
3609 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3610 To avoid overly-overloading MDS, introduce a limit on number of
3611 MDS RPCs in flight for a single client and add /proc controls
3612 to adjust this limit.
3614 Severity : enhancement
3616 Description: client read/write statistics
3617 Details : Add client read/write call usage stats for performance
3618 analysis of user processes.
3619 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3620 file access. extents_stats shows chunk size distribution.
3621 extents_stats_per_process show chunk size distribution per
3624 Severity : enhancement
3626 Description: mds statistics
3627 Details : Add detailed mds operations statistics in
3628 /proc/fs/lustre/mds/*/stats.
3632 Description: Failure of copying files with lustre special EAs.
3633 Details : Client side always return success for setxattr call for lustre
3634 special xattr (currently only "trusted.lov").
3639 Description: Refcount LNET uuids
3640 Details : The global LNET uuid list grew linearly with every startup;
3641 refcount repeated list entries instead of always adding to
3644 Severity : enhancement
3646 Description: Dynamic service threads
3647 Details : Within a small range, start extra service threads
3648 automatically when the request queue builds up.
3651 Frequency : mixed-endian client/server environments
3653 Description: mixed-endian crashes
3654 Details : The new msg_v2 system had some failures in mixed-endian
3657 Severity : enhancement
3659 Description: Easy OST removal
3660 Details : OSTs can be permanently deactivated with e.g. 'lctl
3661 conf_param lustre-OST0001.osc.active=0'
3663 Severity : enhancement
3665 Description: MGS proc entries
3666 Details : Added basic proc entries for the MGS showing what filesystems
3669 Severity : enhancement
3671 Description: provide MGS failover
3672 Details : Added config lock reacquisition after MGS server failover.
3674 Severity : enhancement
3676 Description: add Linux 2.4 support
3677 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3681 Description: a large application tries to do I/O to the same resource and dies
3682 in the middle of it.
3683 Details : Check the req->rq_arrival time after the call to
3684 ost_brw_lock_get(), but before we do anything about
3685 processing it & sending the BULK transfer request. This
3686 should help move old stale pending locks off the queue as
3687 quickly as obd_timeout.
3690 Frequency : when an incorrect nid is specified during startup
3692 Description: ptlrpc connect to non-existant node causes kernel crash
3693 Details : LNET can't be re-entered from an event callback, which
3694 happened when we expire a message after the export has been
3695 cleaned up. Instead, hand the zombie cleanup off to another
3698 Severity : enhancement
3700 Description: plain/inodebits lock performance improvement
3701 Details : Grouping plain/inodebits in granted list by their request modes
3702 and bits policy, thus improving the performance of search through
3706 Frequency : only if OST filesystem is corrupted
3708 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3709 Details : for a short time RPCs with bulk IO are in the replay list,
3710 but replay of bulk IOs is unimplemented. If the OST filesystem
3711 is corrupted due to disk cache incoherency and then replay is
3712 started it is possible to trip an assertion. Avoid putting
3713 committed RPCs into the replay list at all to avoid this issue.
3716 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3718 Description: System hang on startup
3719 Details : This bug allowed the liblustre (e.g. catamount) client to
3720 return to the app before handling all startup RPCs. This
3721 could leave the node unresponsive to lustre network traffic
3722 and manifested as a server ptllnd timeout.
3724 Severity : enhancement
3726 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3727 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3728 boolean to control whether to print peer debug info when a
3729 client's RPC times out.
3732 Frequency : only for kernels with patches from Lustre below 1.4.3
3734 Description: Remove old rdonly API
3735 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3738 Frequency : only for devices with external journals
3740 Description: Set external device read-only also
3741 Details : During a commanded failover stop, we set the disk device
3742 read-only while the server shuts down. We now also set any
3743 external journal device read-only at the same time.
3746 Frequency : when upgrading from 1.4 while trying to change parameters
3748 Description: The wrong (new) MDC name was used when setting parameters for
3749 upgraded MDT's. Also allows changing of OSC (and MDC)
3750 parameters if --writeconf is specified at tunefs upgrade time.
3753 Frequency : when setting specific ost indicies
3755 Description: QOS code breaks on skipped indicies
3756 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3757 created with --index need not be sequential.
3759 Severity : enhancement
3761 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3762 Details : The uninit_groups feature works in conjunction with the kernel
3763 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3764 up the pass1 processing of e2fsck. This is a read-only feature
3765 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3766 mount filesystems that have had this feature enabled.
3768 Severity : enhancement
3770 Description: Improve multi-block allocation algorithm to avoid fragmentation
3771 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3772 allocation locality and avoid filesystem fragmentation.
3774 ------------------------------------------------------------------------------
3776 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3778 * Support for kernels:
3779 2.4.21-47.0.1.EL (RHEL 3)
3780 2.6.5-7.283 (SLES 9)
3781 2.6.9-42.0.10.EL (RHEL 4)
3782 2.6.12.6 vanilla (kernel.org)
3783 2.6.16.27-0.9 (SLES 10)
3784 * Recommended e2fsprogs version: 1.39.cfs5
3786 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3790 Frequency : occasional, depends on client load and configuration
3791 Bugzilla : 12181, 12203
3792 Description: data loss for recently-modified files
3794 Details : In some cases it is possible that recently written or created
3795 files may not be written to disk in a timely manner (this should
3796 normally be within 30s unless client IO load is very high).
3797 The problem appears as zero-length files or files that are a
3798 multiple of 1MB in size after a client crash or client eviction
3799 that are missing data at the end of the file.
3801 This problem is more likely to be hit on clients where files are
3802 repeatedly created and unlinked in the same directory, clients
3803 have a large amount of RAM, have many CPUs, the filesystem has
3804 many OSTs, the clients are rebooted frequently, and/or the files
3805 are not accessed by other nodes after being written.
3807 The presence of the problem can be detected by looking at
3808 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3809 smaller than the second (nr_unused) then dirty files will not
3810 be flushed automatically to disk. "sync; sleep 10" should be
3811 run several times on the node before unmounting it to update
3812 Lustre (this is also safe to run on nodes without this problem).
3814 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3815 that can cause this same problem, so customers using that kernel
3816 also need to update the kernel in addition to Lustre. In order
3817 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3819 It is normal that files written just before a client crash (less
3820 than 30s) may not yet have been flushed to disk, even for local
3824 Frequency : frequent on thin XT3 nodes
3826 Description: UUID collision on thin XT3 Linux nodes
3827 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3828 randomly, since we relied on an insufficiently-seeded PRNG.
3833 Description: OSS hangs after "All ost request buffers busy"
3834 Details : A deadlock between quota and journal operations caused OSS
3835 hangs after printing "All ost request buffers busy."
3838 Frequency : always on liblustre builds
3840 Description: Cleanup compiler warnings on liblustre
3843 Frequency : always on liblustre builds on XT3
3845 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3849 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3850 > 32000 subdirectories
3851 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3852 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3853 subdirectory count crosses 32000. This will aid e2fsck to
3854 correctly handle more than 32000 subdirectories.
3857 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3859 Description: System hang on startup
3860 Details : This bug allowed the liblustre (e.g. catamount) client to
3861 return to the app before handling all startup RPCs. This
3862 could leave the node unresponsive to lustre network traffic
3863 and manifested as a server ptllnd timeout.
3865 Severity : enhancement
3867 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3868 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3869 boolean to control whether to print peer debug info when a
3870 client's RPC times out.
3875 Description: make O_SYNC working on 2.6 kernels
3876 Details : 2.6 kernels use different method for mark pages for write,
3877 so need add a code to lustre for O_SYNC work.
3882 Description: Failure to close file and release space on NFS
3883 Details : Put inode details into lock acquired in ll_intent_file_open.
3884 Use mdc_intent_lock in ll_intent_open to properly
3885 detect all kind of errors unhandled by mdc_enqueue.
3890 Description: proc file read during shutdown sometimes raced obd removal,
3892 Details : Add lock to prevent obd access after proc file removal.
3895 Frequency : Only for files larger than 4GB on 32-bit clients.
3897 Description: improperly doing page alignment of locks
3898 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3899 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3904 Description: RPCs being resent when they shouldn't be
3905 Details : Some RPCs that should not be resent are being resent. This
3906 can cause inconsistencies in the RPC state machine. Do not
3907 resend such requests.
3910 Frequency : rare, only with NFS export
3912 Description: Crash on NFS re-export node
3913 Details : under very unusual load conditions an assertion is hit in
3914 ll_intent_file_open()
3917 Frequency : only if OST filesystem is corrupted
3919 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3920 Details : for a short time RPCs with bulk IO are in the replay list,
3921 but replay of bulk IOs is unimplemented. If the OST filesystem
3922 is corrupted due to disk cache incoherency and then replay is
3923 started it is possible to trip an assertion. Avoid putting
3924 committed RPCs into the replay list at all to avoid this issue.
3929 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3930 Details : Large single O_DIRECT read and write calls can fail to allocate
3931 a sufficiently large buffer to process the request. In case of
3932 allocation failure the allocation is retried with a smaller
3933 buffer and broken into smaller requests.
3935 Severity : enhancement
3937 Description: Add -o localflock option to simulate old noflock behaviour.
3938 Details : This will achieve local-only flock/fcntl locks coherentness.
3943 Description: versioning check is incomplete
3944 Details : Checking the version difference of client vs. server, report
3945 error if the gap is too big.
3950 Description: add support PG_writeback bit
3951 Details : add support for PG_writeback bit for Lustre, for more carefull
3952 work with page cache in 2.6 kernel. This also fix some deadlocks
3953 and remove hack for work O_SYNC with 2.6 kernel.
3955 Severity : enhancement
3957 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3958 Details : The uninit_groups feature works in conjunction with the kernel
3959 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3960 up the pass1 processing of e2fsck. This is a read-only feature
3961 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3962 mount filesystems that have had this feature enabled.
3964 Severity : enhancement
3966 Description: Improve multi-block allocation algorithm to avoid fragmentation
3967 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3968 allocation locality and avoid filesystem fragmentation.
3970 ------------------------------------------------------------------------------
3972 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3974 * Support for kernels:
3975 2.6.9-42.0.3.EL (RHEL 4)
3976 2.6.5-7.276 (SLES 9)
3977 2.4.21-47.0.1.EL (RHEL 3)
3978 2.6.12.6 vanilla (kernel.org)
3979 2.6.16.21-0.8 (SLES10)
3980 * Recommended e2fsprogs version: 1.39.cfs2-0
3982 * The backwards-compatible /proc/sys/portals symlink has been removed
3983 in this release. Before upgrading, please ensure that you change
3984 any configuration scripts or /etc/sysctl.conf files that access
3985 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3986 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3987 in advance of the upgrade on any system running Lustre 1.4.6 or
3988 newer, since /proc/sys/lnet was added in that version.
3989 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3993 Frequency : only when quota is used
3995 Description: avoid scanning export list for quota master
3996 Details : Change the algorithms to avoid scanning export list in order
3997 to improve the efficiency.
4000 Frequency : MDS failover only, very rarely
4002 Description: "went back in time" messages on mds failover
4003 Details : The greatest transno may be lost when the current operation
4004 finishes with an error (transno==0) and the client's last_rcvd
4005 record is over-written. Save the greatest transno in the
4006 mds_last_transno for this case.
4009 Frequency : always for specific kernels and striping counts
4011 Description: client may get "Matching packet too big" without ACL support
4012 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4013 when trying to access files in certain configurations. The
4014 clients should in fact be denied when mounting because they do
4015 not understand ACLs.
4018 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4020 Description: many clients connecting with IO in progress causes connect timeouts
4021 Details : Avoid synchronous journal commits to avoid delays caused by many
4022 clients connecting/disconnecting when bulk IO is in progress.
4023 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4024 OST_IO_PORTAL to avoid delays behind potentially many pending
4028 Frequency : occasionally with multiple writers to a single file
4030 Description: shared writes to file may result in wrong size reported by stat()
4031 Details : Allow growing of kms when extent lock is cancelled
4034 Frequency : always with random mmap IO to multi-striped file
4036 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4037 Details : Only if the hole is at the end of OST object so that kms is too
4038 small. Fix is to increase kms accordingly in ll_nopage.
4041 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4043 Description: writes to a missing object would leak memory on the OST
4044 Details : If there is an inconsistency between the MDS and OST filesystems,
4045 such that the MDS references an object that doesn't exist, writes
4046 to that object will leak memory due to incorrect cleanup in the
4047 error handling path, eventually running out of memory on the OST.
4052 Description: Creating too long symlink causes lustre errors
4053 Details : Check symlink and name lengths before sending requests to MDS.
4056 Frequency : only if flock is enabled (not on by default)
4058 Description: posix locks not released on fd closure on 2.6.9+
4059 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4060 kernels, this caused such locks not to be released on fd close and
4061 then assertions on fs unmount about still used locks.
4064 Frequency : MDS failover only, very rarely
4066 Description: clients may get ASSERTION(granted_lock != NULL)
4067 Details : When request was taking a long time, and a client was resending
4068 a getattr by name lock request. The were multiple lock requests
4069 with the same client lock handle and
4070 mds_getattr_name->fixup_handle_for_resent_request found one of the
4071 lock handles but later failed with ASSERTION(granted_lock != NULL).
4076 Description: handle->h_buffer_credits > 0, assertion failure
4077 Details : h_buffer_credits is zero after truncate, causing assertion
4078 failure. This patch extends the transaction or creates a new
4082 Frequency : NFS re-export or patchless client
4083 Bugzilla : 11179, 10796
4084 Description: Crash on NFS re-export node (__d_move)
4085 Details : We do not want to hash the dentry if we don't have a lock.
4086 But if this dentry is later used in d_move, we'd hit uninitialised
4087 list head d_hash, so we just do this to init d_hash field but
4088 leave dentry unhashed.
4091 Frequency : NFS re-export or patchless client
4093 Description: NFS exports has problem with symbolic link
4094 Details : lustre client didn't properly install dentry when re-exported
4095 to NFS or running patchless client.
4098 Frequency : NFS re-export or patchless client
4100 Description: Various nfs/patchless fixes.
4101 Details : fixes reuse disconected alias for lookup process - this fixes
4102 warning "find_exported_dentry: npd != pd",
4103 fix permission error with open files at nfs.
4104 fix apply umask when do revalidate.
4107 Frequency : occasional
4109 Description: Crash on NFS re-export node
4110 Details : calling clear_page() on the wrong pointer triggered oops in
4111 generic_mapping_read().
4114 Frequency : rarely, using O_DIRECT IO
4116 Description: unaligned directio crashes client with LASSERT
4117 Details : check for unaligned buffers before trying any requests.
4120 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4122 Description: stale data returned from RAID cache
4123 Details : If only a small amount of IO is done to the RAID device before
4124 reading it again it is possible to get stale data from the RAID
4125 cache instead of reading it from disk.
4128 Frequency : always for sles10 kernel
4130 Description: sles10 support
4131 Details : ll_follow_link: compile fixes and using of nd_set_link
4132 under newer kernels.
4135 Frequency : depends on arch, kernel and compiler version, always on sles10
4138 Description: recursive or deep enough symlinks cause stack overflow
4139 Details : getting rid of large stack-allocated variable in
4143 Frequency : depends on hardware
4145 Description: lustre write performance loss in the SLES10 kernel
4146 Details : the performance loss is caused by using of write barriers in the
4147 ext3 code. The SLES10 kernel turns barrier support on by
4148 default. The fix is to undo that change for ldiskfs.
4150 ------------------------------------------------------------------------------
4152 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4154 * Support for kernels:
4155 2.6.9-42.0.3EL (RHEL 4)
4156 2.6.5-7.276 (SLES 9)
4157 2.4.21-47.0.1.EL (RHEL 3)
4158 2.6.12.6 vanilla (kernel.org)
4162 Frequency : quota enabled and large files being deleted
4164 Description: releasing more than 4GB of quota at once hangs OST
4165 Details : If a user deletes more than 4GB of files on a single OST it
4166 will cause the OST to spin in an infinite loop. Release
4167 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4172 Description: statfs data retrieved from /proc may be stale or zero
4173 Details : When reading per-device statfs data from /proc, in the
4174 {kbytes,files}_{total,free,avail} files, it may appear
4175 as zero or be out of date.
4178 Frequency : systems with MD RAID1 external journal devices
4180 Description: lconf's call to blkid is confused by RAID1 journal devices
4181 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4182 of returning all block devices that match the journal UUID.
4185 Frequency : always, for aggregate stripe size over 4GB
4187 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4188 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4189 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4192 Frequency : always if "lfs find" used on a local file/directory
4194 Description: "lfs find" segfaults if used on a local file/directory
4195 Details : The case where a directory component was not specified wasn't
4196 handled correctly. Handle this properly.
4199 Frequency : always on ppc64
4201 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4202 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4203 perform endianness conversion.
4206 Frequency : rarely (truncate to non-zero file size after write under load)
4207 Bugzilla : 10730, 10687
4208 Description: Files padded with zeros to next 4K multiple
4209 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4210 it is possible that files that are truncated to a non-zero size
4211 immediately after being written are filled with zero bytes beyond
4212 the truncated size. No file data is lost.
4214 Severity : enhancement
4216 Description: Allow recovery/failover for liblustre clients.
4217 Details : liblustre clients were unaware of failover configurations until
4220 Severity : enhancement
4222 Description: user file locks should fail when not mounting with flock option
4223 Details : Set up an error-returning stub in ll_file_operations.lock field
4224 to prevent incorrect behaviour when client is mounted without
4225 flock option. Also, set up properly f_op->flock field for
4229 Frequency : always on ia64
4231 Description: "lfs df" loops on printing out MDS statfs information
4232 Details : The obd_ioctl_data was not initialized and in some systems
4233 this caused a failure during the ioctl that did not return
4234 an error. Initialize the struct and return an error on failure.
4237 Frequency : SLES 9 only
4239 Description: Error of copying files with lustre special EAs as root
4240 Details : Client side always return success for setxattr call for lustre
4241 special xattr (currently only "trusted.lov").
4244 Frequency : rarely on clusters with both ia64+i386 clients
4246 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4247 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4248 clients are doing concurrent writes to the same file it is
4249 possible that the ia64 clients may LASSERT because the OST
4250 extent locks are not PAGE_SIZE aligned. Ensure that grown
4251 locks are always aligned on the request boundary.
4254 Frequency : specific use, occasional
4256 Description: Overwriting in use executable truncates on-disk binary image
4257 Details : If one node attempts to overwrite an executable in use by
4258 another node, we now correctly return ETXTBSY instead of
4259 truncating the file.
4261 Severity : enhancement
4263 Description: Async OSC create to avoid the blocking unnecessarily.
4264 Details : If an OST has no remaining object, system will block on the creating
4265 when need to create a new object on this OST. Now, ways use
4266 pre-created objects when available, instead of blocking on an
4267 empty osc while others are not empty. If we must block, we block
4268 for the shortest possible period of time.
4273 Description: chmod on Lustre root is propagated to other clients
4274 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4275 invalid by the follow_mount time.
4278 Frequency : liblustre clients only
4280 Description: Race in 'instant cancel' lock handling could lead to such locks
4281 never to be granted in case of SMP MDS
4282 Details : Do not destroy not yet granted but cbpending locks in
4286 Frequency : replay/resend of open
4288 Description: non null lock assetion failure in mds_intent_policy
4289 Details : Trying to replay/resend lockless open requests resulted in
4290 mds_open() returning 0 with no lock. Now it sets a flag if
4291 a lock is going to be returned.
4293 Severity : enhancement
4295 Description: Checksum enhancements
4296 Details : New checksum enhancements allow for resending RPCs that failed
4299 Severity : enhancement
4301 Description: Tunables on number of dirty pages in cacche
4302 Details : Allow to set limit on number of dirty pages cached.
4307 Description: client crash on unmount - lock still has references
4308 Details : In some error handling cases it was possible to leak a lock
4309 reference on a client while accessing a file. This was not
4310 harmful to the client during operation, but would cause the
4311 client to crash when the filesystem is unmounted.
4314 Frequency : specific case, rare
4316 Description: ETXTBSY on mds though file not in use by client
4317 Details : ETXTBSY is no longer incorrectly returned when attempting to
4318 chmod or chown a directory that the user previously tried to
4319 execute or a currently-executing binary.
4322 Frequency : extremely rare except on liblustre-based clients
4324 Description: Lustre space not freed when files are deleted
4325 Details : Clean up open-unlinked files after client eviction. Previously
4326 the unlink was skipped and the files remained as orphans.
4331 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4332 Details : In some cases it was possible to send a blocking callback to
4333 a client doing a glimpse, even though that client didn't get
4334 a lock granted. When the glimpse lock is cancelled on the OST
4335 the freed lock is left on the waiting list and corrupted the list.
4338 Frequency : all core dumps
4340 Description: Broke core dumps to lustre
4341 Details : Negative dentry may be unhashed if parent does not have UPDATE
4342 lock, but some callers, e.g. do_coredump, expect dentry to be
4343 hashed after successful create, hash it in ll_create_it.
4345 ------------------------------------------------------------------------------
4347 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4349 * Support for kernels:
4350 2.6.9-42.0.2.EL (RHEL 4)
4351 2.6.5-7.276 (SLES 9)
4352 2.4.21-40.EL (RHEL 3)
4353 2.6.12.6 vanilla (kernel.org)
4357 Frequency : always on RHEL 3
4359 Description: Number of open files grows over time
4360 Details : The number of open files grows over time, whether or not
4361 Lustre is started. This was due to a filp leak introduced
4362 by one of our kernel patches.
4364 ------------------------------------------------------------------------------
4366 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4368 * Support for kernels:
4369 2.6.9-42.EL (RHEL 4)
4370 2.6.5-7.276 (SLES 9)
4371 2.4.21-40.EL (RHEL 3)
4372 2.6.12.6 vanilla (kernel.org)
4377 Bugzilla : 5719, 9635, 9792, 9684
4378 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4379 Details : If a server is under heavy load and cannot reply to new
4380 connection requests before the client resends the (re)connect,
4381 the connection handling code can behave badly if two service
4382 threads are concurrently handing separate (re)connections from
4383 the same client. Add better locking to the connection handling
4384 code, and ensure that only a single connection will be processed
4385 for a given client UUID, even if the lock is dropped.
4387 Severity : enhancement
4389 Description: add TCP zero-copy support to kernel
4390 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4391 sends if the hardware supports scatter-gather and checksumming.
4392 This allows socklnd to do client-write and server-read more
4393 efficiently and reduce CPU utilization from skbuf copying.
4396 Frequency : only if NFS exporting from client
4398 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4399 Details : If setting attributes on a file created under NFS that had
4400 never been opened it would be possible to oops the client
4401 if the file had no objects.
4404 Frequency : always for liblustre
4406 Description: liblustre client does MDS+OSTs setattr RPC for each write
4407 Details : When doing a write from a liblustre client, the client
4408 incorrectly issued an RPC to the MDS and each OST the file was
4409 striped over in order to update the timestamps. When writing
4410 with small chunks and many clients this could overwhelm the MDS
4411 with RPCs. In all cases it would slow down the write because
4412 these RPCs are unnecessary.
4414 Severity : enhancement
4416 Description: allow number of MDS service threads to be changed at module load
4417 Details : It is now possible to change the number of MDS service threads
4418 running. Adding "options mds mds_num_threads={N}" to the MDS's
4419 /etc/modprobe.conf will set the number of threads for the next
4420 time Lustre is restarted (assuming the "mds" module is also
4421 reloaded at that time). The default number of threads will
4422 stay the same, 32 for most systems.
4427 Description: OST crash if filesystem is unformatted or corrupt
4428 Details : If an OST is started on a device that has never been formatted
4429 or if the filesystem is corrupt and cannot even mount then the
4430 error handling cleanup routines would dereference a NULL pointer.
4435 Description: NULL pointer deref in llap_from_page.
4436 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4437 incorrect) mapping if the page was truncated/reclaimed while it was
4438 searched for. Check for this condition and skip such pages when
4439 doing readahead. Introduce extra check to llap_from_page() to
4440 verify page->mapping->host is non-NULL (so page is not anonymous).
4443 Frequency : Sometimes when using sys_sendfile
4445 Description: "page not covered by a lock" warnings from ll_readpage
4446 Details : sendfile called ll_readpage without right page locks present.
4447 Now we introduced ll_file_sendfile that does necessary locking
4448 around call to generic_file_sendfile() much like we do in
4452 Frequency : with certain MDS communication failures at client mount time
4454 Description: NULL pointer deref after failed client mount
4455 Details : a client connection request may delayed by the network layer
4456 and not be sent until after the PTLRPC layer has timed out the
4457 request. If the client fails the mount immediately it will try
4458 to clean up before the network times out the request. Add a
4459 reference from the request import to the obd device and delay
4460 the cleanup until the network drops the request.
4463 Frequency : occasionally during client (re)connect
4465 Description: assertion failure during client (re)connect
4466 Details : processing a client connection request may be delayed by the
4467 client or server longer than the client connect timeout. This
4468 causes the client to resend the connection request. If the
4469 original connection request is replied in this interval, the
4470 client may trip an assertion failure in ptlrpc_connect_interpret()
4471 which thought it would be the only running connect process.
4474 Frequency : only with obd_echo servers and clients that are rebooted
4476 Description: kernel BUG accessing uninitialized data structure
4477 Details : When running an obd_echo server it did not start the ping_evictor
4478 thread, and when a client was evicted an uninitialized data
4479 structure was accessed. Start the ping_evictor in the RPC
4480 service startup instead of the OBD startup.
4482 Severity : enhancement
4483 Bugzilla : 10193 (patchless)
4484 Description: Remove dependency on various unexported kernel interfaces.
4485 Details : No longer need reparent_to_init, exit_mm, exit_files,
4486 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4489 Frequency : rare (only users of deprecated and unsupported LDAP config)
4491 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4492 Details : LDAP apparently contains 'lustreName' attributes instead of
4493 'name'. A simple remapping of the name is sufficient.
4496 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4498 Description: waiting_locks_callback trips kernel BUG if client is evicted
4499 Details : Running with the dump_on_timeout debug flag turned on makes
4500 it possible that the waiting_locks_callback() can try to dump
4501 the Lustre kernel debug logs from an interrupt handler. Defer
4502 this log dumping to the expired_lock_main() thread.
4504 Severity : enhancement
4506 Description: Support NFS exporting on 2.6 kernels.
4507 Details : Implement non-rawops metadata methods for NFS server to use without
4508 changing NFS server code.
4511 Frequency : very rare (synthetic metadata workload only)
4513 Description: two racing renames might cause an MDS thread to deadlock
4514 Details : Running the "racer" program may cause one MDS thread to rename
4515 a file from being the source of a rename to being the target of
4516 a rename at exactly the same time that another thread is doing
4517 so, and the second thread has already enqueued these locks after
4518 doing a lookup of the target and is trying to relock them in
4519 order. Ensure that we don't try to re-lock the same resource.
4522 Frequency : only very large systems with liblustre clients
4524 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4525 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4526 all OSTs in parallel. This allows the eviction of stale liblustre
4527 clients to proceed much faster than if they were done in series,
4528 and also offers similar improvements for other set_info RPCs.
4533 Description: excessive CPU usage during initial read phase on client
4534 Details : During the initial read phase on a client, it would agressively
4535 retry readahead on the file, consuming too much CPU and impacting
4536 performance (since 1.4.5.8). Improve the readahead algorithm
4537 to avoid this, and also improve some other common cases (read
4538 of small files in particular, where "small" is files smaller than
4539 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4544 Description: MDS crash when receiving packet with unknown intent.
4545 Details : Do not LBUG in unknown intent case, just return -EFAULT
4547 Severity : enhancement
4548 Bugzilla : 9293, 9385
4549 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4550 Details : Do not serialize getattr (non-intent version) and statfs.
4553 Frequency : occasional, when OST network is overloaded/intermittent
4555 Description: client evicted by OST after bulk IO timeout
4556 Details : If a client sends a bulk IO request (read or write) the OST
4557 may evict the client if it is unresposive to its data GET/PUT
4558 request. This is incorrect if the network is overloaded (takes
4559 too long to transfer the RPC data) or dropped the OST GET/PUT
4560 request. There is no need to evict the client at all, since
4561 the pinger and/or lock callbacks will handle this, and the
4562 client can restart the bulk request.
4565 Frequency : Always when mmapping file with no objects
4567 Description: client crashes when mmapping file with no objects
4568 Details : Check that we actually have objects in a file before doing any
4569 operations on objects in ll_vm_open, ll_vm_close and
4575 Description: Request leak when working with deleted CWD
4576 Details : Introduce advanced request refcount tracking for requests
4577 referenced from lustre intent.
4579 Severity : Enhancement
4581 Description: Cache open file handles on client.
4582 Details : MDS now will return special lock along with openhandle, if
4583 requested and client is allowed to hold openhandle, even if unused,
4584 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4585 closing files for every read/write openration.
4587 Severity : Enhancement
4589 Description: Cache open negative dentries on client when possible.
4590 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4591 negative dentries on lock revocation.
4596 Description: Remounting a client read-only wasn't possible with a zconf mount
4597 Details : It wasn't possible to remount a client read-only with llmount.
4599 Severity : enhancement
4600 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4601 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4602 MPICH is included to add Lustre-specific ADIO interfaces.
4603 This is based closely on the UFS ADIO layer and only differs
4604 in file creation, in order to allow the OST striping to be set.
4605 This is user-contributed code and not supported by CFS.
4610 Description: extended inode attributes (immutable, append-only) work improperly
4611 when 2.4 and 2.6 kernels are used on client/server or vice versa
4612 Details : Introduce kernel-independent values for these flags.
4614 Severity : enhancement
4617 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4618 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4619 granularity. Now, Fractional MB granularity is be supported,
4620 this is very useful for low memory system.
4622 Severity : enhancement
4624 Description: Getattr by fid
4625 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4626 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4627 have locks and avoids getattr rpc requests for it on every stat.
4630 Frequency : Always, for filesystems larger than 2TB
4632 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4633 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4634 filesystems larger than 2TB to be mounted with mballoc enabled.
4637 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4639 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4640 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4641 filesystems larger than 2TB is mounted then the kernel will
4642 silently corrupt the start of the filesystem. CONFIG_LBD is
4643 enabled for all CFS-supported kernels, but the possibility of
4644 this happening with a modified kernel config exists.
4646 Severity : enhancement
4648 Description: add client O_DIRECT support for 2.6 kernels
4649 Details : It is now possible to do O_DIRECT reads and writes to files
4650 in the Lustre client mountpoint on 2.6 kernel clients.
4652 Severity : enhancement
4654 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4655 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4656 parallel, not waiting for response from every OST before sending
4657 a rpc to the next OST.
4662 Description: setattr vs write race when updating file timestamps
4663 Details : Client processes that update a file timestamp into the past
4664 right after writing to the file (e.g. tar) it is possible that
4665 the updated file modification time can be reset to the current
4666 time due to a race between processing the setattr and write RPC.
4668 Severity : enhancement
4670 Description: Bring 'lfs find' closer in line with regular Linux find.
4671 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4672 -print0 options and obtains all the needed info through the lustre
4675 Severity : enhancement
4677 Description: support up to 1024 configured devices on one node
4678 Details : change obd_dev array from statically allocated to dynamically
4679 allocated structs as they are first used to reduce memory usage
4684 Description: Flush dirty partially truncated pages during truncate
4685 Details : Immediatelly flush partially truncated pages in filter_setattr,
4686 this way we completely avoid having any pages in page cache on OST
4687 and can retire ugly workarounds during writes to flush such pages.
4692 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4693 Details : protect inode from truncation within vfs_unlink() context
4694 just take a reference before calling vfs_unlink() and release it
4695 when parent's i_sem is free.
4700 Description: last_id value checked outside lock on OST caused LASSERT failure
4701 Details : If there were multiple MDS->OST object precreate requests in
4702 flight, it was possible that the OST's last object id was checked
4703 outside a lock and incorrectly tripped an assertion. Move checks
4704 inside locks, and discard old precreate requests.
4707 Frequency : always, if extents are used on OSTs
4709 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4710 Details : OSTs using the extents format would not zero the high 16 bits of
4711 the index physical block number. This is not a problem for any
4712 OST filesystems smaller than 16TB, and no kernels support ext3
4713 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4714 new/modified files) and can be fixed for existing filesystems
4715 with e2fsprogs-1.39-cfs1.
4720 Description: import connection selection may be incorrect if timer wraps
4721 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4722 import connections to be ignored if the 32-bit jiffies counter
4723 wraps. Use a 64-bit jiffies counter.
4726 Frequency : very large clusters immediately after boot
4728 Description: LNET request buffers exhausted under heavy short-term load
4729 Details : If a large number of client requests are generated on a service
4730 that has previously never seen so many requests it is possible
4731 that the request buffer growth cannot keep up with the spike in
4732 demand. Instead of dropping incoming requests, they are held in
4733 the LND until the RPC service can accept more requests.
4736 Frequency : Sometimes during replay
4738 Description: Assertion failure in ll_local_open after replay.
4739 Details : If replay happened on an open request reply before we were able
4740 to set replay handler, reply will become not swabbed tripping the
4741 assertion in ll_local_open. Now we set the handler right after
4742 recognising of open request
4745 Frequency : very rare
4747 Description: kernel reports "badness in vsnprintf"
4748 Details : Reading from the "recovery_status" /proc file in small chunks
4749 may cause a negative length in lprocfs_obd_rd_recovery_status()
4750 call to vsnprintf() (which is otherwise harmless). Exit early
4751 if there is no more space in the output buffer.
4753 Severity : enhancement
4755 Description: clear OBD RPC statistics by writing to them
4756 Details : It is now possible to clear the OBD RPC statistics by writing
4757 to the "stats" file.
4762 Description: Client mtime is not the same on different clients after utimes
4763 Details : In some cases, the client was using the utimes() syscall on
4764 a file cached on another node. The clients now validate the
4765 ctime from the MDS + OSTs to determine which one is right.
4770 Description: Inability to activate failout mode
4771 Details : lconf script incorrectly assumed that in pythong string's numeric
4772 value is used in comparisons.
4775 Frequency : always with multiple stripes per file
4777 Description: Inefficient object allocation for mutli-stripe files
4778 Details : When selecting which OSTs to stripe files over, for files with
4779 a stripe count that divides evenly into the number of OSTs,
4780 the MDS is always picking the same starting OST for each file.
4781 Return the OST selection heuristic to the original design.
4786 Description: mount failures may take full timeout to return an error
4787 Details : Under some heavy load conditions it is possible that a
4788 failed mount can wait for the full obd_timeout interval,
4789 possibly several minutes, before reporting an error.
4790 Instead return an error as soon as the status is known.
4792 Frequency : quota enabled and large files being deleted
4794 Description: releasing more than 4GB of quota at once hangs OST
4795 Details : If a user deletes more than 4GB of files on a single OST it
4796 will cause the OST to spin in an infinite loop. Release
4797 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4802 Description: statfs data retrieved from /proc may be stale or zero
4803 Details : When reading per-device statfs data from /proc, in the
4804 {kbytes,files}_{total,free,avail} files, it may appear
4805 as zero or be out of date.
4808 Frequency : systems with MD RAID1 external journal devices
4810 Description: lconf's call to blkid is confused by RAID1 journal devices
4811 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4812 of returning all block devices that match the journal UUID.
4815 Frequency : always, for aggregate stripe size over 4GB
4817 Description: assertion fails when trying to use 4GB stripe size
4818 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4819 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4822 Frequency : always on ppc64
4824 Description: the first write on an ext3 filesystem with mballoc got stuck
4825 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4826 perform endianness conversion.
4828 ------------------------------------------------------------------------------
4830 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4832 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4833 INTEROPERATE with older versions automatically. Please read the
4834 user documentation before upgrading any part of a live system.
4835 * WARNING: Lustre networking configuration changes are required with
4836 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4839 * Support for kernels:
4840 2.6.9-22.0.2.EL (RHEL 4)
4841 2.6.5-7.244 (SLES 9)
4842 2.6.12.6 vanilla (kernel.org)
4845 Severity : enhancement
4846 Bugzilla : 7981/8208
4847 Description: Introduced Lustre Networking (LNET)
4848 Details : LNET is new networking infrastructure for Lustre, it includes
4849 a reorganized network configuration mode (see the user
4850 documentation for full details) as well as support for routing
4851 between different network fabrics. Lustre Networking Devices
4852 (LNDs) for the supported network fabrics have also been
4853 created for this new infrastructure.
4855 Severity : enhancement
4856 Description: Introduced Access control lists
4857 Details : clients can set ACLs on files and directories in order to have
4858 more fine-grained permissions than the standard Unix UGO+RWX.
4859 The MDS must be started with the "-o acl" mount option.
4861 Severity : enhancement
4862 Description: Introduced filesystem quotas
4863 Details : Administrators may now establish per-user quotas on the
4866 Severity : enhancement
4868 Description: Configuration change for the XT3
4869 The PTLLND is now used to run Lustre over Portals on the XT3
4870 The configure option(s) --with-cray-portals are no longer used.
4871 Rather --with-portals=<path-to-portals-includes> is used to
4872 enable building on the XT3. In addition to enable XT3 specific
4873 features the option --enable-cray-xt3 must be used.
4878 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4879 Details : A race between allocating a new llcd and re-getting the llcd_lock
4880 allowed another thread to grab newly-allocated llcd.
4882 Severity : enhancement
4884 Description: 2.6 OST async journal commit and locking fix to improve performance
4885 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4886 2.6 kernels are now async as they already were in 2.4 kernels so
4887 that they can commit concurrently with the network bulk transfer.
4888 For block-allocated files the filter allocation semaphore is held
4889 to avoid filesystem fragmentation during allocation. BKL lock
4890 removed for 2.6 xattr operations where it is no longer needed.
4895 Description: lconf incorrectly determined whether two IP networks could talk
4896 Details : In some more complicated routing and multiple-network
4897 configurations, lconf will avoid trying to make a network
4898 connection to a disjoint part of the IP space. It was doing the
4899 math incorrectly for one set of cases.
4904 Description: Fix for potential infinite loop processing records in an llog.
4905 Details : If an llog record is corrupted/zeroed, it is possible to loop
4906 forever in llog_process(). Validate the llog record length
4907 and skip the remainder of the block on error.
4910 Frequency : occasional (liblustre only)
4912 Description: liblustre could not open files whose last component is a symlink
4913 Details : sysio_path_walk() would incorrectly pass the open intent to
4914 intermediate path components.
4917 Frequency : rare (liblustre only with non-standard tuning)
4918 Bugzilla : 7201 (7350)
4919 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4920 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4921 after it has been released, only do this for referenced locks
4923 Severity : enhancement
4925 Description: specifying an (invalid) directory default stripe_size of -1
4926 would reset the directory default striping
4927 Details : stripe_size -1 was used internally to signal directory stripe
4928 removal, now use "all default" to signal dir stripe removal
4929 as a directory striping of "all default" is not useful
4932 Frequency : common for large clusters running liblustre clients
4934 Description: doing an ls when liblustre clients are running is slow
4935 Details : sending a glimpse AST to a liblustre client waits for every AST
4936 to time out, as liblustre clients will not respond. Since they
4937 cannot cache data we refresh the OST lock LVB from disk instead.
4939 Severity : enhancement
4941 Description: doing an ls at the same time as file IO can be slow
4942 Details : enqueue and other "small" requests can be blocked behind many
4943 large IO requests. Create a new OST IO portal for non-IO
4944 requests so they can be processed faster.
4947 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4949 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4950 created file with mode 000
4951 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4952 allow an ftruncate() as a non-root user to a file with mode 000.
4953 Lustre now respects this flag to disable mode checks when
4954 truncating a file owned by the user
4957 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4960 Description: Revoking locks from clients that went dead or catatonic might take
4962 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4963 cancellation of such locks instant on servers without waiting for
4964 any reply from clients. Clients drops these locks when cancel
4965 notification from server is received without replying.
4968 Frequency : liblustre-only, when liblustre client dies or becomes busy
4970 Description: Doing ls on Linux clients can take a long time with active
4972 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4973 granting such locks to it in the first place if possible. Locks
4974 are taken by proxy on the OST during the read or write and
4975 dropped immediately afterward. Add connect flags handling, do
4976 not grant locks to liblustre clients for glimpse ASTs.
4978 Severity : enhancement
4980 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4981 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4982 sequential reads and behaves badly with applications doing random
4983 reads. Improve it by reading ahead at least read region, and
4984 avoiding excessive large RPC for small reads.
4986 Severity : enhancement
4988 Description: Creating more than 1000 files for a single job may cause a load
4989 imbalance on the OSTs if there are also a large number of OSTs.
4990 Details : qos_prep_create() uses an OST index reseed value that is an
4991 even multiple of the number of available OSTs so that if the
4992 reseed happens in the middle of the object allocation it will
4993 still utilize the OSTs as uniformly as possible.
4998 Description: OST or MDS may oops in ping_evictor_main()
4999 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5000 but doesn't restart at beginning of obd_exports_timed list
5003 Severity : enhancement
5005 Description: improve by-nid export eviction on the MDS and OST
5006 Details : allow multiple exports with the same NID to be evicted at one
5007 time without re-searching the exports list.
5010 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5012 Description: MDS may oops in groups_free()
5013 Details : in rare race conditions a newly allocated group_info struct is
5014 freed again, and this can be NULL. The 2.4 compatibility code
5015 for groups_free() checked for a NULL pointer, but 2.6 did not.
5018 Frequency : common for liblustre clients doing little filesystem IO
5019 Bugzilla : 9352, 7313
5020 Description: server may evict liblustre clients accessing contended locks
5021 Details : if a client is granted a lock or receives a completion AST
5022 with a blocking AST already set it would not reply to the AST
5023 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5026 Frequency : lfs setstripe, only systems with more than 160 OSTs
5028 Description: unable to set striping with a starting offset beyond OST 160
5029 Details : llapi_create_file() incorrectly limited the starting stripe
5030 index to the maximum single-file stripe count.
5033 Frequency : LDAP users only
5035 Description: lconf did not handle in-kernel recovery with LDAP properly
5036 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5038 Severity : enhancement
5040 Description: bind OST threads to NUMA nodes to improve performance
5041 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5042 node and do their allocations there to localize memory access
5044 Severity : enhancement
5046 Description: llmount can determine client NID directly from Myrinet (GM)
5047 Details : the client NID code from gmnalnid was moved directly into
5048 llmount, removing the need to use this or specifying the
5049 client NID explicitly when mounting GM clients with zeroconf
5052 Frequency : if client is started with down MDS
5054 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5055 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5056 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5057 ptlrpc_import_delayed_req() because we didn't check if the
5058 request was interrupted, and we also didn't break out of the
5059 event loop if there was no timeout
5064 Description: data loss during non-page-aligned writes to a single file from
5065 both multiple nodes and multiple threads on one node at same time
5066 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5067 inconsistency led to false short-reads, that were cached and later
5068 used by ->prepare_write() to fill in partially written page,
5069 leading to data loss.
5072 Frequency : always, if lconf --abort_recovery used
5074 Description: lconf --abort_recovery fails with 'Operation not supported'
5075 Details : lconf was attempting to abort recovery on the MDT device and not
5078 Severity : enhancement
5080 Description: remove cleanup logs
5081 Details : replace lconf-generated cleanup logs with lustre internal
5082 cleanup routines. Eliminates the need for client-cleanup and
5085 Severity : enhancement
5087 Description: add support for EAs (user and system) on lustre filesystems
5088 Details : it is now possible to store extended attributes in the Lustre
5089 client filesystem, and with the user_xattr mount option it
5090 is possible to allow users to store EAs on their files also
5092 Severity : enhancement
5094 Description: Add possibility (config option) to show minimal available OST free
5096 Details : When compiled with --enable-mindf configure option, statfs(2)
5097 (and so, df) will return least minimal free space available from
5098 all OSTs as amount of free space on FS, instead of summary of
5099 free spaces of all OSTs.
5101 Severity : enhancement
5103 Description: do not expand extent locks acquired on OST-side
5104 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5105 by server: they are not cached anyway.
5108 Frequency : when mmap is used/binaries executed from Lustre
5110 Description: Unmmap pages before throwing them away from read cache.
5111 Details : llap_shrink cache now attempts to unmap pages before discarding
5112 them (if unmapping failed - do not discard). SLES9 kernel has
5113 extra checks that trigger if this unmapping is not done first.
5118 Description: lconf didn't resolve symlinks before checking to see whether a
5119 given mountpoint was already in use
5122 Frequency : when migrating failover services
5123 Bugzilla : 6395, 9514
5124 Description: When migrating a subset of services from a node (e.g. failback
5125 from a failover service node) the remaining services would
5126 time out and evict clients.
5127 Details : lconf --force (implied by --failover) sets the global obd_timeout
5128 to 5 seconds in order to quickly disconnect, but this caused
5129 other RPCs to time out too quickly. Do not change the global
5130 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5132 Severity : enhancement
5133 Frequency : if MDS is started with down OST
5134 Bugzilla : 9439,5706
5135 Description: Allow startup/shutdown of an MDS without depending on the
5136 availability of the OSTs.
5137 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5138 Add appropriate locking and lov-osc refcounts for safe
5139 cleaning. Add osc abort_inflight calls in case the
5140 synchronize never started.
5143 Frequency : occasional (Cray XT3 only)
5145 Description: root not authorized to access files in CRAY_PORTALS environment
5146 Details : The client process capabilities were not honoured on the MDS in
5147 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5148 been accessed by an authorized user then root was able to access
5149 the file on the local client also. The root user capabilities
5150 are now allowed on the MDS, as this environment has secure UID.
5153 Frequency : occasional
5155 Description: ldiskfs "too long searching" message happens too often
5156 Details : A debugging message (otherwise harmless) prints too often on
5157 the OST console. This has been reduced to only happen when
5158 there are fragmentation problems on the filesystem.
5163 Description: Division by zero in statfs when all OSCs are inactive
5164 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5165 lov_statfs divides by value returned by lov_get_stripecnt().
5169 Bugzilla : 9489, 3273
5170 Description: First write from each client to each OST was only 4kB in size,
5171 to initialize client writeback cache, which caused sub-optimal
5172 RPCs and poor layout on disk for the first written file.
5173 Details : Clients now request an initial cache grant at (re)connect time
5174 and so that they can start streaming writes to the cache right
5175 away and always do full-sized RPCs if there is enough data.
5176 If the OST is rebooted the client also re-establishes its grant
5177 so that client cached writes will be honoured under the grant.
5182 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5183 Details : Now I/O RPCs go to different portal number and (presumably) fast
5184 lock requests (and glimses) and other RPCs get their own service
5185 threads pool that should be able to service those RPCs
5188 Severity : enhancement
5190 Description: Ability to exchange lustre version between client and servers and
5191 issue warnings at client side if client is too old. Also for
5192 liblustre clients there is ability to refuse connection of too old
5194 Details : New 'version' field is added to connect data structure that is
5195 filled with version info. That info is later checked by server and
5199 Frequency : rare, liblustre only.
5200 Bugzilla : 9296, 9581
5201 Description: Two simultaneous writes from liblustre at offset within same page
5202 might proceed at the same time overwriting eachother with stale
5204 Details : I/O lock withing llu_file_prwv was released too early, before data
5205 actually was hitting the wire. Extended lock-holding time until
5206 server acknowledges receiving data.
5209 Frequency : extremely rare. Never observed in practice.
5211 Description: avoid generating lustre_handle cookie of 0.
5212 Details : class_handle_hash() generates handle cookies by incrementing
5213 global counter, and can hit 0 occasionaly (this is unlikely, but
5214 not impossible, because initial value of cookie counter is
5215 selected randonly). Value of 0 is used as a sentinel meaning
5216 "unassigned handle" --- avoid it. Also coalesce two critical
5217 sections in this function into one.
5219 Severity : enhancement
5221 Description: allow liblustre clients to delegate truncate locking to OST
5222 Details : To avoid overhead of locking, liblustre client instructs OST to
5223 take extent lock in ost_punch() on client's behalf. New connection
5224 flag is added to handle backward compatibility.
5226 Severity : enhancement
5227 Bugzilla : 4928, 7341, 9758
5228 Description: allow number of OST service threads to be specified
5229 Details : a module parameter allows the number of OST service threads
5230 to be specified via "options ost ost_num_threads={N}" in the
5231 OSS's /etc/modules.conf or /etc/modprobe.conf.
5235 Bugzilla : 6146, 9635, 9895
5236 Description: servers crash with bad pointer in target_handle_connect()
5237 Details : In rare cases when a client is reconnecting it was possible that
5238 the connection request was the last reference for that export.
5239 We would temporarily drop the export reference and get a new
5240 one, but this may have been the last reference and the export
5241 was just destroyed. Get new reference before dropping old one.
5243 Severity : enhancement
5244 Frequency : if client is started with failover MDS
5246 Description: Allow multiple MDS hostnames in the mount command
5247 Details : Try to read the configuration from all specified MDS
5248 hostnames during a client mount in case the "primary"
5251 Severity : enhancement
5253 Description: Stop sending data to evicted clients as soon as possible.
5254 Details : Check if the client we are about to send or are sending data to
5255 was evicted already. (Check is done every second of waiting,
5256 for which l_wait_event interface was extended to allow checking
5257 of exit condition at specified intervals).
5260 Frequency : rare, normally only when NFS exporting is done from client
5262 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5264 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5267 Frequency : always, if srand() is called before liblustre initialization
5269 Description: Liblustre uses system PRNG disturbing its usage by user application
5270 Details : Introduce internal to lustre fast and high-quality PRNG for
5271 lustre usage and make liblustre and some other places in generic
5272 lustre code to use it.
5274 Severity : enhancement
5275 Bugzilla : 9477, 9557, 9870
5276 Description: Verify that the MDS configuration logs are updated when xml is
5277 Details : Check if the .xml configuration logs are newer than the config
5278 logs stored on the MDS and report an error if this is the case.
5279 Request --write-conf, or allow starting with --old_conf.
5281 Severity : enhancement
5283 Description: Handle symlinks in the path when checking if Lustre is mounted.
5284 Details : Resolve intermediate symlinks when checking if a client has
5285 mounted a filesystem to avoid duplicate client mounts.
5290 Description: lconf can hit an error exception but still return success.
5291 Details : The lconf command catches the Command error exception at the top
5292 level script context and will exit with the associated exit
5293 status, but doesn't ensure that this exit status is non-zero.
5298 Description: failure of ptlrpc thread startup can cause oops
5299 Details : Starting a ptlrpc service thread can fail if there are a large
5300 number of threads or the server memory is very fragmented.
5301 Handle this without oopsing.
5304 Frequency : always, only if liblustre and non-default acceptor port was used
5306 Description: liblustre cannot connect to servers with non-default acceptor port
5307 Details : tcpnal_set_default_params() was not called and was therefore
5308 ignoring the environment varaible TCPNAL_PORT, as well as other
5309 TCPNAL_ environment variables
5314 Description: two objects could be created on the same OST for a single file
5315 Details : If an OST is down, in some cases it was possible to create two
5316 objects on a single OST for a single file. No problems other
5317 than potential performance impact and spurious error messages.
5321 Bugzilla : 5681, 9562
5322 Description: Client may oops in ll_unhash_aliases
5323 Details : Client dcache may become inconsistent in race condition.
5324 In some cases "getcwd" can fail if the current directory is
5330 Description: Inode refcounting problems in NFS export code
5331 Details : link_raw functions used to call d_instantiate without obtaining
5332 extra inode reference first.
5336 Bugzilla : 9942, 9903
5337 Description: Referencing freed requests leading to crash, memleaks with NFS.
5338 Details : We used to require that call to ll_revalidate_it was always
5339 followed by ll_lookup_it. Also with revalidate_special() it is
5340 possible to call ll_revalidate_it() twice for the same dentry
5341 even if first occurence returned success. This fix changes semantic
5342 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5343 reference on a request referred from the intent.
5344 ll_intent_release() then releases such a request.
5347 Frequency : rare, normally benchmark loads only
5349 Description: unlinked inodes were kept in memory on the client
5350 Details : If a client is repeatedly creating and unlinking files it
5351 can accumulate a lot of stale inodes in the inode slab cache.
5352 If there is no other client load running this can cause the
5353 client node to run out of memory. Instead flush old inodes
5354 from client cache that have the same inode number as a new inode.
5357 Frequency : SLES9 2.6.5 kernel and long filenames only
5358 Bugzilla : 9969, 10379
5359 Description: utime reports stale NFS file handle
5360 Details : SLES9 uses out-of-dentry names in some cases, which confused
5361 the lustre dentry revalidation. Change it to always use the
5365 Frequency : rare, unless heavy write-truncate concurrency is continuous
5366 Bugzilla : 4180, 6984, 7171, 9963, 9331
5367 Description: OST becomes very slow and/or deadlocked during object unlink
5368 Details : filter_destroy() was holding onto the parent directory lock
5369 while truncating+unlinking objects. For very large objects this
5370 may block other threads for a long time and slow overall OST
5371 responsiveness. It may also be possible to get a lock ordering
5372 deadlock in this case, or run out of journal credits because of
5373 the combined truncate+unlink. Solution is to do object truncate
5374 first in one transaction without parent lock, and then do the
5375 final unlink in a new transaction with the parent lock. This
5376 reduces the lock hold time dramatically.
5379 Frequency : rare, 2.4 kernels only
5381 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5382 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5383 before unmounting their filesystem in order to avoid deadlock.
5384 The kernel_locked() function in 2.4 kernels only checks whether
5385 the kernel lock is held, not whether it is this process that is
5386 holding it as 2.6 kernels do.
5391 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5392 Details : The client ptlrpc code may be trying to reconnect to a down
5393 server before a previous connection attempt has timed out.
5394 Increase the reconnect interval to be longer than the connection
5395 timeout interval to avoid sending duplicate connections to
5399 Frequency : echo_client brw_test command
5401 Description: fix echo_client to work with OST preallocated code
5402 Details : OST preallocation code (5137) didn't take echo_client IO path
5403 into account: echo_client calls filter methods outside of any
5404 OST thread and, hence, there is no per-thread preallocated
5405 pages and buffers to use. Solution: hijack pga pages for IO. As
5406 a byproduct, this avoids unnecessary data copying.
5410 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5411 Description: Client can oops in mdc_commit_close() after open replay
5412 Details : It was possible for the MDS to return an open request with no
5413 transaction number in mds_finish_transno() if the client was
5414 evicted, but without actually returning an error. Clients
5415 would later try to replay that open and may trip an assertion
5416 Simplify the client close codepath, and always return an error
5417 from the MDS in case the open is not successful.
5420 Frequency : rare, 2.6 OSTs only
5422 Description: OST may deadlock under high load on fragmented files
5423 Details : If there was a heavy load and highly-fragmented OST filesystems
5424 it was possible to have all the OST threads deadlock waiting on
5425 allocation of biovecs, because the biovecs were not released
5426 until the entire RPC IO was completed. Instead, release biovecs
5427 as soon as they are complete to ensure forward IO progress.
5429 Severity : enhancement
5431 Description: Support for specifying external journal device at mount
5432 Details : If an OST or MDS device is formatted with an external journal
5433 device, this device major/minor is stored in the ext3 superblock
5434 and may not be valid for failover. Allow detecting and
5435 specifying the external journal at mount time.
5440 Description: Mounting an MDS with pending unlinked files may cause oops
5441 Details : target_finish_recovery() calls mds_postrecov() which returned
5442 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5443 considers this an error and immediately begins cleaning up the
5444 lov, just after starting the mds_lov process
5446 Severity : enhancement
5448 Description: Implement 'lfs df' to report actual free space on per-OST basis
5449 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5450 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5451 usage of objects; '-h' to report in human readable format.
5453 ------------------------------------------------------------------------------
5455 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5462 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5463 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5464 at startup during orphan unlinking. Instead, skip buddy update
5465 before setup, as it will be regenerated after recovery is complete.
5470 Description: If an OST is inactive, its locks might reference stale inodes.
5471 Details : lov_change_cbdata() must iterate over all namespaces, even if
5472 they are inactive to clear inode references from the lock.
5474 Severity : enhancement
5475 Frequency : occasional, if non-standard max_dirty_mb used
5477 Description: Client will block write RPCs if not enough grant
5478 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5479 then the client will block writes while waiting for another RPC
5480 to complete instead of consuming its dirty limit. With change
5481 we get improved performance when max_dirty_mb is small.
5483 Severity : enhancement
5484 Bugzilla : 3389, 6253
5485 Description: Add support for supplementary groups on the MDS.
5486 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5487 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5488 lookups for user supplementary groups into a cache.
5492 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5493 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5494 for liblustre O_EXCL is enforced by MDS.
5498 Description: readdir never returns NULL in liblustre.
5499 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5500 dentry in struct dirent.
5504 Frequency : liblustre only, depends on application IO pattern
5505 Description: liblustre clients evicted if not contacting servers
5506 Details : Don't put liblustre clients into the ping_evictor list, so
5507 they will not be evicted by the pinger ever.
5509 Severity : enhancement
5511 Description: Add ability to evict clients by NID from MDS.
5512 Details : By echoing "nid:$NID" string into
5513 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5514 $NID would be instantly evicted from this MDS and from all active
5515 OSTs connected to it.
5519 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5520 Details : lookup_it_finish() used to query file size from OSTs that was not
5525 Description: service threads change working directory to that of init
5526 Details : Starting lustre service threads may pin the working directory
5527 of the parent thread, making that filesystem busy. Threads
5528 now change to the working directory of init to avoid this.
5532 Frequency : during shutdown only
5533 Description: shutdown with a failed MDS or OST can cause unmount to hang
5534 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5539 Frequency : 2.6 only
5540 Description: chmod/chown may include an extra supplementary group
5541 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5542 supplementary group and if none is specified this is used.
5545 Bugzilla : 5479 (6816)
5547 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5548 Details : If lookup is in progress on a file that is unlinked we might try
5549 to revalidate the inode and fail in revalidate after lookup is
5550 complete and ll_file_open() enqueues the open again but
5551 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5554 Frequency : always, if lconf --abort_recovery used
5556 Description: lconf --abort_recovery fails with 'Operation not supported'
5557 Details : lconf was attempting to abort recovery on the MDT device and not
5560 ------------------------------------------------------------------------------
5562 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5567 Frequency : rare (only unsupported configurations with a node running as an
5569 Bugzilla : 6514, 5137
5570 Description: Mounting a Lustre file system on a node running as an OST could
5572 Details : OSTs now preallocates memory needed to write out data at
5573 startup, instead of when needed, to avoid having to
5574 allocate memory in possibly low memory situations.
5575 Specifically, if the file system is mounted on on OST,
5576 memory pressure could force it to try to write out data,
5577 which it needed to allocate memory to do. Due to the low
5578 memory, it would be unable to do so and the node would
5579 become unresponsive.
5581 Severity : enhancement
5583 Description: Addition of lconf --service command line option
5584 Details : lconf now accepts a '--service <arg>' option, which is
5585 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5587 Severity : enhancement
5589 Description: Failover mode is now the default for OSTs.
5590 Details : By default, OSTs will now run in failover mode. To return to
5591 the old behaviour, add '--failout' to the lmc line for OSTs.
5593 Severity : enhancement
5595 Description: Health checks are now provided for MDS and OSTs
5596 Details : Additional detailed health check information on MSD and OSTs
5597 is now provided through the procfs health_check value.
5600 Frequency : occasional, depends on IO load
5602 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5603 after numerous create/delete cycles
5604 Details : The ext3 inode allocation policy would not allocate new inodes
5605 very well on the OSTs because there are no new directories
5606 being created. Instead we look for groups with free space if
5607 the parent directories are nearly full.
5612 Description: Network or server problems during mount may cause partially
5613 mounted clients instead of returning an error.
5614 Details : The config llog parsing code may overwrite the error return
5615 code during mount error handling, returning success instead
5621 Description: MDS can fail to allocate large reply buffers
5622 Details : After long uptimes the MDS can fail to allocate large reply
5623 buffers (e.g. zconf client mount config records) due to memory
5624 fragmentation or consumption by the buffer cache. Preallocate
5625 some large reply buffers so that these replies can be sent even
5626 under memory pressure.
5630 Frequency : rare (liblustre)
5631 Description: fsx running with liblustre complained that using truncate() to
5632 extend the file doesn't work. This patch corrects that issue.
5633 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5634 ATTR_SIZE and lsm use in llu_setattr_raw.
5638 Frequency : rare, only 2.6 kernels
5639 Description: Unusual file access patterns on the MDS may result in inode
5640 data being lost in very rare circumstances.
5641 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5642 code in the 2.6 kernel caused the inode and/or EA data not to
5643 be read from disk, causing single-file corruption.
5647 Frequency : rare, only 2.6 filesystems using extents
5648 Description: Heavy concurrent write and delete load may cause data corruption.
5649 Details : It was possible under high-load situations to have an extent
5650 metadata block in the block device cache from a just-unlinked
5651 file overwrite a newly-allocated data block. We now unmap any
5652 metadata buffers that alias just-allocated data blocks.
5656 Frequency : filesystems with default stripe_count larger than 77
5657 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5658 Details : lconf specifies an inode size of 4096 bytes when the default
5659 stripe_count is larger than 77. This conflicts with the default
5660 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5661 this case to avoid pinning too much memory for large EAs.
5663 ------------------------------------------------------------------------------
5665 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5670 Frequency : rare (extremely heavy IO load with hundreds of clients)
5672 Description: Client is evicted, gets IO error writing to file
5673 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5674 caused clients to be evicted for AST timeouts. The fixes in
5675 bug 5192 mean we no longer need to have such short AST timeouts
5676 so ldlm_timeout has been increased.
5679 Frequency : occasional during --force or --failover shutdown under load
5680 Bugzilla : 5949, 4834
5681 Description: Server oops/LBUG if stopped with --force or --failover under load
5682 Details : a collection of import/export refcount and cleanup ordering
5683 issues fixed for safer force cleanup
5686 Frequency : only filesystems larger than 120 OSTs
5687 Bugzilla : 5990, 6223
5688 Description: lfs getstripe would oops on a very large filesystem
5689 Details : lov_getconfig used kfree on vmalloc'd memory
5692 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5693 Bugzilla : 6242, 6243
5694 Description: reading from files that had been truncated to a non-zero size
5695 but never opened returned no data
5696 Details : ll_file_read() reads zeros from no-object files to EOF
5701 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5702 Details : There's little harm in aborting MDS/OSS recovery and letting it
5703 try again, so I removed the LASSERT and return an error instead.
5705 Severity : enhancement
5707 Description: New debugging infrastructure for tracking down data corruption
5708 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5709 (b) cover more of the client-side code path, and (c) try to narrow
5710 down where problems occurred
5714 Bugzilla : 3819, 4364, 4397, 6313
5715 Description: Racing close and eviction MDS could cause assertion in mds_close
5716 Details : It was possible to get multiple mfd references during close and
5717 client eviction, leading to one thread referencing a freed mfd.
5719 Severity: : enhancement
5720 Bugzilla : 3262, 6359
5721 Description: Attempts to reconnect to servers are now more aggressive.
5722 Details : This builds on the enhanced upcall-less recovery that was added
5723 in 1.4.2. When trying to reconnect to servers, clients will
5724 now try each server in the failover group every 10 seconds. By
5725 default, clients would previously try one server every 25 seconds.
5730 Description: After recovery, certain operations trigger a failed
5731 assertion on a client.
5732 Details : Failing over an mds, using lconf -d --failover, while a
5733 client was doing a readdir() call would cause the client to
5734 LBUG after recovery completed and the readdir() was resent.
5736 Severity : enhancement
5738 Description: Default groups are now added by lconf
5739 Details : You can now run lconf --group <servicename> without having to
5740 manually add groups with lmc.
5743 Frequency : occasional
5745 Description: Nodes with an elan id of 0 trigger a failed assertion
5748 Frequency : always when accessing e.g. tty/console device nodes
5750 Description: tty and some other devices nodes cannot be used on lustre
5751 Details : file's private_data field is used by device data and lustre
5752 values in there got lost. New field was added to struct file to
5753 store fs-specific private data.
5756 Frequency : when exporting Lustre via NFS
5758 Description: NFSD failed occasionally when looking up a path component
5759 Details : NFSD is looking up ".." which was broken in ext3 directories
5760 that had grown large enough to become hashed.
5763 Frequency : Clusters with multiple interfaces not on the same subnet
5765 Description: Nodes will repeatedly try to reconnect to an interface which it
5766 cannot reach and report an error to the log.
5767 Details : Extra peer list entries will be created by lconf with some peers
5768 unreachable. lconf now validates the peer before adding it.
5771 Frequency : Only if a default stripe is set on the filesystem root.
5773 Description: Setting a default stripe on the filesystem root prevented the
5774 filesystem from being remounted.
5775 Details : The client was sending extra request flags in the root getattr
5776 request and did not allocate a reply buffer for the dir EA.
5779 Frequency : occasional, higher if lots of files are accessed by one client
5780 Bugzilla : 6159, 6097
5781 Description: Client trips assertion regarding lsm mismatch/magic
5782 Details : While revalidating inodes the VFS looks up inodes with ifind()
5783 and in rare cases can find an inode that is being freed.
5784 The ll_test_inode() code will free the lsm during ifind()
5785 when it finds an existing inode and then the VFS later attaches
5786 this free lsm to a new inode.
5790 Bugzilla : 6422, 7030
5791 Description: MDS deadlock between mkdir and client eviction
5792 Details : Creating a new file via mkdir or mknod (starting a transaction
5793 and getting the ns lock) can deadlock with client eviction
5794 (gets ns lock and trying to finish a synchronous transaction).
5797 Frequency : occasional
5798 Description: While starting a server, the fsfilt_ext3 module could not be
5800 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5801 kernels. Previously, lconf would still use the ext3 name
5802 when trying to load modules. Now, it will correctly use
5803 ext3 on 2.4 and ldiskfs on 2.6.
5805 Severity : enhancement
5806 Description: The default stripe count has been changed to 1
5807 Details : The interpretation of the default stripe count (0, to lfs
5808 or lmc) has been changed to mean striping across a single
5809 OST, rather than all available. For general usage we have
5810 found a stripe count of 1 or 2 works best.
5812 Severity : enhancement
5813 Description: Add support for compiling against Cray portals.
5814 Details : Conditional compiling for some areas that are different
5818 Frequency : occasional
5819 Bugzilla : 6409, 6834
5820 Description: Creating files with an explicit stripe count may lead to
5821 a failed assertion on the MDS
5822 Details : If some OSTs are full or unavailable, creating files may
5823 trigger a failed assertion on the MDS. Now, Lustre will
5824 try to use other servers or return an error to the
5828 Frequency : occasional
5830 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5831 caused lock timeout during readahead (since 1.4.2).
5832 Details : Processes doing readahead might match a lock that hasn't been
5833 granted yet if there are overlapping and conflicting lock
5834 requests. The readahead process waits on ungranted lock
5835 (original lock is CBPENDING), while OST waits for that process
5836 to cancel CBPENDING read lock and eventually evicts client.
5838 Severity : enhancement
5840 Description: Initial enabling of flock support for clients
5841 Details : Implements fcntl advisory locking and file status functions.
5842 This feature is provided as an optional mount flag (default
5843 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5844 locking are implemented yet, and those that are are not guaranteed
5845 to be completely correct in production environments.
5846 mount -t lustre -o [flock|noflock] ...
5849 Frequency : occasional
5851 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5852 assertion in the ext3 JBD (journaling) layer.
5853 Details : The b_committed_data struct is protected by the big kernel lock
5854 in 2.4 kernels, serializing journal_commit_transaction() and
5855 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5856 access to this struct. In 2.6 kernels there is finer grained
5857 locking to improve SMP performance of the JBD layer.
5861 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5866 Description: merge and cleanup kernel patches.
5869 -----------------------------------------------------------------------------
5871 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5873 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5874 versions of Lustre. Please update all servers and clients to
5875 version 1.4.2 or later at the same time. You must also run
5876 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5877 to update the configuration logs.
5879 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5880 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5881 - swab llog records of type '0' so we get proper header size/idx (5861)
5882 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5883 - fix rename of one directory over another leaking an inode (5953)
5884 - avoid SetPageDirty on 2.6 (5981)
5885 - don't re-add just-being-destroyed locks to the waiting list (5653)
5886 - when creating new directories, inherit the parent's custom
5887 striping settings if present parent (3048)
5888 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5889 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5890 - don't hold client locks on temporary worklist from l_lru (5666)
5891 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5892 - automatically evict dead clients (5921)
5893 - Update file size properly in create+truncate+fstat case (6196)
5894 - Do not unhash mountpoint dentries, do not allow removal of
5896 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5897 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5898 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5899 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5900 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5901 - Correctly handle reads of files with no objects (6243)
5902 - lctl recover will also mark a device active if deactivate used (5933)
5904 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5905 - allow --write-conf on an MDS with different nettype than client (5619)
5906 - don't write config llogs to MDS for mounts not from that MDS (5617)
5907 - lconf should create multiple TCP connections from a client (5201)
5908 - init scripts are now turned off by default; run chkconfig --on
5909 lustre and chkconfig --on lustrefs to use them
5910 - upcalls are no longer needed for clients to recover to failover
5912 - add --abort-recovery option to lconf to abort recovery on device
5914 - add support for an arbitrary number of OSTs (3026)
5915 - Quota support protocol changes.
5916 - forward compatibility changes to wire structs (6007)
5917 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5918 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5919 - improved "lustre status" script
5920 - initialize blocksize for non-regular files (6062)
5921 - added --disable-server and --disable-client configure options (5782)
5922 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5923 - Vanilla 2.4.29 support
5924 - increase maximum number of obd devices to 520 (6242)
5925 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5926 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5927 SLES 9 2.6.5 kernels
5928 - limit stripes per file to 160 (the maximum EA size) (6093)
5930 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5933 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5934 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5935 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5936 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5937 - small ext3 extents cleanups and fixes (5733)
5938 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5939 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5940 - avoid needless client->OST connect, fix handle mismatch (5317)
5941 - fix DLM error path that led to out-of-sync client, long delays (5779)
5942 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5943 - fix several locking issues related to i_size (5492,5624,5654,5672)
5944 - don't move pending lock onto export if it is already evicted (5683)
5945 - fix kernel oops when creating .foo in unlinked directory (5548)
5946 - fix deadlock in obdfilter statistics vs. object create (5811)
5947 - use time_{before,after} to avoid timer jiffies wrap (5882)
5948 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5949 - Do not leak request if server was not able to process it (5154)
5950 - If mds_open unable to find parent dir, make that negative lookup(5154)
5951 - don't create new directories with extent-mapping (5909, 5936)
5953 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5954 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5955 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5956 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5957 - support for RHEL4 kernel on i686 (5773)
5958 - provide error messages when incompatible logs are encountered (5898)
5960 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5961 * version 1.4.0.10 (1.4.1 release candidate 1)
5963 - don't keep a lock reference when lock is not granted (4238)
5964 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5965 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5966 - drop import inflight refcount on signal_completed_replay error (5255)
5967 - unlock page after async write error during send (3677)
5968 - handle missing objects in filter_preprw_read properly (5265)
5969 - no transno return for symlink open, don't save no-trasno open (3440)
5970 - don't try to complete elan receive that already failed (4012)
5971 - free RPC server reply state on error (5406)
5972 - clean up thread from ptlrpc_start_thread() on error (5160)
5973 - readahead could read extra page into cache that wasn't ejected (5388)
5974 - prevent races in class_attach/setup/cleanup/detach (5260)
5975 - don't dereference de->d_inode after l_dput of de (5458)
5976 - use "int" for stripe value returned from lock_to_stripe (5544)
5977 - mballoc allocation and error-checking fixes in 2.6 (5504)
5978 - block device patches to fix I/O request sizes in 2.6 (5482)
5979 - look up hostnames for IB nals (5602)
5980 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5981 - don't start multiple acceptors for the same port (5277)
5982 - fix incorrect LASSERT in mds_getattr_name (5635)
5983 - export a proc file for general "ping" checking (5628)
5984 - fix "lfs check" to not block when the MDS is down (5628)
5986 - service request history (4965)
5987 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5988 - create an "evict_client" /proc entry on OSTs, like the MDS has
5989 - fix mount usage message, return errors per mount(8) (5168)
5990 - change grep [] to grep "[]" in tests so they work in more UMLs
5991 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5992 - remove ancient LOV_MAGIC_V0 EA support (5047)
5993 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5994 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5995 - allow readahead allocations to fail when low on memory (5383)
5996 - mmap locking landed again, after considerable improvement (2828)
5997 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5999 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6002 - send OST transaction number in read/write reply to free req (4966)
6003 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6004 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6005 - the watchdog thread now runs as interruptible (5246)
6006 - flock/lockf fixes (but it's still disabled, pending 5135)
6007 - don't use EXT3 constants in llite code (5094)
6008 - memory shortage at startup could cause assertion (5176)
6010 - reorganization of lov code
6011 - single portals codebase
6013 - add extents/mballoc support (5025)
6014 - direct I/O reads in the obdfilter (4048)
6015 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6017 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6020 - send OST transaction number in read/write reply to free req (4966)
6021 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6022 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6023 - the watchdog thread now runs as interruptible (5246)
6024 - handle missing objects in filter_preprw_read properly (5265)
6025 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6026 - drop import inflight refcount on signal_completed_replay error (5255)
6027 - unlock page after async write error during send (3677)
6028 - return original error code on reconstructed replies (3761)
6029 - no transno return for symlink open, don't save no-trasno open (3440)
6031 - add pid to ldlm debugging output (4922)
6032 - bump the watchdog timeouts -- we can't handle 30sec yet
6033 - extra debugging for orphan dentry/inode bug (5259)
6035 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6038 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6039 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6040 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6041 - fix NULL dereference in /proc/sys/portals/routes (4827)
6042 - allow failed mdc_close() operations to be interrupted (4561)
6043 - stop precreate on OST before MDS would time out on it (4778)
6044 - don't send partial-page writes before EOF from client (4410)
6045 - discard client grant for sub-page writes on large-page clients (4520)
6046 - don't free dentries not owned by NFS code, check generation (4806)
6047 - fix lsm leak if mds_create_objects() fails (4801)
6048 - limit debug_daemon file size, always print CERROR messages (4789)
6049 - use transno after validating reply (3892)
6050 - process timed out requests if import state changes (3754)
6051 - update mtime on OST during writes, return in glimpse (4829)
6052 - add mkfsoptions to LDAP (4679)
6053 - use ->max_readahead method instead of zapping global ra (5039)
6054 - don't interrupt __l_wait_event() during strace
6056 - add software watchdogs to catch hung threads quickly (4941)
6057 - make lustrefs init script start after nfs is mounted
6058 - fix CWARN/ERROR duplication (4930)
6059 - return async write errors to application if possible (2248)
6060 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6061 - print NAL number in %x format (4645)
6062 - update barely-supported suse-2.4.21-171 series (4842)
6063 - support for sles 9 %post scripts
6064 - support for building 2.6 kernel-source packages
6065 - support for sles km_* packages
6067 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6070 - ignore -ENOENT errors in osc_destroy (3639)
6071 - notify osc create thread that OSC is being cleaned up (4600)
6072 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6073 - reconstruct ost_handle() like mds_handle() (4657)
6074 - create a new thread to do import eviction to avoid deadlock (3969)
6075 - let lconf resolve symlinked-to devices (4629)
6076 - don't unlink "objects" from directory with default EA (4554)
6077 - hold socknal file ref over connect in case target is down (4394)
6078 - allow more than 32000 subdirectories in a single directory (3244)
6079 - fix blocks count for O_DIRECT writes (3751)
6080 - OST returns ENOSPC from object create when no space left (4539)
6081 - don't send truncate RPC if file size isn't changing (4410)
6082 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6083 - bind to privileged port in socknal and tcpnal (3689)
6085 - rate limit CERROR/CWARN console message to avoid overload (4519)
6086 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6087 - basic mmap support (3918)
6088 - kernel patch series update from b1_4 (4711)
6090 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6093 - avoid crash during MDS cleanup with OST shut down (2775)
6094 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6095 - don't use bad inodes on the MDS (3744)
6096 - dynamic object preallocation to improve recovery speed (4236)
6097 - don't hold spinlock over lock dumping or change debug flags (4401)
6098 - don't zero obd_dev when it is force cleaned (3651)
6099 - print grants to console if they go negative (4431)
6100 - "lctl deactivate" will stop automatic recovery attempts (3406)
6101 - look for existing locks in ldlm_handle_enqueue() (3764)
6102 - don't resolve lock handle twice in recovery avoiding race (4401)
6103 - revalidate should check working dir is a directory (4134)
6105 - don't always mark "slow" obdfilter messages as errors (4418)
6107 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6110 - don't close LustreDB during write_conf until it is done (3860)
6111 - fix typo in lconf for_each_profile (3821)
6112 - allow dumping logs from multiple threads at one time (3820)
6113 - don't allow multiple threads in OSC recovery (3812)
6114 - fix debug_size parameters (3864)
6115 - fix mds_postrecov to initialize import for llog ctxt (3121)
6116 - replace config semaphore with spinlock (3306)
6117 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6118 - don't allow enqueue to complete on a destroyed export (3822)
6119 - down write_lock before checking llog header bitmap (3825)
6120 - recover from lock replay timeout (3764)
6121 - up llog sem before sending rpc (3652)
6122 - reduce ns lock hold times when setting kms (3267)
6123 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6124 - fix NULL deref and obd_dev leak on setup error (3312)
6125 - replace some LBUG about llog ops with error handling (3841)
6126 - don't match INVALID dentries from d_lookup and spin (3784)
6127 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6128 - fix invalid assertion in ptlrpc_set_wait (3880)
6130 - add libwrap support for the TCP acceptor (3996)
6131 - add /proc/sys/portals/routes for non-root route listing (3994)
6132 - allow setting MDS UUID in .xml (2580)
6133 - print the stack of a process that LBUGs (4228)
6135 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6138 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6139 - make sure to unset replay flag from failed open requests (3440)
6140 - if default stripe count is 0, use OST count for inode size (3636)
6141 - update parent mtime/ctime on client for create/unlink (2611)
6142 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6143 - free recovery state on server during a forced cleanup (3571)
6144 - unregister_reply for resent reqs (3063)
6145 - loop back devices mounting and status check on 2.6 (3563)
6146 - fix resource-creation race that can provoke i_size == 0 (3513)
6147 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6148 - more debugging for page-accounting assertion (3746)
6149 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6150 - avoid deadlock after precreation failure (3758)
6151 - fix race and lock order deadlock in orphan handling (3450, 3750)
6152 - add validity checks when grabbing inodes from l_ast_data (3599)
6154 - add /proc/.../recovery_status to obdfilter (3428)
6155 - lightweight CDEBUG infrastructure, debug daemon (3668)
6156 - change default OSC RPC parameters to be better on small clusters
6157 - turn off OST read cache for files smaller than 32MB
6158 - install man pages and include them in rpms (3100)
6159 - add new init script for (un)mounting lustre filesystems (2593)
6160 - run chkconfig in %post for init scripts (3701)
6161 - drop scimac NAL (unmaintained)
6163 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6166 - clean kiobufs before and after use (3485)
6167 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6168 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6169 - change init script to fail more clearly if not run as root (1528)
6170 - allow clients to reconnect during replay (1742)
6171 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6172 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6173 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6174 - fix cond_resched (3554)
6175 - only evict unfinished clients after recovery (3515)
6176 - allow bulk resend, prevent data loss (3570)
6177 - dynamic ptlrpc request buffer allocation (2102)
6178 - don't allow unlinking open directory if it isn't empty (2904)
6179 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6180 - remove extraneous obd dereference causing LASSERT failure (3334)
6181 - don't use get_cycles() when creating temp. files on the mds (3156)
6182 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6183 - handle EEXIST for set-stripe, set proper directory name (3336)
6185 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6186 - fix ksocknal_fmb_callback() error messages (2918)
6188 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6191 - don't copy lvb into (possibly NULL) reply on error (2983)
6192 - don't deref dentry after dput, don't free lvb on error (2922)
6193 - use the kms to determine writeback rpc length (2947)
6194 - increment oti_logcookies when osc is inactive (2948)
6195 - update client's i_blocks count via lvb messages (2543)
6196 - handle intent open/close of special files properly (1557)
6197 - mount MDS with errors=remount-ro, like obdfilter (2009)
6198 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6199 - don't use cancelling-locks' kms values (2947)
6200 - use highest lock extent for kms, not last one (2925)
6201 - don't dereference ERR_PTR() dentry in error handling path (3107)
6202 - fix thread race in portals_debug_dumplog() (3122)
6203 - create lprocfs device entries at setup instead of at attach (1519)
6204 - common AST error handler, don't evict client on completion race (3145)
6205 - zero nameidata in detach_mnt in 2.6 (3118)
6206 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6207 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6208 - initialize RPC timeout timer earlier for 2.6 (3219)
6209 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6210 - print nal/nid for unknown nid (3258)
6211 - additional checks for oscc recovery before doing precreate (3284)
6212 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6213 - don't crash in mdc_close for bad permissions on open (3285)
6214 - zero i_rdev for non-device files (3147)
6215 - clear page->private before handing to FS, better assertion (3119)
6216 - tune the read pipeline (3236)
6217 - fix incorrect decref of invalidated dentry (2350)
6218 - provide read-ahead stats and refine rpc in flight stats (3328)
6219 - don't hold journal transaction open across create RPC (3313)
6220 - update atime on MDS at close time (3265)
6221 - close LDAP connection when recovering to avoid server load (3315)
6222 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6223 - don't leak open file on MDS after open resend (3325)
6224 - serialize filter_precreate and filter_destroy_precreated (3329)
6225 - loop device shouldn't call sync_dev() for nul device (3092)
6226 - clear page cache after eviction (2766)
6227 - resynchronize MDS->OST in background (2824)
6228 - refuse to mount the same filesystem twice on same mountpoint (3394)
6229 - allow llmount to create routes for mounting behind routers (3320)
6230 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6231 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6232 - fix rare problem with rename on htree directories (3417)
6234 - allow default OST striping configuration per directory (1414)
6235 - fix compilation for qswnal for 2.6 kernels (3125)
6236 - increase maximum number of MDS request buffers for large systems
6237 - change liblustreapi to be useful for external progs like lfsck (3098)
6238 - increase local configuration timeout for slow disks (3353)
6239 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6241 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6244 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6245 - don't overwrite extent policy data in reply if lock was blocked (2901)
6246 - drop filter export grants atomically with removal from device (2663)
6247 - del obd_self_export from work_list in class_disconnect_exports (2908)
6248 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6249 - swab reply message in mdc_close, other PPC fixes (2464)
6250 - fix destroying of named logs (2325)
6251 - overwrite old logs when running lconf --write_conf (2264)
6252 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6253 - fix race in target_handle_connect (2898)
6254 - mds_reint_create() should take same inode create lock (2926)
6255 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6256 - don't close files for self_export to avoid uninitialized obd (2936)
6257 - allow MDS with the same name as client node (2939)
6258 - hold dentry reference for closed log files for unlink (2325)
6259 - reserve space for all logs during transactions (2059)
6260 - don't evict page beyond end of stripe extent (2925)
6261 - don't oops on a deleted current working directory (2399)
6262 - handle hard links to targets without a parent properly (2517)
6263 - don't dereference NULL lock when racing during eviction (2867)
6264 - don't grow lock extents when lots of conflicting locks (2919)
6266 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6269 - account for cache space usage on clients to avoid data loss (974)
6270 - lfsck support in lustre kernel code (2349)
6271 - reduce journal credits needed for BRW writes (2370)
6272 - orphan handling to avoid losing space on client/server crashes
6273 - ptlrpcd can be blocked, stopping ALL progress (2477)
6274 - use lock value blocks to assist in proper KMS, faster stat (1021)
6275 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6276 - recovery for initial connections (2355)
6277 - fixes for mds_cleanup_orphans (1934)
6278 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6279 - block all file creations until orphan recovery completes (1901)
6280 - client remove rq_connection from request struct (2423)
6281 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6282 - recovery timer race (2670)
6283 - mdc_close recovey bug (2532)
6284 - ptlrpc cleanup bug (2710)
6285 - mds timeout on local locks (2588)
6286 - namespace lock held during RPCs (2431)
6287 - handle interrupted sync write properly (2503)
6288 - don't try to handle a message that hasn't been replied to (2699)
6289 - client assert failure during cleanup after abort recovery (2701)
6290 - leak mdc device after failed mount (2712)
6291 - ptlrpc_check_set allows timedout requests to complete (2714)
6292 - wait for inflight reqs when ptlrpcd finishes (2710)
6293 - make sure unregistered services are removed from the srv_list
6294 - reset bulk XID's when resending them (caught by 1138 test)
6295 - unregister_bulk after timeout
6296 - fix lconf error (2694)
6297 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6298 - readahead locks pages, leaves pending causing memory pressure (2673)
6299 - increase OST request buffers to 4096 on large machines (2729)
6300 - fix up permission of existing directories in simple_mkdir (2661)
6301 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6302 - don't assign transno to errored transactions (2742)
6303 - don't delete objects on OST if given a bogus objid from MDS (2751)
6304 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6305 - if rq_no_resend, then timeout request after recovery (2432)
6306 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6307 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6308 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6309 - MDS recovery shouldn't delete objects that it has given out (2730)
6310 - if enqueue arrives after completion, don't clobber LVB (2819)
6311 - don't unlock pages twice when trigger_group_io returns error (2814)
6312 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6313 - don't write pages to disk if there was an error (1450)
6314 - don't ping imports that have recovery disabled (2676)
6315 - take buffered bytes into account when balancing socknal conn (2817)
6316 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6317 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6318 - remove little-endian swabbing of llog records (1987)
6319 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6320 - retry reposting request buffers if they fail (1191)
6321 - grow extent at grant time to avoid granting a revoked lock (2809)
6322 - lock revoke doesn't evict page if covered by a second lock (2765)
6323 - disable VM readahead to avoid reading outside lock extents (2805)
6325 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6326 - updated kernel patches for hp-2.4.20 kernel (2681)
6328 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6331 - fix truncated write corruption (2366)
6332 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6334 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6335 - ensure len(uuid) < 37 in lmc (1171)
6336 - fix ia64 OOPS in llog_test (2255)
6337 - zero end of page at obdfilter for partial page writes (2648)
6338 - don't leave stale dentries around after renames (bug 2428)
6339 - fix timeouts when evicting a client with a single lock held (2642)
6340 - set deadline for the initial HELLO message to drain (2634)
6341 - print out dotted-quad IP addresses in the socknal (2302)
6343 - additional debugging for MDS client eviction problem (2443)
6344 - fix mkfsoptions support for osts (2603, 2604)
6346 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6349 - add series for the vanilla 2.6.0 kernel
6350 - add series for the vanilla 2.4.24 kernel
6351 - add series for a cray x86/64 UL kernel drop
6352 - fix xattr patches for the vanilla 2.4.19 series
6354 - generate true UUIDs in lmc (1171)
6355 - have portals stack dumping break in UML (2466)
6356 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6357 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6358 - wake callback waiting threads on client eviction (2460)
6359 - Add --ptldebug and --subsystem to lmc (1719)
6360 - update assertion to allow safe interrupt allocation
6361 - set rq_no_resend for cancel requests (2432)
6362 - recalculate ptlrpcd timeout after resend (2494)
6363 - call vfs_rmdir when removing pending directories (2368)
6364 - fix renaming a file to itself (2429)
6365 - lmc creates a default one-stripe lov (2454)
6366 - expand procfs space to handle large clusters (2326)
6367 - increase UML stack to avoid overflow
6368 - update lconf's list of debug and subsystem masks
6369 - fix lfs find --obd (2510)
6370 - /proc tunable for disabling filter read caching (2591)
6371 - stop rpm packages from altering slapd.conf (2301)
6372 - disable nagle in the socknal under 0conf (2578)
6373 - choose mds inode size based on stripe count (2572)
6374 - fix kernel-source rpm problems (2516)
6376 - add --disable-doc to avoid pdf generation (2421)
6377 - update documentation, tests, type-os, comments
6378 - avoid format warnings on ia64
6379 - remove the TOE NAL
6380 - tiny code cleanups by removing unused fields
6382 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6385 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6386 - remove the most blatant lies from BUILDING file (2371)
6387 - change default debug level to reasonable production setting
6388 - reduce client side cache size to reduce cache flush time
6389 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6390 - make TCP zerocopy and pinger support enabled by default (2476)
6391 - sync writes completed after process exits caused crashes (2319)
6392 - maintain correct mount count on the MDS (2356)
6393 - backout 1557, because 2316 wasn't really fixed
6394 - better file I/O statistics gathering in /proc
6395 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6396 - another kernel patch to fix zero-copy TCP function export
6397 - don't take duplicate lock when processing re-sent getattr (2420)
6398 - lctl uses obd_self_export instead of creating new conn (2353)
6399 - MDS/OST recovery case which requires object creation asserted (2425)
6400 - move lfs from /usr/sbin to /usr/bin in packages
6401 - fix race between mds_client_add and mds_client_free (2417)
6402 - use kmalloc instead of slabs in portals (2430)
6403 - don't create duplicate records when a failover MDS is present (2442)
6404 - remove unnecessary mount age check (2332)
6405 - don't remove directory inodes from locks prematurely (2451)
6406 - don't break if MDS service name is the same as hostname (2103)
6407 - fix races in client write RPC generation when cache full (2482)
6409 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6412 - remove now-unused request->rq_obd (278)
6413 - if an allocation fails, print out how much memory we've used (1933)
6414 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6415 - add the "configurable stack size" patch to most series files (1256)
6416 - ability to write large log records, for 100+ OST configs (2306)
6417 - fix NULL deref when filter_prep fails (2314)
6418 - fix operator precedence error in filter_sync
6419 - dynamic allocation of socknal TX descriptors (2315)
6420 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6421 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6422 - ensure that CWARN messages in Portals always get to the syslog
6423 - __init/__exit are not for prototype decls (ldlm_init/exit)
6424 - x86-64 compile warning fixes
6425 - fix gateway LMC keyword conflict (2318)
6426 - fix MDS lock inversions in getattr/reint paths (1844)
6427 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6428 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6429 - DLM race condition prevented some lock evictions (2328)
6430 - ENOMEM detection and retry on socknal sends (2230)
6431 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6432 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6433 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6434 - better error messages when a client is rejected during recovery (1505)
6435 - avoid cancelling locks which were never granted, after failure (2330)
6436 - fix i_sem/journal inversion in mds_client_add (2333)
6437 - fix truncate/getattr lock cycle deadlock (2334)
6438 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6439 - fix two rare exit paths which could leak an l_lock() ref (2321)
6440 - fencepost error in MDS/OST orphan recovery (2226)
6441 - make log record alignment 8 bytes (1988)
6442 - lstripe now fails when requested offset > ost_count (2237)
6443 - ensure that all kernel series have a complete list.h (1607)
6444 - fix crashes in special-file operations (2316)
6445 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6447 - allow configurable automake binary, for testing new versions
6448 - small update to the lfs documentation
6450 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6452 * fix negative export reference count in fsfilt_sync (2312)
6454 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6455 * release candidate 0.9.1
6457 - orphans are moved into the PENDING directory for possible recovery
6458 - replayed opens now open by fid for orphan/rename safety (1042)
6459 - last close of an orphan inode generates a transno (683)
6460 - chdir() and mount() now pin the directory entry (1020)
6461 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6462 - discard very old requests without processing them (1502)
6463 - remove some common, well-understood CERRORs (1505)
6464 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6465 - clear "grant" flags in OST replies until OST grant code lands (1644)
6466 - fix read performance by not clobbering i_blksize on client (1598)
6467 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6468 - make LRU size a /proc tunable, clears locks when reduced (707)
6469 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6470 - server locks take references on exports now (1558)
6471 - build fixes for 2.4.20-rh trees (1663)
6472 - return an error from lov_create if all OSCs are inactive (1751)
6473 - fix import levels when a reconnect happens without a timeout (1597)
6474 - exit early from mds_open if we get a lookup error (1749)
6475 - partial page read at EOF wouldn't wait for disk before sending (1642)
6476 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6477 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6478 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6479 - write/truncate lock inversion (1639)
6480 - Don't auto-load obdclass, portals modules during cleanup (1495)
6481 - fix timestamps from jumping to "now" (1763)
6482 - extra journal assertions (1648)
6483 - add an extra multiunlink test (1771)
6484 - fix read_record/write_record API (1776)
6485 - fix leak of offset_extent, possible incorrect i_size later (1772)
6486 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6487 - Debugging for the kqswnal_get_idle_tx problems (1820)
6488 - Allow recovery to be attempted multiple times (1536)
6489 - Write out MDS last_rcvd file after it is first created (1600)
6490 - Fix tx_descriptor leak in failed transmit situations (1827)
6491 - ext3 journaling fixes for assertion failure after IO error (1871)
6492 - class_export_put() on freed export after completion AST error (1896)
6493 - Fix revalidate looping in VFS (1322)
6494 - Don't access a freed export during MDS_REINT timeout (1521)
6495 - Add open-unlink recovery support on the MDS (1673,1764)
6496 - Return an error if no MDS data was read from last_rcvd (1946)
6497 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6498 - Don't setup a disk device that doesn't match exported UUID (317)
6499 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6500 - avoid committing NULL handle in force close
6501 - local.sh is now a one-stripe LOV configuration
6502 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6503 - fix ext3 htree duplicate directory entry corruption (1516)
6504 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6505 - update to new LOV EA format (2097)
6506 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6508 2003-06-15 Phil Schwan <phil@clusterfs.com>
6511 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6512 - per-import/export recovery handling (958, 931, 959)
6513 - multiple last-rcvd slots, for serving multiple FSes (949)
6514 - connections are again shared between multiple imp/exports (963, 964)
6515 - "umount -f" would hang if any requests needed to be sent (393, 978)
6516 - avoid pinning large req buffer by copying for queued messages (989)
6517 - add "uuid" to "lctl device" command to help upcalls (991)
6518 - "open" RPCs with transnos would confuse recovery counters (1037)
6519 - do proper endian conversion of all wire messages (288, 340, 891)
6520 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6521 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6522 - fix ptlrpc_connection leak in target_handle_connect (1174)
6523 - fix import refcounting bug in OST and MDS cleanup (1134)
6524 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6525 - fix very unlikely obd_types race condition (501)
6526 - remove osc_open hack for echo_client (1187)
6527 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6528 - a failure in read_inode2 leads to deadlock (1139)
6529 - cancel ack-locks as soon as transaction is committed (1072)
6530 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6531 - make sure to commitrw after any preprw to avoid deadlock (1162)
6532 - failing to execute a file in a lustre FS would lock inode (1203)
6533 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6534 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6535 - obd_uuid2tgt would walk off the end of the list (1255)
6536 - on IA64 the osc would give portals incorrect bulk size (1258)
6537 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6538 - fix lock inversion caused by new llite matching code (1282)
6539 - limit the number of dirty pages on a client to 10MB (1286)
6540 - timed out locks were not being corrected cancelled (1289)
6541 - fix O_DIRECT above 4GB on IA-32 (1292)
6542 * major user-visible changes
6543 - fail out/fail over policy now controlled by the upcall (993)
6545 - add OBD_PING to check server availability and failure (954)
6546 - lustre messages are now sent in sending host order (288, 340, 891)
6547 - add eadatalen to MDS getattr reply (340)
6548 - OST read replies may contain second buffer, with per-page status (593)
6550 2003-03-11 Phil Schwan <phil@clusterfs.com>
6553 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6554 - failed open()s could cause deadlock; fixed (867, 869)
6555 - stop cancelling OST locks when files are closed (481)
6556 - overlapping XID spaces caused network corruption (851, 853)
6557 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6558 - setattr_raw added, to do single-RPC, server-side setattrs
6559 - lmc/lconf syntax change for OST UUIDs
6560 - fix crashy race condition between ptlrpc_free_req and osc_close
6561 - don't use request in mdc_enqueue if we hit a timeout (889)
6562 - don't set the inode i_size for regular files from the MDS (896)
6563 - handle out of order completion AST (842)
6564 - don't LBUG if a lock request times out after receiving AST (913)
6565 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6566 - if a bad lock AST arrives, send an error instead of dropping entirely
6567 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6568 - fix leak in bulk IO when only partially completed (899, 900, 926)
6569 - fix O_DIRECT for ia64 (55)
6570 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6571 - C-z after timeout could hang a process forever; fixed (977)
6573 - client-side I/O cache (678, 924, 929, 941, 970)
6575 - READPAGE and SETATTRs which don't take server-side locks get
6578 2003-02-11 Phil Schwan <phil@clusterfs.com>
6581 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6582 - Fix fsfilt_extN_readpage() to read a full page of directory
6583 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6584 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6585 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6586 - Most intent operations were removed, in favour of a new RPC mode
6587 that does a single RPC to the server and bypasses most of the VFS
6588 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6589 - Aggressively cancel local locks on DLM servers
6590 - mds_reint_unlink sends EA to the client if it's the last nlink.
6591 client uses that EA to unlink OST objects.
6592 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6593 - recursive symlinks were fixed (439)
6594 - fixed NULL deref in DEBUG_REQ
6595 - filter_update_lastobjid no longer calls sync, which annoyed extN
6596 - fixed multi-client small-writes to a single file problem (445)
6597 - fixed mtime updates during file writes (607)
6598 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6599 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6600 - fixed Linux OST size reporting problem (444, 656)
6601 - OST now updates object mtime with writes or setattr (607, 619)
6602 - client verifies file size before zeroing page past EOF (445)
6603 - OST now writes last allocated objid to disk with allocation (108)
6604 - LOV on echo now works (409)
6606 - mds_reint_unlink sends a new buffer, with the EA included. this
6607 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6608 set if a regular file was being unlinked, and it was the last link
6609 - use PtlGet from the target for bulk writes (315)
6610 - OST now updates object mtime with writes or setattr (607, 619)
6611 - LDLM now has a grant-time callback to revalidate locked items, if
6613 - Many MDS operations were reorganized to combat race conditions
6615 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6616 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6618 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6621 - Fully reactivate OST imports after reconnection (512, others)
6622 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6623 - More graceful error handling for truncating on dead OST (515)
6624 - Don't error out unless we're actually accessing dead stripes (474)
6625 - Fix garbage sizes when stripes are missing (410)
6626 - LRU counters were broken, causing constant lock purge (433, 432)
6627 - garbage on read from stripes with failed OSTs (441)
6628 - mark OSCs as active before reconnecting during recovery (438)
6629 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6630 - lfind did not preserve OST order in output (443)
6631 - symlinks cause hung clients, incorrect data (439)
6632 - stop dereferencing request after dropping refcount (457)
6633 - don't LASSERT(spin_is_locked) on non-SMP (455)
6634 - fixes for many rename() bugs
6635 - fstat didn't correctly synchronize attributes (399)
6636 - server must handle lock cancellation during blocking AST prep (487)
6637 - bulk descriptors were free()d too soon (511)
6638 - fix paths in lconf, which would load incorrect modules (451, 507)
6639 - fix confusing lconf 'host not found' error message (386)
6640 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6641 - fix race condition in mdc_blocking_ast() for inode access (526)
6642 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6643 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6644 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6645 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6646 - fix lstripe in several ways (kernel side) (527)
6647 - fix request leak in ldlm_cli_enqueue (262)
6648 - incorrect OSC was marked inactive after OST failure
6649 - call mds_fs_cleanup before unmounting filesystem (524)
6650 - fix races between taking ns_lock and ldlm_lock_change_resource
6651 - fix races updating LOV export open file list
6652 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6653 - fix recovery NULL deref in ldlm_cli_cancel_unused
6654 - fix some DLM races by using new hash table for lock handles (419)
6655 - permit the client to specify desired inodes, at replay
6656 - duplicate requests when we queue them for replay reintegration
6657 - fix last_rcvd offset calculation
6658 - sync after each recovered transaction, so we always make progress
6659 - never, not always, ERESTART requests without transnos
6660 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6662 - skip replay if the MDS says that the client is already connected
6663 - don't check for a recovery-enabled export to match lctl's UUID
6664 - don't INC_USE_COUNT for phantom exports
6665 - don't crash when cleaning up phantom exports (567)
6666 - don't double-finish or set replay data for errored mdc_open requests
6667 - abort requests when they time out, so we don't get old replies
6668 - send/receive replies for AST messages again
6669 - if the client says that it doesn't have the lock, cancel it on the
6671 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6672 mark it as destroyed, it will all work out when decref is called
6673 - fix module use counts (22, 581)
6675 - ASTs now expect a reply (server cancels lock on error reply)
6677 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6680 - fix many simultaneous client startup (392)
6681 - fix dentry->d_it clobbering
6682 - credentials weren't being shipped for readdir/getattr operations
6683 - remove invalid assertions triggered during some concurrent MD
6685 - proper Lustre versions added (336, 389)
6686 - fix memory leak for create error case (398)
6687 - fix LOV locking bug that would get cli/srv out of sync
6688 - fix echo client over LOV (409)
6689 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6690 - fix double-O_EXCL intent crash (424)
6691 - avoid sending multiple lock CANCELs (352)
6693 - MDS can do multi-client recovery (modulo bugs in new code)
6695 - many updates, edits, cleanups
6697 2002-11-18 Phil Schwan <phil@clusterfs.com>
6700 - fix null d_it dereference (346)
6701 - fix full OST/dbench hang (333)
6702 - fix permission problem with file removal (286)
6703 - fix removal of OSCs from LOV when they fail
6704 - fix NULL deref during bulk timeout (214)
6705 - fix problems related to multiple filesystems on one MDS (241)
6706 - fixed serious subtle metadata locking bugs
6707 - free locks on clients when inodes are removed due to memory
6709 - fix inode pointer in lock data (285)
6710 - partial support for multiple MDS on a single host (241)
6711 - data locks weren't cancelled at clear_inode time (290, 311)
6712 - intent locks could lead to unbounded lock growth (205)
6713 - added a maximum lock count, an LRU list, and a flusher
6714 - fix multiple rename (365)
6715 - properly abstracted the echo client
6716 - OSC locked 1 byte too many; fixed
6717 - rewrote brw callback code:
6718 - fixed recovery bugs related to LOVs (306)
6719 - fixed too-many-pages-in-one-write crash (191)
6720 - fixed (again) crash in sync_io_timeout (214)
6721 - probably fixed callback-related race (385)
6723 - Add capability to MDS protocol
6724 - LDLM cancellations and callbacks on different portals
6726 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6729 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6730 - defer open object destruction to close time (601981, 138)
6731 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6732 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6733 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6734 - don't call obd_setattr in ll_file_release for destroyed objects
6735 * protocol change to lustre_msg: move |version| and add |flags|
6736 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6737 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6738 * added replay of create, unlink, link and rename operations during
6739 MDS failover; recovery should be much more robust now
6740 * remove failed OSCs from LOVs (only lov_create uses this so far)
6741 * the lustre-HOWTO was brought (more) up to date (582544)
6743 2002-10-23 Phil Schwan <phil@clusterfs.com>
6746 - in-use dentries weren't being reused properly (617851)
6747 - prevent multiple LDLM setup (599178)
6748 - fix LOV size calculations for truncate (617853)
6749 - fix client handling of MDS intent errors (POSIX)
6750 - fix permission bug in lovstripe.c test (624321)
6751 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6752 - truncate past end of file could corrupt data
6753 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6754 - a race in recovery could return ETIMEDOUT to apps (623947)
6755 - building outside the source directory was fixed
6756 * the lustre-HOWTO was brought (more) up to date (582544)
6757 * major progress was made on recovery functionality
6759 2002-10-10 Phil Schwan <phil@clusterfs.com>
6762 - recovery deadlock fix
6763 - rm -rf causes LBUG fix (617817)
6764 - file open by multiple tasks fix (618962)
6765 - directory permissions bugs (602707 and 620007)
6766 - journal_stop fixed with locking (611313)
6767 - O_APPEND failures resolved (618273, perhaps 614459)
6768 - lconf PATH fix (619770)
6769 - IA64 build fix (621450)
6770 - RPC buffer sizes scale with amount of memory
6772 2002-10-01 Phil Schwan <phil@clusterfs.com>
6775 - locks would be cancelled without throwing away data pages,
6776 resulting in inconsistent data (605627)
6777 - inode attributes were not always being refreshed (605627, 612449)
6778 - lconf now continues to cleanup after lctl reports an error
6779 - MDS now enforces user permissions (602707)
6780 - lprocfs cleanup fixed, but not yet enabled (614157)
6781 - fixed infinite server hang, should a client not respond to an AST
6782 - avoid going into recovery if user calls readlink() with a buffer
6783 that's too small (613941)
6784 - AST RPCs no longer require replies (614867) -- this may be changed
6785 - don't crash server if client sends an IOV that's too big (611336)
6786 - fixed lock conversion deadlock (611892)
6787 - fixed the following of symlinks (614622)
6788 * recovery: the server can remove locks from a client that dies, other
6789 clients can make progress
6790 * more extN patch fixes
6791 * compile-time configurable ptlrpc buffer allocations
6793 - collaborative read cache document
6794 - Lustre Lite Performance CDR document-in-progress
6796 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6799 - fix typo in patch-2.4.18
6801 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6804 - clear ptlrpc request each time in handle_incoming_request()
6805 - unlink of files now destroys the object on the OST
6807 2002-09-19 Peter Braam <braam@clusterfs.com>
6809 * add hard link support
6810 * change obdfile creation method
6811 * kernel patch changed
6813 2002-09-19 Peter Braam <braam@clusterfs.com>
6816 - stack overflow bug in extN fixed
6818 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6820 * documentation updates
6821 - add man pages for config tools
6822 - update tests/README to describe testing with new config tools
6823 - finish metadata API descriptions
6824 * bug fixes and cleanups
6825 - statfs workaround for 16TB limit
6826 - LOV stripe allocation improved, can stripe on subset of OSTs
6827 - LOV file size/IO offset was wrong for files > 4GB in size
6828 - object EA data was being dropped, caused files to be unreadable
6829 - memory overflow with non-LOV OST caused memory corruption
6830 - fixed regression tests to work with new config tools, obdfilter
6831 - fixed bug when directory size became larger than 1 block
6832 - fixed bug (for single client case) when PWD was deleted
6833 - invalidate local directory pages when doing intent-based ops
6834 - avoid LDLM oops when lock callback contained bad data
6836 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6838 * documentation updates
6839 * bug fixes and cleanups
6840 - configuration tools
6843 - 64-bit compile warnings
6844 - 64-bit internal statfs data
6846 * test_brw on persistent OST devices
6848 * lprocfs (disabled)
6850 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6852 * documentation updates
6853 * bug fixes and cleanups
6854 * configuration tools
6856 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6859 * many small fixes to 0_5_4
6860 * io/network handling
6861 * thinkos in MDS operations
6863 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6866 * crucial basic fixes to 0.5.3
6867 * IOR, Iozone work over Elan
6870 2002-08-07 Phil Schwan <phil@clusterfs.com>
6871 * version 0_5_3, our first alpha
6872 * we use the new Portals iovs
6873 * documentation updates
6874 * bug fixes and cleanups
6875 * small changes in the DLM wire protocol
6877 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6878 * version 0_5_1 with some initial stability,
6879 * locking on MD and file I/O.
6880 * documentation updates
6881 * several bug fixes since 0.5.0
6882 * small changes in wire protocol
6884 2002-07-18 Phil Schwan <phil@clusterfs.com>
6886 * delivered as Lustre Light Alpha
6887 * fixed a crash after handling invalid MDS requests
6888 * fixed directory pages for architectures with non-4k pages sizes
6890 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6891 * release version v0_4_4
6892 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6894 * Fixes a number of interruption problems with OST operations.
6895 * Update documentation for portals header changes
6896 * Move all wire protocol structs/defines to lustre_idl.h
6897 * Fixes symlink length bug.
6898 * Add tcpdump to repository.
6900 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6901 * release version v0_4_3
6902 * Fixes statfs for inodes on extN.
6903 * Fixes bug in runtests which would delete /etc/hosts.
6904 * Use 64-bit object IDs wherever possible (not into VFS though)
6905 Remove ost_get_info, which is unused by lustre, and out of date.
6907 2002-07-03 Peter Braam <braam@clusterfs.com>
6908 * release version v0_4_2 Fixes a lookup error (type not passed)
6909 * move forward to head of Portals
6910 * move forward to latest Lustre kernel
6912 2002-06-25 Peter Braam <braam@clusterfs.com>
6913 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob