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 5.14.0-284.25.1.el9 (RHEL9.2)
7 5.14.0-162.23.1.el9 (RHEL9.1)
8 4.18.0-477.21.1.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 5.14.0-284.25.1.el9 (RHEL9.2)
37 5.14.0-162.23.1.el9 (RHEL9.1)
38 4.18.0-477.21.1.el8 (RHEL8.8)
39 5.4.0-37 (Ubuntu 20.04)
40 5.14.21-150400.24.46 (SLES15 SP4)
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 4.4.0-131 (Ubuntu 16.04)
66 4.15.0-48 (Ubuntu 18.04)
67 5.8.0-53 (Ubuntu 20.04.2 HWE)
68 5.11.0-31 (Ubuntu 20.04.3 HWE)
69 5.11.0 (vanilla kernel.org)
70 5.10.0 (openEuler 22.03 LTS)
71 * Recommended e2fsprogs version: 1.47.0-wc1 or newer
72 * Recommended ZFS version: 2.1.11
73 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
74 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
75 stack overflow. For more information, please refer to bugzilla 17630.
76 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
77 the re-exporting nodes to be mounted with "32bitapi" mount option
79 --------------------------------------------------------------------------------
84 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
85 for currently supported client and server kernel versions.
86 * Server primary kernels built and tested during release cycle:
87 4.18.0-348.2.1.el8 (RHEL8.5)
88 * Other server kernels known to build and work at some point (others may also work):
89 3.10.0-862.14.4.el7 (RHEL7.5)
90 3.10.0-957.27.2.el7 (RHEL7.6)
91 3.10.0-1062.18.1.el7 (RHEL7.7)
92 3.10.0-1127.19.1.el7 (RHEL7.8)
93 3.10.0-1160.49.1.el7 (RHEL7.9)
94 4.18.0-80.11.2.el8 (RHEL8.0)
95 4.18.0-147.8.1.el8 (RHEL8.1)
96 4.18.0-193.28.1.el8 (RHEL8.2)
97 4.18.0-240.22.1.el8 (RHEL8.3)
98 4.18.0-305.25.1.el8 (RHEL8.4)
99 4.4.120-92.70 (SLES12 SP2)
100 4.4.180-94.100 (SLES12 SP3)
101 4.4.0-131 (Ubuntu 16.04)
102 4.15.0-32 (Ubuntu 18.04)
103 5.4.0-48 (Ubuntu 20.04)
104 vanilla linux 5.4.0 (ZFS + ldiskfs)
105 vanilla linux 5.4.21 (ZFS + ldiskfs)
106 vanilla linux 5.4.136 (ZFS + ldiskfs)
107 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
108 * Client primary kernels built and tested during release cycle:
109 4.18.0-348.2.1.el8 (RHEL8.5)
110 5.4.0-37 (Ubuntu 20.04)
111 5.3.18-59.27 (SLES15 SP3)
112 * Other clients known to build on these kernels at some point (others may also work):
113 3.10.0-862.14.4.el7 (RHEL7.5)
114 3.10.0-957.27.2.el7 (RHEL7.6)
115 4.14.0-49.13.1.el7a (RHEL7.5)
116 4.14.0-115.2.2.el7a (RHEL7.6)
117 3.10.0-1062.18.1.el7 (RHEL7.7)
118 3.10.0-1127.19.1.el7 (RHEL7.8)
119 3.10.0-1160.49.1.el7 (RHEL7.9)
120 4.18.0-80.11.2.el8 (RHEL8.0)
121 4.18.0-147.8.1.el8 (RHEL8.1)
122 4.18.0-193.28.1.el8 (RHEL8.2)
123 4.18.0-240.22.1.el8 (RHEL8.3)
124 4.18.0-305.25.1.el8 (RHEL8.4)
125 4.4.120-92.70 (SLES12 SP2)
126 4.4.180-94.100 (SLES12 SP3)
127 4.12.14-95.48 (SLES12 SP4)
128 4.12.14-122.91 (SLES12 SP5)
129 4.12.14-197.75 (SLES15 SP1)
130 5.3.18-24.96 (SLES15 SP2)
131 4.4.0-131 (Ubuntu 16.04)
132 4.15.0-48 (Ubuntu 18.04)
133 5.8.0-53 (Ubuntu 20.04.2 HWE)
134 5.11.0-31 (Ubuntu 20.04.3 HWE)
135 5.11.0 (vanilla kernel.org)
136 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
137 * Recommended ZFS version: 2.0.7
138 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
139 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
140 stack overflow. For more information, please refer to bugzilla 17630.
141 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
142 the re-exporting nodes to be mounted with "32bitapi" mount option
144 --------------------------------------------------------------------------------
148 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
149 for currently supported client and server kernel versions.
150 * Server primary kernels built and tested during release cycle:
151 3.10.0-1062.18.1.el7 (RHEL7.7)
152 3.10.0-1127.19.1.el7 (RHEL7.8)
153 3.10.0-1160.6.1.el7 (RHEL7.9)
154 * Other server kernels known to build and work at some point (others may also work):
155 3.10.0-862.14.4.el7 (RHEL7.5)
156 3.10.0-957.27.2.el7 (RHEL7.6)
157 4.18.0-80.11.2.el8 (RHEL8.0)
158 4.18.0-147.8.1.el8 (RHEL8.1)
159 4.18.0-193.28.1.el8 (RHEL8.2)
160 4.18.0-240.1.1.el8 (RHEL8.3)
161 4.4.120-92.70 (SLES12 SP2)
162 4.4.180-94.100 (SLES12 SP3)
163 4.4.0-131 (Ubuntu 16.04)
164 4.15.0-32 (Ubuntu 18.04)
165 5.4.0-48 (Ubuntu 20.04)
166 vanilla linux 5.4.0 (ZFS + ldiskfs)
167 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
168 * Client primary kernels built and tested during release cycle:
169 3.10.0-1062.18.1.el7 (RHEL7.7)
170 3.10.0-1127.19.1.el7 (RHEL7.8)
171 3.10.0-1160.6.1.el7 (RHEL7.9)
172 4.12.14-95.48 (SLES12 SP4)
173 4.12.14-122.20 (SLES12 SP5)
174 4.12.14-197.75 (SLES15 SP1)
175 5.3.18-24.24 (SLES15 SP2)
176 4.15.0-48 (Ubuntu 18.04)
177 * Other clients known to build on these kernels at some point (others may also work):
178 3.10.0-862.14.4.el7 (RHEL7.5)
179 3.10.0-957.27.2.el7 (RHEL7.6)
180 4.14.0-49.13.1.el7a (RHEL7.5)
181 4.14.0-115.2.2.el7a (RHEL7.6)
182 4.18.0-80.11.2.el8 (RHEL8.0)
183 4.18.0-147.8.1.el8 (RHEL8.1)
184 4.18.0-193.28.1.el8 (RHEL8.2)
185 4.18.0-240.1.1.el8 (RHEL8.3)
186 4.4.120-92.70 (SLES12 SP2)
187 4.4.180-94.100 (SLES12 SP3)
188 4.4.0-131 (Ubuntu 16.04)
189 5.4.0-37 (Ubuntu 20.04)
190 5.4.0 (vanilla kernel.org)
191 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
192 * Recommended ZFS version: 2.0.0
193 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
194 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
195 stack overflow. For more information, please refer to bugzilla 17630.
196 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
197 the re-exporting nodes to be mounted with "32bitapi" mount option
199 --------------------------------------------------------------------------------
203 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
204 for currently supported client and server kernel versions.
205 * Server primary kernels built and tested during release cycle:
206 3.10.0-1062.1.1.el7 (RHEL7.7)
207 * Other server kernels known to build and work at some point (others may also work):
208 3.10.0-862.14.4.el7 (RHEL7.5)
209 3.10.0-957.27.2.el7 (RHEL7.6)
210 4.18.0-80.7.1.el8 (RHEL8.0)
211 4.4.120-92.70 (SLES12 SP2)
212 4.4.180-94.100 (SLES12 SP3)
213 4.4.0-131 (Ubuntu 16.04)
214 4.15.0-32 (Ubuntu 18.04)
215 vanilla linux 4.6.7 (ZFS only)
216 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
217 * Client primary unpatched kernels built and tested during release cycle:
218 3.10.0-1062.1.1.el7 (RHEL7.7)
219 4.12.14-95.29 (SLES12 SP4)
220 4.15.0-48 (Ubuntu 18.04)
221 * Other clients known to build on unpatched kernels at some point (others may also work):
222 3.10.0-862.14.4.el7 (RHEL7.5)
223 3.10.0-957.27.2.el7 (RHEL7.6)
224 4.14.0-49.13.1.el7a (RHEL7.5)
225 4.14.0-115.2.2.el7a (RHEL7.6)
226 4.18.0-80.7.1.el8 (RHEL8.0)
227 4.4.120-92.70 (SLES12 SP2)
228 4.4.180-94.100 (SLES12 SP3)
229 4.4.0-131 (Ubuntu 16.04)
230 vanilla linux 4.15.0 (ZFS only)
231 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
232 * Recommended ZFS / SPL version: 0.7.13
233 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
234 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
235 stack overflow. For more information, please refer to bugzilla 17630.
236 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
237 the re-exporting nodes to be mounted with "32bitapi" mount option
238 * Note for Ubuntu 14.04.5:
239 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
240 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
242 --------------------------------------------------------------------------------
246 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
247 for currently supported client and server kernel versions.
248 * Server known to build on the following kernels (others may also work):
249 2.6.32-696.18.7.el6 (RHEL6.9)
250 3.10.0-862.14.4.el7 (RHEL7.5)
251 3.10.0-957.el7 (RHEL7.6)
252 4.4.120-92.70 (SLES12 SP2)
253 4.4.162-94.69 (SLES12 SP3)
254 4.4.0-131 (Ubuntu 16.04)
255 4.15.0-32 (Ubuntu 18.04)
256 vanilla linux 4.6.7 (ZFS only)
257 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
258 * Client known to build on unpatched kernels (others may also work):
259 2.6.32-696.18.7.el6 (RHEL6.9)
260 2.6.32-754.6.3.el6 (RHEL6.10)
261 3.10.0-862.14.4.el7 (RHEL7.5)
262 3.10.0-957.el7 (RHEL7.6)
263 4.14.0-49.13.1.el7a (RHEL7.5)
264 4.14.0-115.2.2.el7a (RHEL7.6)
265 4.4.120-92.70 (SLES12 SP2)
266 4.4.162-94.69 (SLES12 SP3)
267 4.4.0-131 (Ubuntu 16.04)
268 4.15.0-32 (Ubuntu 18.04)
269 vanilla linux 4.15.0 (ZFS only)
270 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
271 * Recommended ZFS / SPL version: 0.7.9
272 * Tested with ZFS / SPL version: 0.7.9
273 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
274 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
275 stack overflow. For more information, please refer to bugzilla 17630.
276 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
277 the re-exporting nodes to be mounted with "32bitapi" mount option
278 * Note for Ubuntu 14.04.5:
279 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
280 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
282 --------------------------------------------------------------------------------
284 2018-04-03 Intel Corporation
286 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
287 for currently supported client and server kernel versions.
288 * Server known to build on patched kernels:
289 2.6.32-431.29.2.el6 (RHEL6.5)
290 2.6.32-504.30.3.el6 (RHEL6.6)
291 2.6.32-573.26.1.el6 (RHEL6.7)
292 2.6.32-642.15.1.el6 (RHEL6.8)
293 2.6.32-696.18.7.el6 (RHEL6.9)
294 3.10.0-693.21.1.el7 (RHEL7.4)
295 3.0.101-0.47.71 (SLES11 SP3)
296 3.0.101-107 (SLES11 SP4)
297 3.12.74-60.64.40 (SLES12 SP1)
298 4.4.114-92.64 (SLES12 SP2)
299 4.4.114-94.11 (SLES12 SP3)
300 3.13.0-101 (Ubuntu 14.04, ZFS only)
301 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
302 4.4.0-116 (Ubuntu 16.04)
303 vanilla linux 4.6.7 (ZFS only)
304 * Client known to build on unpatched kernels:
305 2.6.32-431.29.2.el6 (RHEL6.5)
306 2.6.32-504.30.3.el6 (RHEL6.6)
307 2.6.32-573.26.1.el6 (RHEL6.7)
308 2.6.32-642.15.1.el6 (RHEL6.8)
309 2.6.32-696.18.7.el6 (RHEL6.9)
310 3.10.0-693.21.1.el7 (RHEL7.4)
311 3.0.101-0.47.71 (SLES11 SP3)
312 3.0.101-107 (SLES11 SP4)
313 3.12.74-60.64.40 (SLES12 SP1)
314 4.4.114-92.64 (SLES12 SP2)
315 4.4.114-94.11 (SLES12 SP3)
316 3.13.0-101 (Ubuntu 14.04)
317 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
318 4.4.0-116 (Ubuntu 16.04)
320 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
321 * Recommended ZFS / SPL version: 0.7.6
322 * Tested with ZFS / SPL version: 0.7.6
323 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
324 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
325 stack overflow. For more information, please refer to bugzilla 17630.
326 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
327 the re-exporting nodes to be mounted with "32bitapi" mount option
328 * Note for Ubuntu 14.04.5:
329 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
330 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
332 --------------------------------------------------------------------------------
334 2017-07-05 Intel Corporation
336 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
337 for currently supported client and server kernel versions.
338 * Server known to build on patched kernels:
339 2.6.32-431.29.2.el6 (RHEL6.5)
340 2.6.32-504.30.3.el6 (RHEL6.6)
341 2.6.32-573.26.1.el6 (RHEL6.7)
342 2.6.32-642.15.1.el6 (RHEL6.8)
343 2.6.32-696.3.1.el6 (RHEL6.9)
344 3.10.0-514.21.1.el7 (RHEL7.3)
345 3.0.101-0.47.71 (SLES11 SP3)
346 3.0.101-100 (SLES11 SP4)
347 3.12.74-60.64.40 (SLES12 SP1)
348 4.4.59-92.17 (SLES12 SP2)
349 vanilla linux 4.6.7 (ZFS only)
350 * Client known to build on unpatched kernels:
351 2.6.32-431.29.2.el6 (RHEL6.5)
352 2.6.32-504.30.3.el6 (RHEL6.6)
353 2.6.32-573.26.1.el6 (RHEL6.7)
354 2.6.32-642.15.1.el6 (RHEL6.8)
355 2.6.32-696.3.1.el6 (RHEL6.9)
356 3.10.0-514.21.1.el7 (RHEL7.3)
357 3.0.101-0.47.71 (SLES11 SP3)
358 3.0.101-100 (SLES11 SP4)
359 3.12.74-60.64.40 (SLES12 SP1)
360 4.4.59-92.17 (SLES12 SP2)
362 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
363 * Recommended ZFS / SPL version: 0.6.5.9
364 * Tested with ZFS / SPL version: 0.6.5.9
365 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
366 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
367 stack overflow. For more information, please refer to bugzilla 17630.
368 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
369 the re-exporting nodes to be mounted with "32bitapi" mount option
371 --------------------------------------------------------------------------------
373 12-07-2016 Intel Corporation
375 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
376 for currently supported client and server kernel versions.
377 * Server known to build on patched kernels:
378 2.6.32-431.29.2.el6 (RHEL6.5)
379 2.6.32-504.30.3.el6 (RHEL6.6)
380 2.6.32-573.26.1.el6 (RHEL6.7)
381 2.6.32-642.11.1.el6 (RHEL6.8)
382 3.10.0-514.el7 (RHEL7.3)
383 3.0.101-0.47.71 (SLES11 SP3)
384 3.0.101-80 (SLES11 SP4)
385 3.12.62-60.64.8 (SLES12 SP1)
386 vanilla linux 4.5.7 (ZFS only)
387 * Client known to build on unpatched kernels:
388 2.6.32-431.29.2.el6 (RHEL6.5)
389 2.6.32-504.30.3.el6 (RHEL6.6)
390 2.6.32-573.26.1.el6 (RHEL6.7)
391 2.6.32-642.11.1.el6 (RHEL6.8)
392 3.10.0-514.el7 (RHEL7.3)
393 3.0.101-0.47.71 (SLES11 SP3)
394 3.0.101-80 (SLES11 SP4)
395 3.12.62-60.64.8 (SLES12 SP1)
397 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
398 * Recommended ZFS / SPL version: 0.6.5.8
399 * Tested with ZFS / SPL version: 0.6.5.8
400 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
401 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
402 stack overflow. For more information, please refer to bugzilla 17630.
403 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
404 the re-exporting nodes to be mounted with "32bitapi" mount option
406 --------------------------------------------------------------------------------
408 02-29-2016 Intel Corporation
410 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
411 for currently supported client and server kernel versions.
412 * Server known to build on patched kernels:
413 2.6.32-431.29.2.el6 (RHEL6.5)
414 2.6.32-504.30.3.el6 (RHEL6.6)
415 2.6.32-573.12.1.el6 (RHEL6.7)
416 3.10.0-327.3.1.el7 (RHEL7.2)
417 3.0.101-0.47.71 (SLES11 SP3)
418 3.0.101-68 (SLES11 SP4)
419 vanilla linux 4.2.1 (ZFS only)
420 * Client known to build on unpatched kernels:
421 2.6.32-431.29.2.el6 (RHEL6.5)
422 2.6.32-504.30.3.el6 (RHEL6.6)
423 2.6.32-573.12.1.el6 (RHEL6.7)
424 3.10.0-327.3.1.el7 (RHEL7.2)
425 3.0.101-0.47.71 (SLES11 SP3)
426 3.0.101-68 (SLES11 SP4)
429 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
430 * Recommended ZFS / SPL version: 0.6.4.2
431 * Tested with ZFS / SPL version: 0.6.4.2
432 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
433 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
434 stack overflow. For more information, please refer to bugzilla 17630.
435 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
436 the re-exporting nodes to be mounted with "32bitapi" mount option
438 --------------------------------------------------------------------------------
440 03-10-2015 Intel Corporation
442 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
443 for currently supported client and server kernel versions.
444 * Server known to build on patched kernels:
445 2.6.32-431.29.2.el6 (RHEL6.5)
446 2.6.32-504.8.1.el6 (RHEL6.6)
447 3.0.101-0.46 (SLES11 SP3)
448 * Client known to build on unpatched kernels:
449 2.6.32-431.29.2.el6 (RHEL6.5)
450 2.6.32-504.8.1.el6 (RHEL6.6)
451 3.10.0-123.20.1.el7 (RHEL7)
452 3.0.101-0.46 (SLES11 SP3)
453 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
454 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
455 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
456 stack overflow. For more information, please refer to bugzilla 17630.
457 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
458 the re-exporting nodes to be mounted with "32bitapi" mount option
460 Severity : enhancement
462 Description: control OST-index in IDIF via ROCOMPAT flag.
463 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
464 last_rcvd file. For new formatted OST device, it will be auto set;
465 for the case of upgrading from old OST device, you can enable it
466 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
467 enabled, for new created OST-object, its IDIF-in-LMA will contain
468 the OST-index; for the existing OST-object, the OSD will convert
469 old format IDIF as new format IDIF with OST-index stored in the
470 LMA EA when accessing such OST-object or via OI scrub. Once such
471 flag is enabled, it cannot be reverted back, so the system cannot
472 be downgraded to the orignal incompatible version.
474 --------------------------------------------------------------------------------
476 07-30-2014 Intel Corporation
478 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
479 for currently supported client and server kernel versions.
480 * Server known to build on patched kernels:
481 2.6.32-431.20.3.el6 (RHEL6.5)
482 3.0.101-0.31 (SLES11 SP3)
483 * Client known to build on unpatched kernels:
484 2.6.32-431.20.3.el6 (RHEL6.5)
485 3.0.101-0.31 (SLES11 SP3)
488 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
489 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
490 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
491 stack overflow. For more information, please refer to bugzilla 17630.
492 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
493 the re-exporting nodes to be mounted with "32bitapi" mount option
495 --------------------------------------------------------------------------------
497 10-31-2013 Intel Corporation
499 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
500 * Server builds for kernels:
501 2.6.32-358.18.1.el6 (RHEL6.4)
502 * Client builds for unpatched kernels:
503 2.6.32-358.18.1.el6 (RHEL6.4)
504 3.0.92-0.8 (SLES11 SP3)
506 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
507 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
508 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
509 stack overflow. For more information, please refer to bugzilla 17630.
510 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
511 the re-exporting nodes to be mounted with "32bitapi" mount option
513 --------------------------------------------------------------------------------
515 05-30-2013 Intel Corporation
517 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
518 * Server builds for kernels:
519 2.6.32-358.6.2.el6 (RHEL6.4)
520 * Client builds for unpatched kernels:
521 2.6.32-358.6.2.el6 (RHEL6.4)
522 3.0.74-0.6.6 (SLES11 SP2)
524 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
525 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
526 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
527 stack overflow. For more information, please refer to bugzilla 17630.
528 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
529 the re-exporting nodes to be mounted with "32bitapi" mount option
531 --------------------------------------------------------------------------------
533 9-30-2012 Whamcloud, Inc.
535 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
536 * Server support for kernels:
537 2.6.32-279.5.1.el6 (RHEL6)
538 * Client support for unpatched kernels:
539 2.6.32-279.5.1.el6 (RHEL6)
541 2.6.18-238.19.1.el5 (RHEL5)
542 2.6.16.60-0.69.1 (SLES10)
543 * Recommended e2fsprogs version: 1.41.90.wc4
544 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
545 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
546 stack overflow. For more information, please refer to bugzilla 17630.
547 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
548 the re-exporting nodes to be mounted with "32bitapi" mount option
550 Severity : enhancement
552 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
554 Severity : enhancement
556 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
558 Severity : enhancement
560 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
562 Severity : enhancement
564 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
566 Severity : enhancement
568 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
570 --------------------------------------------------------------------------------
572 04-30-2012 Whamcloud, Inc.
574 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
575 * Server support for kernels:
576 2.6.32-220.4.2.el6 (RHEL6)
577 * Client support for unpatched kernels:
578 2.6.32-220.4.2.el6 (RHEL6)
580 2.6.18-238.19.1.el5 (RHEL5)
581 2.6.16.60-0.69.1 (SLES10)
582 * Recommended e2fsprogs version: 1.41.90.wc4
583 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
584 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
585 stack overflow. For more information, please refer to bugzilla 17630.
586 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
587 the re-exporting nodes to be mounted with "32bitapi" mount option
589 Severity : enhancement
591 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
593 Severity : enhancement
595 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
597 Severity : enhancement
599 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
601 Severity : enhancement
603 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
605 --------------------------------------------------------------------------------
607 09-30-2011 Whamcloud, Inc.
609 * Server support for kernels:
610 2.6.18-238.19.1.el5 (RHEL5)
611 2.6.32-131.6.1.el6 (RHEL6)
612 * Client support for unpatched kernels:
613 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
614 2.6.18-238.19.1.el5 (RHEL5)
615 2.6.32-131.6.1.el6 (RHEL6)
616 2.6.29.4-167.fc11 (FC11)
617 2.6.16.60-0.69.1 (SLES10)
619 * Recommended e2fsprogs version: 1.41.90.wc3
620 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
621 Lustre filesystem with 4K stack may cause a stack overflow. For more
622 information, please refer to bugzilla 17630.
623 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
624 mounted with "32bitapi" mount option
626 Severity : enhancement
627 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
629 Severity : enhancement
630 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
632 Severity : enhancement
633 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
636 Frequency : only with SLES10
638 Description: Use OFED "KMP" provided by Novell
639 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
640 Lustre is now built against this package. That means you need to
641 install the ofed-kmp package from Novell for the patchless client.
642 Note that the ofed-kmp that Novell ships may not exactly match the
643 kernel version but should still be compatible.
645 Severity : enhancement
647 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
648 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
650 --------------------------------------------------------------------------------
652 2010-07-15 Oracle, Inc.
654 * Server support for kernels:
655 2.6.18-164.11.1.el5 (RHEL 5)
656 2.6.18-164.11.1.0.1.el5 (OEL 5)
657 * Client support for unpatched kernels:
658 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
659 2.6.18-164.11.1.el5 (RHEL 5),
660 2.6.18-164.11.1.0.1.el5 (OEL 5)
661 2.6.16.60-0.42.8 (SLES 10),
663 2.6.29.4-167.fc11 (FC11)
664 * Recommended e2fsprogs version: 1.41.10-sun2
666 Severity : enhancement
668 Description: Distribute lfs_migrate script with lustre
669 Details : The lfs_migrate mentioned in the manual is now included in the
670 lustre tree in lustre/scripts and will distributed in lustre RPMs.
672 Severity : enhancement
674 Description: Failover nid list added to import proc file.
676 Severity : enhancement
678 Description: "som_preview" mount option required to use SOM
679 Details : The clients will need to use som_preview mount option
683 Description: Lustre does not do 1MB IOs to HW RAID
684 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
686 Severity : enhancement
688 Description: lfs quota output is not very convenient for awk/sed-parsing
689 Details : Some positions in lfs quota output table could be empty or
690 non-empty which made it hard to parse it with scripts, now
691 a dash is put instead of space where there is not supposed
696 Description: Update to OFED 1.5.1
697 Details : OFED 1.5.1 will be used in environments that don't use
698 vendor supplied OFED.
702 Description: Remove lnet. prefix from lctl params display
703 Details : Also renamed lnet.memused to lnet_memused to
708 Description: Use conf_param -d to remove permanant settings.
712 Description: Add list_param -R
713 Details : lctl list_param now supports recursive mode.
716 Bugzilla : 21829, 22312
717 Description: Fix llobdstat.
718 Details : 1. Now updating the write_bytes counter
719 2. Port script fixes from 1.8.
723 Description: decrease the usage of memory on clients.
724 Details : 1. On clients, recycle dentries and inodes unused.
725 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
730 Description: llobdstat fix and enhancement
731 Details : add a counter to set a limit to how many samples will be returned
732 fix a wildcard in the path to limit to obdfilter stats only
736 Description: lfs check only allowed for root
737 Details : Code cleanups and added a regresion test of non-root use of
742 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
743 Details : Fix hr->hr_index overload.
747 Description: Fix FIEMAP functionality.
751 Description: FC11 patchless client support added
755 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
756 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
760 Description: New acc-sm script and various test-framework enhancments.
764 Description: Previously landed patch was reverted.
768 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
772 Description: Add ldiskfs maxdirsize mount option.
776 Description: Fix mkfs.lustre support for 16TB LUNs.
780 Description: Print mmp_check_interval
781 Details : print mmp_check_interval and make it possible to abort
782 mount operation in case it takes too long.
786 Description: Add cascading_rw to lustre mpi tests.
790 Description: Store SOM data in EA.
791 Details : Also several SOM related cleanups.
795 Description: Add refcount to md_open_data
799 Description: Use refcount on lov_request_set to prevent memory corruption.
803 Description: Statahead: use parent dir's i_mutex for synchronization.
807 Description: loadgen improvements.
811 Description: Enable peer health detection in ksocklnd and ko2iblnd.
815 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
819 Description: Fix race on obd_devcie:md_stats
823 Description: Fix SLES 11 client builds.
824 Details : Resolved uninitalized pointer gcc warnings.
828 Description: Allow non-root access for "lfs check".
829 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
833 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
835 Severity : enhancement
837 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
838 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
842 Description: Sanity test_27p fails.
843 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
844 Sanity tests 27 were cleaned up and improved.
848 Description: Unify req format on client/servers
849 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
850 buffers and swabbers approach..
855 Description: drop join file stuff.
857 Severity : enhancement
859 Description: Add 2.6.27 and SLES11 (patchless client) support
863 Description: Fix in ptlrpc_expire_one_request() to print the signed time
866 Severity : enhancement
868 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
871 Severity : enhancement
873 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
875 Severity : enhancement
877 Description: correctly handle big reply message.
878 Details : send LNet event if reply is bigger than buffer and adjust this buffer
883 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
885 Severity : enhancement
887 Description: Add support for OEL5.
891 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
893 Severity : enhancement
895 Description: "weak-modules" support
896 Details : Implement "weak-modules" support which enables kernel modules
897 to be used with any kernel that implements the same kABI. In
898 order to achieve this modules are now installed in
899 /lib/modules/$(uname -r)/updates/kernel on all distributions.
903 Description: client could not reconnect to OST because of an active request.
904 Details : The client now retries to reconnect to the same server, if a
905 connect request failed with EBUSY or -EAGAIN. Test case 35b
906 conf-sanity.sh has been created.
910 Description: Don't allow make backward step on assignin osc next id.
911 Details : race between allocation next id and ll_sync thread can be cause
912 of set wrong osc next id and can be kill valid ost objects.
916 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
920 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
924 Description: Error handling in osc_statfs_interpret() has been improved.
925 Details : Check in osc_statfs_interpret() for EBADR.
929 Description: Do not send statfs() requests to OSTs disabled by administrator.
930 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
932 Severity : enhancement
934 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
936 Severity : enhancement
938 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
940 Severity : enhancement
942 Description: Enhancement for lfs(1) command to use numeric uid/gid.
946 Description: Improved cleanup procedure for conf-sanity test_46a.
950 Description: Do not start bulk transfer, if client tried to reconnect.
951 Details : Do not start bulk transfer, if client tried to reconnect,
952 but failed due to active rpcs. Check for exp_abort_active_req
953 has been implemented.
957 Description: Prevent inconsistences between linux and lustre mount structures.
958 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
959 Make the sleep interruptible.
961 Severity : enhancement
963 Description: provide server to client comms path
964 Details : server to client communications path via new LDLM_SET_INFO rpc
968 Description: truncate starts GFP_FS allocation under transaction causing deadlock
969 Details : ldiskfs_truncate calls grab_cache_page which may start page
970 allocation under an open transaction. This may lead to
971 calling prune_icache with consequent lustre reentrance.
975 Description: send correctly lsm on open replay
976 Details : MDS is trust to LSM size on replay open, but client can set wrong size
981 Description: enable client interface failover
982 Details : When a child reconnects from another NID, properly update export
983 nid hash position and ldlm reverse import.
985 Severity : enhancement
987 Description: Communicate OST degraded/readonly state via statfs to MDS
988 Details : Flags in the statfs returned from OSTs indicate whether the
989 OST is in a degraded RAID state, or if the filesystem has
990 turned read-only after a filesystem error is detected.
995 Description: don't panic if EPROTO was hit when reading symlink
996 Details : correctly handling request reference in error cases.
998 Severity : enhancement
1000 Description: Add LustreNetLink, a kernel-userspace communcation path.
1002 Severity : enhancement
1004 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1009 Description: access to llog context before init.
1010 Details : move handling CATALOGS file at osc layer and forbid access to llog
1011 context before init.
1014 Frequency : rare, only when client is writing to OST on local node
1016 Description: Avoid deadlock for local client writes
1017 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1018 memory freeing context. This allows OST threads to set the
1019 PF_MEMALLOC flag on task structures in order to allocate memory
1020 from reserved pools and complete IO.
1021 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1022 so that the OST threads generate memory pressure and allow
1023 inactive pages to be reclaimed.
1025 Severity : enhancement
1027 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1032 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1033 Details : move ldlm namespace creation in setup phase to avoid grab
1034 _lprocfs_lock with cli_sem held.
1038 Description: Temporarily disable grant shrink.
1039 Details : Disable the feature for debugging.
1042 Frequency : only when formatting test filesystems
1044 Description: Unable to run several mkfs.lustre on loop devices at the same time
1045 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1046 formatting. The solution is to proper handle the error.
1048 Severity : enhancement
1050 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1052 Severity : enhancement
1054 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1059 Description: resolve race between obd_disconnect and class_disconnect_exports
1060 Details : if obd_disconnect will be called to already disconnected export he
1061 forget release one reference and osc module can't unloaded.
1063 Severity : enhancement
1065 Description: Allow tuning service thread via /proc
1066 Details : For each service a new
1067 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1068 created that can be used to set min/max thread counts, and get the
1069 current number of running threads.
1074 Descriptoin: don't return error if have partially created objects for file.
1075 Details : lov_update_create_set uses set->set_success as index for created
1076 objects, so if some requests will be failed, they will have hole
1077 at end of array and qos_shrink_lsm for allocate correct lsm.
1079 Severity : enhancement
1081 Description: Update OFED support to 1.4.1
1086 Description: Reduce small size read RPC
1087 Details : Set read-ahead limit for every file and only do read-ahead when
1088 available read-ahead pages are bigger than 1M to avoid small size
1091 Severity : enhancement
1093 Description: move AT tunable parameters for more consistent usage
1094 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1096 Severity : enhancement
1098 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1099 Details : allow skip disconnected ost for send statfs request and hide error
1105 Description: fix racy locking of mballoc block bitmaps causing BUG
1106 Details : The locking of the mballoc buddy bitmap and the in-memory
1107 block bitmap was using two different spin locks in some
1108 cases. This made it possible to incorrectly access the
1109 mballoc bitmap while another process was modifying it,
1110 causing a sanity assertion to fail. While no on-disk corruption
1111 was reported, there was some risk of this happening.
1114 Frequency : rare, on llog test 6
1116 Descriptoin: don't allow connect to already connected import
1117 Details : allowing connect to already connected import is hide connecting problem.
1122 Descriptoin: allow kill process which wait statahead result
1123 Details : for some reasons 'ls' can stick in waiting result from statahead,
1124 in this case need way for kill this process.
1126 Severity : enhancement
1128 Description: Add state history info file, enhance import info file
1129 Details : Track import connection state changes in a new osc/mdc proc file;
1130 add overview-type data to the osc/mdc import proc file.
1132 Severity : enhancement
1134 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1136 Severity : enhancement
1138 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1143 Descriptoin: OST grant too much space to client even there are not enough space.
1144 Details : Client will shrink its grant cache to OST if there are no write
1145 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1146 this grant cache if there are already not enough avaible space
1147 (left_space < total_clients * 32M).
1150 Frequency : start MDS on uncleanly shutdowned MDS device
1152 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1153 Details : stay in waiting mds<>ost recovery finished produce random bugs
1154 due race between two ll_sync thread for one lov target. send
1155 ACTIVATE event only if connect realy finished and import have
1159 Frequency : rare, connect and disconnect target at same time
1161 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1162 Details : don't call obd_disconnect under lov_lock. this long time
1163 operation and can block ptlrpcd which answer to connect request.
1168 Descriptoin: don't lose wakeup for imp_recovery_waitq
1169 Details : recover_import_no_retry or invalidate_import and import_close can
1170 both sleep on imp_recovery_waitq, but we was send only one wakeup
1174 Frequency : always with long access acl
1176 Descriptoin: mds can't pack reply with long acl.
1177 Details : mds don't control size of acl but they limited by reint/getattr
1180 Severity : enhancement
1182 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1184 Severity : enhancement
1186 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1189 Frequency : start MDS on uncleanly shutdowned MDS device
1191 Descriptoin: aborting recovery hang on MDS
1192 Details : don't throttle destroy RPCs for the MDT.
1195 Frequency : on remount
1197 Description: external journal device not working after the remount
1198 Details : clear dev_rdonly flag for external journal devices in
1204 Description: shutdown vs evict race
1205 Details : client_disconnect_export vs connect request race.
1206 if client will evicted at this time - we start invalidate
1207 thread without referece to import and import can be freed
1211 Frequency : rare, need acl's on inode.
1213 Description: client can't handle ost additional correctly
1214 Details : if ost was added after client connected to mds client can have
1215 hit lnet_try_match_md ... to big messages to wide striped files.
1216 in this case need teach client to handle config events about add
1217 lov target and update client max ea size at that event.
1219 Severity : enhancement
1221 Description: Changelogs
1222 Details : Changelogs are a lightweight mechanism to track filesystem
1223 metadata and namespace changes. The changelog is recorded
1224 permanently on the MDTs, and is periodically "consumed" / purged
1225 when records are no longer needed.
1227 Severity : enhancement
1229 Description: compact fld format with extents
1230 Details : Store range of seq rather than every seq in FLD. Seq
1231 controller update FLD rather than clients. In Case of CMD, mdt0
1232 has FLD, all other metadata server act as non persistent proxy
1233 for FLD queries and cache fld entries in fld cache.
1238 Description: don't skip ost target if they assigned to file
1239 Details : Drop slow OSCs if we can, but not for requested start idx.
1240 This means "if OSC is slow and it is not the requested
1241 start OST, then it can be skipped, otherwise skip it only
1242 if it is inactive/recovering/out-of-space.
1246 Description: more cleanup in mds_lov
1247 Details : not send LOV EA under replay, we can't know about they size at this
1248 time. Don't allow client connect to mds before any ost connected,
1249 for avoid problems with LOV EA size and returning EIO to client.
1251 Severity : enhancement
1253 Description: Interoperability at server side (Disk interoperability)
1255 Severity : enhancement
1257 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1259 Severity : enhancement
1261 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1263 Severity : enhancement
1265 Description: New client IO stack (CLIO).
1267 Severity : enhancement
1269 Description: Commit on sharing. Eliminate inter-client dependencies between
1270 uncommitted transactions by doing transaction commits.
1271 Thereby clients may recovery independently.
1274 Frequency : Create a symlink file with a very long name
1276 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1277 Details : If there is no extra space in the request for early cancels,
1278 ldlm_req_handles_avail() returns 0 instead of a negative value.
1280 Severity : enhancement
1282 Description: Add /proc entry for import status
1283 Details : The mdc, osc, and mgc import directories now have
1284 an import directory that contains useful import data for debugging
1285 connection problems.
1287 Severity : enhancement
1289 Description: Re-disable certain /proc logging
1290 Details : Enable and disable client's offset_stats, extents_stats and
1291 extents_stats_per_process stats logging on the fly.
1294 Frequency : Only on FC kernels 2.6.22+
1296 Description: oops in statahead
1297 Details : Do not drop reference count for the dentry from VFS when lookup,
1298 VFS will do that by itself.
1300 Severity : enhancement
1302 Description: Generic /proc file permissions
1303 Details : Set /Proc file permissions in a more generic way to enable non-
1304 root users operate on some /proc files.
1308 Description: Hitting mdc_commit_close() ASSERTION
1309 Details : Properly handle request reference release in
1310 ll_release_openhandle().
1314 Description: quota recovery deadlock during mds failover
1315 Details : This patch includes att18982, att18236, att18237 in bz14840.
1317 1. fix osts hang when mds does failover with quotaon
1318 2. prevent watchdog storm when osts threads wait for the
1323 Frequency : only patchless client
1324 Description: add workaround for race between add/remove dentry from hash
1326 Severity : enhancement
1328 Description: Allow OST glimpses to return PW locks
1332 Description: LBUG when llog conf file is full
1333 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1337 Description: Prevent import from entering FULL state when server in recovery
1341 Description: service mount cannot take device name with ":"
1342 Details : Only when device name contains ":/" will mount treat it as
1348 Description: replace ptlrpcd with the statahead thread to interpret the async
1349 statahead RPC callback
1353 Frequency : on recovery
1354 Description: I/O failures after umount during fail back
1355 Details : if client reconnected to restarted server we need join to recovery
1356 instead of find server handler is changed and process self
1357 eviction with cancel all locks.
1359 Severity : enhancement
1361 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1363 Severity : enhancement
1365 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1367 Severity : enhancement
1369 Description: Upcall on Lustre log has been dumped
1370 Details : Allow for a user mode script to be called once a Lustre log has
1371 been dumped. It passes the filename of the dumped log to the
1372 script, the location of the script can be specified via
1373 /proc/sys/lnet/debug_log_upcall.
1378 Description: avoid idr_remove called for id which is not allocated.
1379 Details : Move assigment s_dev for clustered nfs to end of initialization,
1380 to avoid problem with error handling.
1385 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1386 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1387 into CONN_USED_HASH and this prodice warning when put connection
1388 again in unused hash.
1391 Severity : enhancement
1393 Description: File striping can now be set to use an arbitrary pool of OSTs.
1395 Severity : enhancement
1397 Description: Export bytes_read/bytes_write count on OSC/OST.
1401 Description: Early reply size mismatch, MGC loses connection
1402 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1403 the connect flags are properly negotiated.
1408 Description: quotas are not honored with O_DIRECT
1409 Details : all writes with the flag O_DIRECT will use grants which leads to
1410 this problem. Now using OBD_BRW_SYNC to guard this.
1414 Description: add quota statistics
1415 Details : 1. sort out quota proc entries and proc code.
1416 2. add quota statistics
1418 Severity : enhancement
1420 Description: enable quota support for HEAD.
1424 Description: Properly propagate oinfo flags from lov to osc for statfs
1425 Details : restore missing copy oi_flags to lov requests.
1427 Severity : enhancement
1429 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1430 : l_getgroups(8), lst(8), routerstat(8)
1431 Details : included man pages for llobdstat(8), llstat(8),
1432 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1434 Severity : enhancement
1436 Description: configure's --enable-quota should check the
1437 : kernel .config for CONFIG_QUOTA
1438 Details : configure is terminated if --enable-quota is passed but
1439 : no quota support is in kernel
1443 Description: Remove portals compatibility
1444 Details : Remove portals compatibility, not interoperable with releases
1449 Description: Resolve device initialization race
1450 Details : Prevent proc handler from accessing devices added to the
1451 obd_devs array but yet be intialized.
1453 Severity : enhancement
1455 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1457 Severity : enhancement
1459 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1464 Description: Using wrong pointer in osc_brw_prep_request
1465 Details : Access to array[-1] can produce panic if kernel compiled with
1466 CONFIG_PAGE_ALLOC enabled
1470 Description: Client runs out of low memory
1471 Details : Consider only lowmem when counting initial number of llap pages
1475 Description: *optional* service tags registration
1476 Details : if the "service tags" package is installed on a Lustre node
1477 When the filesystem is mounted, a local-node service tag will
1478 be created. See http://inventory.sun.com/ for more information
1479 about the Service Tags asset management system.
1483 Description: Kernel BUG tries to release flock
1484 Details : Lustre does not destroy flock lock before last reference goes
1485 away. So always drop flock locks when client is evicted and
1486 perform unlock regardless of successfulness of speaking to MDS.
1490 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1494 Description: sanity test 65a fails if stripecount of -1 is set
1495 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1500 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1501 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1504 Severity : enhancement
1506 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1510 Description: do not process already freed flock
1511 Details : flock can possibly be freed by another thread before it reaches
1512 to ldlm_flock_completion_ast.
1516 Description: LBUG during stress test
1517 Details : Need properly lock accesses the flock deadlock detection list.
1521 Description: oops in page fault handler
1522 Details : kernel page fault handler can return two special 'pages' in error
1523 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1527 Description: timeout with invalidate import.
1528 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1529 should be handled by ptlrpcd. This produce long age waiting and
1530 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1532 Severity : enhancement
1534 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1538 Description: enable MGS and MDT services start separately
1539 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1540 a co-located MGS without starting the MGS, which is a complement
1541 to 'nosvc' mount option.
1545 Frequency : after recovery
1546 Description: precreate to many object's after del orphan.
1547 Details : del orphan st in oscc last_id == next_id and this triger growing
1548 count of precreated objects. Set flag LOW to skip increase count
1549 of precreated objects.
1553 Frequency : rare, on clear nid stats
1554 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1555 Details : when clean nid stats sometimes try destroy live entry,
1556 and this produce panic in free.
1560 Description: Stack overflow during MDS log replay
1561 ease stack pressure by using a thread dealing llog_process.
1565 Description: wait until IO finished before start new when do lock cancel.
1566 Details : VM protocol want old IO finished before start new, in this case
1567 need wait until PG_writeback is cleared until check dirty flag and
1568 call writepages in lock cancel callback.
1570 Severity : enhancement
1572 Description: using special macro for print time and cleanup in includes.
1576 Description: mds_mfd_close() ASSERTION(rc == 0)
1577 Details : In mds_mfd_close(), we need protect inode's writecount change
1578 within its orphan write semaphore to prevent possible races.
1582 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1586 Frequency : rare, on shutdown ost
1587 Description: don't hit live lock with umount ost.
1588 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1589 use shrink_dcache_sb instead.
1593 Description: don't panic with use echo client
1594 Details : echo client pass NULL as client nid pointer and this produce null
1595 pointer dereference.
1599 Description: fix build on ppc32
1600 Details : compile code with -m64 flag produce wrong object file for ppc32.
1604 Description: add message levels for liblustreapi
1608 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1609 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1610 will result in an -EAGAIN and in the exit call path we call
1611 alloc_rr() which will with increasing aggressiveness attempt to
1612 aquire precreated objects on the minimum number of required OSCs.
1616 Description: Use old size assignment to avoid deadlock
1617 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1618 the scheduling while holding a spinlock. We do not need locking
1619 for size in ll_update_inode() because size is only updated from
1620 the MDS for directories or files without objects, so there is no
1621 other place to do the update, and concurrent access to such inodes
1622 are protected by the inode lock.
1626 Description: resolve "_IOWR redefined" build error on SLES10
1630 Description: dump the memory debugging after all modules are unloaded to
1631 suppress false negative in conf_sanity test 39
1633 Severity : enhancement
1635 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1640 Description: fix SLES kernel versioning
1641 Details : the kernel version for our SLES 10 kernel did not include a "-"
1642 before the "smp" at the end. while this was not a problem in
1643 general, it did mean that software trying to use the kernel
1644 version to try to detect a vendor specific kernel would fail.
1645 this was most evident by the OFED build scripts.
1649 Description: Don't update lov_desc members until making sure they are valid
1650 Details : When updating lov_desc members via proc fs, need fix their
1651 validities before doing the real update.
1655 Description: don't put request into delay list while invalidate in flight.
1656 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1657 import in flight. this produce timeout for invalidate and sometimes
1658 can cause stale data.
1662 Frequency : on ppc only
1663 Description: not convert ost objects for directory because it's not exist.
1664 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1668 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1670 Severity : enhancement
1672 Description: Update to support 2.6.22.14 vanilla kernel.
1676 Frequency : rare, on recovery
1677 Description: read procfs can produce deadlock in some situation
1678 Details : Holding lprocfs lock which send rpc can produce block for destroy
1679 obd objects and this also block reconnect with -EALREADY. This
1680 isn't fix all lprocfs bugs - but make it rare.
1682 Severity : enhancement
1684 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1687 Frequency : frequent on X2 node
1689 Description: mdc_set_open_replay_data LBUG
1690 Details : Set replay data for requests that are eligible for replay.
1694 Description: lustre_mgs: operation 101 on unconnected MGS
1695 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1696 MGC, and late on MGC cannot successfully connect to MGS and a lot
1697 of the error messages complaining that MGS is not connected.
1701 Frequency : on network error
1702 Description: panic with double free request if network error
1703 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1704 but it's true only for synchronus enqueue, for async enqueue
1705 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1708 Severity : enhancement
1710 Description: client-side metadata stat-ahead during readdir(directory readahead)
1711 Details : perform client-side metadata stat-ahead when the client detects
1712 readdir and sequential stat of dir entries therein
1715 Frequency : on start mds
1717 Description: Implement get_info(last_id) in obdfilter.
1720 Frequency : occasional
1722 Description: Correctly check stale fid, not start epoch if ost not support SOM
1723 Details : open with flag O_CREATE need set old fid in op_fid3 because
1724 op_fid2 was overwritten with new generated fid, but mds can answer
1725 with one of these two fids and both is not stale. Setattr
1726 incorrectly started an epoch and assume will be called
1727 done_writing, but without SOM done_writing ever being called.
1730 Frequency : rare, depends on device drivers and load
1732 Description: MDS or OSS nodes crash due to stack overflow
1733 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1734 In some cases, in conjunction with device drivers that use a lot
1735 of stack the MDS (or possibly OSS) service threads could overflow
1736 the stack. One change which was identified to consume additional
1737 stack has been reworked to avoid the extra stack usage.
1740 Frequency : occasional
1742 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1743 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1744 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1745 oscc_has_objects and see OSC has no objests and call
1746 oscc_internal_create to resend create request.
1748 Severity : enhancement
1750 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1752 Severity : enhancement
1754 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1757 Frequency : very rare
1759 Description: Fix replay if there is an un-replied request and open
1760 Details : In some cases, older replay request will revert the
1761 mcd->mcd_last_xid on MDS which is used to record the client's
1762 latest sent request.
1764 Severity : enhancement
1766 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1768 Severity : enhancement
1770 Description: Add rhel5 support to HEAD.
1772 Serverity : enhancement
1774 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1779 Description: Don't allow skipping OSTs if index has been specified.
1780 Details : Don't allow skipping OSTs if index has been specified, make
1781 locking in internal create lots better.
1785 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1786 Details : access to bitfield in structure is always rounded to long
1787 and this produce problem with not atomic change any bit.
1791 Description: Lustre make rpms failed.
1792 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1793 builds Lustre rpms from tarball.
1796 Frequency : rare on shutdown ost
1798 Description: If llog cancel was not send before clean_exports phase, this can
1799 produce deadlock in llog code.
1800 Details : If llog thread has last reference to obd and call class_import_put
1801 this produce deadlock because llog_cleanup_commit_master wait when
1802 last llog_commit_thread exited, but this never success because was
1803 called from llog_commit_thread.
1807 Description: allow userland application know is lost one of stripes.
1808 Details : fill lvb_blocks with error code on ost and return it to
1809 application if error flag found.
1813 Description: NULL lov_tgts causing MDS oops
1814 Details : more safe checks for NULL lov_tgts for avoid oops.
1816 Severity : enhancement
1818 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1822 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1823 Details : If an OST is down the MDS will hang indefinitely in
1824 obd_statfs_rqset() waiting for the statfs data. While for
1825 MDS QOS usage of statfs, it should not stuck in waiting.
1827 Severity : enhancement
1829 Description: remote_acl support
1830 Details : Support ACL-based permission check for remote user.
1831 Support setfacl/getfacl for remote user with the utils
1832 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1833 the system "{s,g}etfacl" utils.
1835 Severity : enhancement
1837 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1839 Severity : enhancement
1841 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1845 Description: make mgs_setparam() handle fsname containing dash
1846 Details : fsname containing a dash does not work with lctl conf_param
1848 Severity : enhancement
1850 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1852 Severity : enhancement
1854 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1858 Frequency : rare, at shutdown
1859 Description: access already free / zero obd_namespace.
1860 Details : if client_disconnect_export was called without force flag set,
1861 and exist connect request in flight, this can produce access to
1862 NULL pointer (or already free pointer) when connect_interpret
1863 store ocd flags in obd_namespace.
1867 Frequency : only at startup
1868 Description: not alloc memory with spinlock held.
1869 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1870 if any spinlock held.
1872 Severity : enhancement
1874 Description: make lustre randomly fail allocating memory
1875 Details : Make lustre randomly failed allocating memory for testing purpose.
1877 Severity : enhancement
1879 Description: lost problems with lov objid file
1880 Details : Fixes some scability and access to not inited memory problems
1881 in work with lov objdid file.
1886 Description: lfs find does not continue on file error
1887 Details : Continue other files processing when a file/dir is absent.
1891 Description: Inconsistent usage of lustre_pack_reply()
1892 Details : Standardize the usage of lustre_pack_reply() such that it
1893 always generate a CERROR on failure.
1896 Frequency : occasional
1898 Description: MDS hang or stay in waiting lock
1899 Details : If client receive lock with CBPENDING flag ldlm need send lock
1900 cancel as separate rpc, to avoid situation when cancel request
1901 can't processed due all i/o threads stay in wait lock.
1905 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1909 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1913 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1917 Description: Update to RHEL4 latest kernel.
1919 Severity : enhancement
1921 Description: Build SLES10 patchless client fails
1922 Details : The configure was broken by run ./configure with
1923 --with-linux-obj=.... argument for patchless client. When the
1924 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1925 can't search header adequately. Use absolute path such as
1926 -I($LINUX)/include instead.
1930 Description: interrupt oig_wait produce painc on resend.
1931 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1932 private set, and this produce situation when rq_ptlrpcd_data not
1933 copyed to new allocated request and triggered LBUG on assert
1934 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1935 ptlrpcd set if request is changed and can be safety changed to use
1940 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1941 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1942 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1943 these macros and replace them with their difinitions instead.
1947 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1948 Details : Update which_patch & target file for SLES10 latest kernel.
1951 Description: Remove client patches from SLES 10 kernel.
1952 Details : This causes SLES 10 clients to behave as patchless clients
1953 even on a Lustre-patched (server) kernel.
1955 Severity : enhancement
1957 Description: self-adjustable client's lru lists
1958 Details : use adaptive algorithm for managing client cached locks lru
1959 lists according to current server load, other client's work
1960 pattern, memory activities, etc. Both, server and client
1961 side namespaces provide number of proc tunables for controlling
1964 Severity : enhancement
1966 Description: light-weight GSS support
1967 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1968 and reduce performance overhead.
1970 Severity : enhancement
1972 Description: Linux keyring support
1973 Details : Support using service of Linux keyring for Lustre GSS internal
1974 context refresh/cache mechanism.
1978 Description: Fix errors in lfs documentation
1979 Details : Fixes man pages
1983 Description: don't use GFP_* in generic Lustre code.
1984 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1988 Description: obdclass is limited by single OBD_ALLOC(idarray)
1989 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1993 Description: warnings with build patchless client with vanila 2.6.19 and up
1994 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1995 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1999 Frequency : Only for SLES
2000 Description: sanity_quota fail test_1
2001 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2002 file, which makes TSTID[2] inunique.
2006 Frequency : Only for SLES9
2007 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2008 Details : Update target/ChangeLog/which_patch .
2012 Frequency : Only for test_10 in sanity-quota.sh
2013 Description: a bug in quota test code
2014 Details : checking a test flag in a wrong place
2018 Frequency : Only for test_13 in sanity-quota.sh
2019 Description: a bug in quota test script
2020 Details : shouldn't check overfull allocated quota
2024 Description: jbd statistics
2025 Details : Port older jbd statistics patch for sles10
2027 Severity : enhancement
2029 Description: Add printing inode info into message about error in writepage.
2034 Description: reply_lock_interpret crash due to race with it and lock cancel.
2035 Details : Do not replay locks that are being cancelled. Do not reference
2036 locks by their address during replay, just by their handle.
2040 Frequency : When flocks are used.
2041 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2042 Details : Flock locks might destroy just granted lock if it could be merged
2043 with another existing flock, this is done in completion handler,
2044 so teach ldlm_cli_enquque_fini that this is a valid case for
2050 Description: Oops in read and write path when failing to allocate lock.
2051 Details : Check if lock allocation failed and return error back.
2055 Description: lstripe command fails for valid OST index
2056 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2057 instead of lov->desc.ld_active_tgt_count.
2061 Description: sanity.sh failed test 103
2062 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2067 Description: df doesn't work properly if diskfs blocksize != 4K
2068 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2072 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2073 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2075 Severity : enhancement
2077 Description: Adaptive timeouts
2078 Details : RPC timeouts adapt to changing server load and network
2079 conditions to reduce resend attempts and improve recovery time.
2083 Description: llapi_file_create() does not allow some changes
2084 Details : add llapi_file_open() that allows specifying the mode and
2085 open flags, and also returns an open file handle.
2089 Description: merge and cleanup kernel patches.
2090 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2094 Description: Add journal checksum support.(Kernel part)
2095 Details : The journal checksum feature adds two new flags i.e
2096 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2097 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2098 indicates that the commit block contains the checksum for
2099 the blocks described by the descriptor blocks. Now commit
2100 record can be sent to disk without waiting for descriptor
2101 blocks to be written to disk. This behavior is controlled
2102 using JBD2_FEATURE_ASYNC_COMMIT flag.
2106 Description: OSS needs mutliple precreate threads
2107 Details : Add ability to start more than one create thread per OSS.
2111 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2112 Details : execute lfs setstripe on client
2116 Description: mds_obd_create error creating tmp object
2117 Details : When the user sets quota on root, llog will be affected and can't
2118 create files and write files.
2121 Frequency : Always on ia64 patchless client, and possibly others.
2123 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2124 Details : This allows the patchless client to be loaded on architectures
2125 without this export.
2129 Description: RedHat Update kernel for RHEL5
2130 Details : Add the kernel config file for RHEL5.
2134 Description: RedHat Update kernel for RHEL5
2135 Details : Modify the kernel config file more closer RHEL5.
2139 Description: Build failure against Centos5 (RHEL5)
2140 Details : Define PAGE_SIZE when it isn't present.
2144 Description: test_7 on sanity-quota.sh can't be used on separate machines
2145 Details : add facet to handle it
2149 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2150 Details : with NFS, the anon dentry's parent was set to itself in
2151 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2152 corresponding dentry other than use rec->ur_name.
2154 Severity : enhancement
2156 Description: lfs setstripe enhancement
2157 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2161 Description: enable data checksumming by default
2162 Details : enable checksum by default, allow --disable-checksum
2163 configure option and "-o nochecksum" mount option. Checksums
2164 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2168 Description: lustre support for RHEL5
2169 Details : Add support for RHEL5.
2173 Description: Client eviction due to failover config
2174 Details : after a connection loss, the lustre client should attempt to
2175 reconnect to the last active server first before trying the
2176 other potential connections.
2180 Description: when mds and osts use different quota unit(32bit and 64bit),
2181 quota will be released repeatly.
2182 Details : void sending multiple quota reqs to mds, which will keep the
2183 status between the reqs.
2187 Description: osts not allocated evenly to files
2188 Details : change the condition to increase offset_idx
2191 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2193 Description: Data corruption for OSTs that are formatted larger than 2TB
2195 Details : When generating the bio request for lustre file writes the
2196 sector number would overflow a temporary variable before being
2197 used for the IO. The data reads correctly from Lustre (which
2198 will overflow in a similar manner) but other file data or
2199 filesystem metadata may be corrupted in some cases.
2203 Description: Tune the kernel for good SCSI performance.
2204 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2205 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2210 Description: rewrite ext2-derived code in obdclass/uuid.c
2211 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2212 from scratch preserving functionality.
2216 Description: "lfs find -obd UUID" prints directories
2217 Details : "lfs find -obd UUID" will return all directory names instead
2218 of just file names. It is incorrect because the directories
2219 do not reside on the OSTs.
2223 Description: use i_size_read and i_size_write in 2.6 port
2224 Details : replace inode->i_size access with i_size_read/write()
2227 Frequency : when removing large files
2229 Description: scheduling issue during removal of large Lustre files
2230 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2231 It causes scheduling issues when removing large files (17TB in the
2235 Frequency : only with liblustre clients on XT3
2237 Description: evictions taking too long
2238 Details : allow llrd to evict clients directly on OSTs
2241 Frequency : only on ppc
2243 Description: /proc/fs/lustre/devices broken on ppc
2244 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2245 We should make sure the type of 'index' is loff_t and then cast
2246 explicitly as needed below. Do not assign an explicitly cast
2250 Frequency : only for rhel5
2252 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2253 Details : Modify the target file & which_kernel.
2255 Severity : enhancement
2257 Description: omit set fsid for export NFS
2258 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2259 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2263 Frequency : Always, for kernels after 2.6.16
2264 Description: Fix warning idr_remove called for id=.. which is not allocated.
2265 Details : Last kernels save old s_dev before kill super and not allow
2266 to restore from callback - restore it before call kill_anon_super.
2270 Description: buffer overruns could theoretically occur
2271 Details : llapi_semantic_traverse() modifies the "path" argument by
2272 appending values to the end of the origin string, and a
2273 overrun may occur. Adding buffer overrun check in liblustreapi.
2277 Description: Fix error on 'ls .' at the top of the Lustre mount.
2278 Details : Don't revalidate dentry if it is a root dentry.
2282 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2283 Details : Modify vm-tunables-rhel4.patch.
2287 Description: Kernel config for 2.6.18-vanilla.
2288 Details : Modify targets/2.6-vanilla.target.in.
2289 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2290 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2291 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2292 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2296 Description: improve handling recoverable errors
2297 Details : if request processig with error which can be recoverable on server
2298 request should be resend, otherwise page released from cache and
2303 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2304 Details : Modify target file & which_patch.
2305 A flaw was found in the IA32 system call emulation provided
2306 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2307 value could be stored in the %RAX register, which could trigger an
2308 out-of-bounds system call table access. An untrusted local user
2309 could exploit this flaw to run code in the kernel
2310 (ie a root privilege escalation). (CVE-2007-4573).
2314 Description: O_DIRECT bypasses client statistics.
2315 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2316 read_ahead_stats were updated. copied the stats section from
2317 osc_send_oap_rpc() into async_internal().
2321 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2325 Description: change order of libsysio includes
2326 Details : '#include sysio.h' should always come before '#include xtio.h'
2330 Description: handle "serious error: objid * already exists" more gracefully
2331 Details : If LAST_ID value on disk is smaller than the objects existing in
2332 the O/0/d* directories, it indicates disk corruption and causes an
2333 LBUG(). If the object is 0-length, then we should use the existing
2334 object. This will help to avoid a full fsck in most cases.
2336 Severity : enhancement
2338 Description: adapt the lustre_config script to support the upgrade case
2339 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2340 server targets to 1.6 in parallel.
2344 Description: To avoid grant space > avaible space when the disk is almost
2345 full. Without this patch you might see the error "grant XXXX >
2346 available" or some LBUG about grant, when the disk is almost
2348 Details : In filter_check_grant, for non_grant cache write, we should
2349 check the left space by if (*left > ungranted + bytes), instead
2350 of (*left > ungranted), because only we are sure the left space
2351 is enough for another "bytes", then the ungrant space should be
2352 increase. In client, we should update cl_avail_grant only there
2353 is OBD_MD_FLGRANT in the reply.
2357 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2358 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2359 described in CVE-2007-4573. This problem could allow untrusted
2360 local users to gain root access.
2363 Frequency : when using O_DIRECT and quotas
2365 Description: Incorrect file ownership on O_DIRECT output files
2366 Details : block usage reported by 'lfs quota' does not take into account
2367 files that have been written with O_DIRECT.
2372 Description: touch file failed when fs is not full
2373 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2374 otherwise we can get ENOSP while fs is not full.
2378 Description: parallel lock callbacks
2379 Details : Instead of sending blocking and completion callbacks as separated
2380 requests, adding them to a set and sending in parallel.
2383 Frequency : only for Cray XT3
2384 Bugzilla : 12829/13455
2385 Description: Changing primary group doesn't change the group lustre assigns to
2387 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2388 overridden with the primary group provided by the group upcall,
2389 whereas the supplied fsgid can be trusted if it is in the list of
2390 supplementary groups returned by the group upcall.
2392 Severity : enhancement
2394 Description: Allow masking D_WARNING, D_ERROR messages from console
2395 Details : Console messages can now be disabled via lnet.printk.
2399 Description: User code with malformed file open parameter crashes client node
2400 Details : Before packing join_file req, all the related reference should be
2401 checked carefully in case some malformed flags cause fake
2402 join_file req on client.
2406 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2407 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2408 mode is equal to the granted mode. However, it should also check
2409 that LDLM_FL_CP_REQD is not set.
2413 Description: Heavy nfs access might result in deadlocks
2414 Details : After ELC code landed, it is now improper to enqueue any mds
2415 locks under och_sem, because enqueue might want to decide to
2416 cancel open locks for same inode we are holding och_sem for.
2420 Description: Client eviction while running blogbench
2421 Details : A lot of unlink operations with concurrent I/O can lead to a
2422 deadlock causing evictions. To address the problem, the number of
2423 oustanding OST_DESTROY requests is now throttled to
2424 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2425 callbacks are processed in priority.
2429 Description: enable ACLs on MDS by default
2430 Details : ACLs must be enabled on MDS by default.
2433 Frequency : PPC/PPC64 only
2435 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2436 Details : fix duplicated definitions between asm-ppc64/types.h and
2437 lustre_types.h on PPC.
2440 Frequency : PPC/PPC64 only
2442 Description: asm-ppc/segment.h does not exist
2443 Details : fix compile issue on PPC.
2447 Description: better handle error messages in extents code
2450 Frequency : RHEL4 only
2452 Description: mkfs is very slow on IA64/RHEL4
2453 Details : A performance regression has been discovered in the MPT Fusion
2454 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2455 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2456 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2458 Severity : enhancement
2460 Description: SNMP support enhancement
2461 Details : Adding total number of sampled request for an MDS node in snmp
2464 Severity : enhancement
2466 Description: Optimize ldlm waiting list processing for PR extent locks
2467 Details : When processing waiting list for read extent lock and meeting read
2468 lock that is same or wider to it that is not contended, skip
2469 processing rest of the list and immediatelly return current
2470 status of conflictness, since we are guaranteed there are no
2471 conflicting locks in the rest of the list.
2475 Description: Time out and refuse to reconnect
2476 Details : When the failover node is the primary node, it is possible
2477 to have two identical connections in imp_conn_list. We must
2478 compare not conn's pointers but NIDs, otherwise we can defeat
2479 connection throttling.
2483 Description: port llog fixes from b1_6 into HEAD
2484 Details : Port llog reference couting and some llog cleanups from b1_6
2485 (bug 10800) into HEAD, for protect from panic and access to already
2486 free llog structures.
2490 Description: Detect stride IO mode in read-ahead
2491 Details : When a client does stride read, read-ahead should detect that and
2492 read-ahead pages according to the detected stride pattern.
2496 Description: data checksumming impacts single node performance
2497 Details : add support for several checksum algorithm. Currently, only CRC32
2498 and Adler-32 are supported. The checksum type can be changed on
2499 the fly via /proc/fs/lustre/osc/*/checksum_type.
2503 Description: use adler32 for page checksums
2504 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2509 Description: build for x2 fails
2510 Details : fix compile issue on Cray systems.
2514 Description: Properly match for duplicate locks
2515 Details : Due to different lock order from skiplists code, we need to
2516 traverse entire list for now
2519 Frequency : only on PPC/SLES10
2521 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2522 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2523 be defined. Add a hack in mkfs_lustre.c to work around this compile
2528 Description: LASSERT on MDS when client holding flock lock dies
2529 Details : ldlm pool logic depends on number of granted locks equal to
2530 number of released locks which is not true for flock locks, so
2531 just exclude such locks from consideration.
2535 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2536 Details : Use fsfilt_sync() for both the whole filesystem sync and
2537 individual file sync to eliminate dangerous inode locking
2538 with I_LOCK that can lead to a deadlock.
2542 Description: performance in 1.6.3
2543 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2544 sure that 1MB requests are not fragmented by the block layer.
2546 Severity : enhancement
2548 Description: organize the server-side client stats on per-nid basis
2549 Details : Change the structure of stats under obdfilter and mds to
2557 The "uuid"s file would list the uuids of _active_ exports.
2558 And the clear entry is to clear all stats and stale nids.
2560 Severity : enhancement
2562 Description: eliminate client locks in face of contention
2563 Details : file contention detection and lockless i/o implementation
2564 for contended files.
2568 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2571 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2573 Description: MDS may lose file striping (and hence file data) in some cases
2574 Details : If there are additional extended attributes stored on the MDS,
2575 in particular ACLs, SELinux, or user attributes (if user_xattr
2576 is specified for the client mount options) then there is a risk
2577 of attribute loss. Additionally, the Lustre file striping
2578 needs to be larger than default (e.g. striped over all OSTs),
2579 and an additional attribute must be stored initially in the
2580 inode and then increase in size enough to be moved to the
2581 external attribute block (e.g. ACL growing in size) for file
2586 Description: skiplist implementation simplification
2587 Details : skiplists are used to group compatible locks on granted list
2588 that was implemented as tracking first and last lock of each
2589 lock group the patch changes that to using doubly linked lists
2593 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2594 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2595 because LASSERT on both the data supplied by a client, and the
2596 data on disk is dangerous and incorrect.
2598 Severity : enhancement
2600 Description: Slow truncate/writes to huge files at high offsets.
2601 Details : Directly associate cached pages to lock that protect those pages,
2602 this allows us to quickly find what pages to write and remove
2603 once lock callback is received.
2607 Description: more ldlm soft lockups
2608 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2609 starve other threads from the resource lock for a long time in
2610 case of long waiting queue, so change the debug level from
2611 D_OTHER to the less frequently used D_INFO.
2613 Severity : enhancement
2615 Description: add -gid, -group, -uid, -user options to lfs find
2619 Description: Hung threads in invalidate_inode_pages2_range
2620 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2621 once one is completed. As a result, some RPCs are stuck in the
2622 queue and are never sent.
2626 Description: filter threads hungs on waiting journal commit
2627 Details : Cleanup filter group llog code, then only filter group llog will
2628 be only created in the MDS/OST syncing process.
2632 Description: Procfs and llog threads access destoryed import sometimes.
2633 Details : Sync the import destoryed process with procfs and llog threads by
2634 the import refcount and semaphore.
2636 Severity : enhancement
2638 Description: openlock cache of b1_6 port to HEAD
2643 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2644 Details : If insertion of an extent fails, then discard the inode
2645 preallocation and free data blocks else it can lead to duplicate
2650 Description: don't always update ctime in ext3_xattr_set_handle()
2651 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2652 In some cases the ctime should not be updated, for example for
2653 2.0->1.8 compatibility it is necessary to delete an xattr and it
2654 should not update the ctime.
2658 Bugzilla : 15713/16362
2659 Description: Assertion in iopen_connect_dentry in 1.6.3
2660 Details : looking up an inode via iopen with the wrong generation number can
2661 populate the dcache with a disconneced dentry while the inode
2662 number is in the process of being reallocated. This causes an
2663 assertion failure in iopen since the inode's dentry list contains
2664 both a connected and disconnected dentry.
2668 Description: assertion failure in ldlm_handle2lock()
2669 Details : fix a race between class_handle_unhash() and class_handle2object()
2670 introduced in lustre 1.6.5 by bug 13622.
2675 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2676 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2681 Description: oops in sd_iostats_seq_show()
2682 Details : unloading/reloading the scsi low level driver triggers a kernel
2683 bug when trying to access the sd iostat file.
2688 Description: Kernel panics during QLogic driver reload
2689 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2690 patch, causing memory corruption.
2695 Description: journal_dev option does not work in b1_6
2696 Details : pass mount option during pre-mount.
2698 Severity : enhancement
2700 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2701 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2702 extent information of a file. It can be used to map logical blocks
2703 in a file to physical blocks in the block device.
2707 Description: LDLM soft lockups - improvement
2708 Details : It is be possible to send the lock handle along with each read
2709 or write request because the client is already doing a lock match
2710 itself so there isn't any reason the OST should have to re-do that
2714 Frequency : only on Cray X2
2716 Description: X2 build failures
2717 Details : fix build failures on Cray X2.
2721 Description: xid & resent requests
2722 Details : Initialize RPC XID from clock at startup (randomly if clock is
2725 Severity : enhancement
2727 Description: Add lustre_start utility to start or stop multiple Lustre servers
2732 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2733 Details : In case of memory pressure, list_del() can be called twice on
2734 req->rq_history_list, causing a kernel oops.
2738 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2739 Details : fix stack overflow in the distributed lock manager by defering
2740 export eviction after a failed AST to the elt thread instead of
2741 handling it in the dlm interpret routine.
2745 Description: Convert some comments to new format.
2746 Details : Update documenting comments to match doxygen conventions.
2750 Description: Grammar fixes.
2751 Details : A couple of trivial spelling fixes.
2755 Description: OSD_COUNTERS-mandatory
2756 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2757 to simplify the code.
2761 Description: simplify cmm_device freeing logic.
2762 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2766 Description: Add lockdep support to dt_object_operations locking interface.
2767 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2768 parameter indicating lock ordering. Update mdd code to use new
2773 Description: Introduce failloc constants for lockless IO tests.
2774 Details : Add two new failloc constants to test lockless IO. Only one of
2775 them in implemented---another is checked in yet to be landed
2780 Description: Add lockdep support for inode mutex.
2781 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2782 in the situations where Lustre has to lock more than one inode
2787 Description: Add optional invariants checking support.
2788 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2789 configure switch. This macro is to be used for consistency and
2790 sanity checks that are too expensive to be left in `production'
2795 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2796 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2801 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2802 Details : Use spin_lock_nested() in (the only) situation where more than
2803 one ldlm_lock is locked simultaneously. Also, fix possible
2804 dead-lock in ldlm_lock_change_resource() by enforcing particular
2809 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2810 Details : Instead of specifying each ldlm_lock call-back through separate
2811 parameter, wrap them into struct ldlm_callback_suite.
2815 Description: Kill join_lru obd method and its callers.
2816 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2817 in memory---a requirement implemented through ->o_join_lru() obd
2818 method in HEAD. Remove this method and its users.
2822 Description: Add asynchronous ldlm ENQUEUE completion handler.
2823 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2824 case that stock ldlm_completion_ast() cannot handle as it waits
2825 until lock is granted. Introduce new ldlm_completion_ast_async()
2826 for this. Also comment ldlm_completion_ast().
2830 Description: ldlm_error <-> errno conversion.
2831 Details : Add functions to map (rather arbitrary) between LDLM error codes
2832 and standard errno values. CLIO needs this to prevent LDLM specific
2833 constants from escaping ldlm and osc.
2837 Description: Kill unused ldlm_handle2lock_ns() function.
2838 Details : Kill unused ldlm_handle2lock_ns() function.
2842 Description: Add lu_ref support to ldlm_lock
2843 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2844 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2845 and ldlm_resource. LDLM interface has to be changed, because code that
2846 releases a reference on a lock, has to "know" what reference this is.
2847 In the most frequent case
2849 lock = ldlm_handle2lock(handle);
2851 LDLM_LOCK_PUT(lock);
2853 no changes are required. When any other reference (received _not_ from
2854 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2855 instead of LDLM_LOCK_PUT().
2857 Arguably, changes are pervasive, and interface requires some discipline
2858 for proper use. On the other hand, it was very instrumental in finding
2859 a few leaked lock references.
2863 Description: Add ldlm_lock_addref_try().
2864 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2865 attempts to addref a lock that might be being canceled
2870 Description: Add ldlm_weigh_callback().
2871 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2872 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2873 hard-coded `number of pages' logic.
2877 Description: Add lockdep annotations to llog code.
2878 Details : Use appropriately tagged _nested() locking calls in the places
2879 where llog takes more than one ->lgh_lock lock.
2883 Description: Add loi_kms_set().
2884 Details : Wrap kms updates into a helper function.
2888 Description: Constify instances of struct lsm_operations.
2889 Details : Constify instances of struct lsm_operations.
2893 Description: lu_conf support.
2894 Details : On a server, a file system object is uniquely identified
2895 by a fid, which is sufficient to locate and load all object
2896 state (inode). On a client, on the other hand, more data are
2897 necessary instantiate an object. Change lu_object_find() and
2898 friends to take additional `lu_conf' argument describing object.
2899 Typically this includes layout information.
2903 Description: lu_context fixes.
2904 Details : Introduce new lu_context functions that are needed on the client
2905 side, where some system threads (ptlrpcd) are shared by multiple
2906 modules, and so cannot be stopped during module shutdown.
2910 Description: Add start and stop methods to lu_device_type_operations.
2911 Details : Introduce two new methods in lu_device_type_operations, that are
2912 invoked when first instance of a given type is created and last one
2913 is destroyed respectively. This is need by CLIO.
2917 Description: Add lu_ref support to struct lu_device.
2918 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2919 track leaked references.
2923 Description: Introduce lu_kmem_descr.
2924 Details : lu_kmem_descr and its companion interface allow to create
2925 and destroy a number of kmem caches at once.
2929 Description: Fix lu_object finalization race.
2930 Details : Fix a race between lu_object_find() finding an object and its
2931 concurrent finalization. This race is (most likely) not possible
2932 on the server, but might happen on the client.
2936 Description: Introduce lu_ref interface.
2937 Details : lu_ref is a debugging module allowing to track references to
2938 a given object. It is quite cpu expensive, and has to be
2939 explicitly enabled with --enable-lu_ref. See usage description
2944 Description: Factor lu_site procfs stats into a separate function.
2945 Details : Separate lu_site stats printing code into a separate function
2946 that can be reused on a client.
2950 Description: Constify instances of struct {lu,dt,md}_device_operations.
2951 Details : Constify instances of struct {lu,dt,md}_device_operations.
2955 Description: Introduce struct md_site and move meta-data specific parts of
2956 struct lu_site here.
2957 Details : Move md-specific fields out of struct lu_site into special struct
2958 md_site, so that lu_site can be used on a client.
2962 Description: Kill mdd_lov_destroy().
2963 Details : Remove unused mdd code.
2967 Description: Add st_block checking to multistat.c.
2968 Details : Add st_block checking to multistat.c.
2972 Description: Add lu_ref support to struct obd_device.
2973 Details : Add lu_ref tracking to obd_device.
2977 Description: Kill obd_set_fail_loc().
2978 Details : Remove unused code.
2982 Description: Add special type for ptlrpc_request interpret functions.
2983 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2984 there. Actual usage will be in CLIO.
2988 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2989 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2990 doesn't work with lockdep.
2994 Description: Add rwv.c test program.
2995 Details : New testing program exercising readv(2) and writev(2) (Qian).
2999 Description: Add sendfile.c test program.
3000 Details : New testing program exercising sendfile(2) (Jay).
3004 Description: Ratelimit a message that can be very frequent.
3005 Details : Ratelimit a memory allocation failure message that can
3010 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3011 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3015 Description: Liblustre build fixes.
3016 Details : Liblustre build fixes.
3020 Description: libcfs: add cfs_{need,cond}_resched() interface.
3021 Details : libcfs: add cfs_{need,cond}_resched() definition and
3022 implementations for Linux, NT, and liblustre.
3024 Severity : enhancement
3026 Description: More exported tunables for mballoc
3027 Details : Add support for tunable preallocation window and new tunables for
3028 large/small requests
3032 Description: Detect corruption of block bitmap and checking for preallocations
3033 Details : Checks validity of on-disk block bitmap. Also it does better
3034 checking of number of applied preallocations. When corruption is
3035 found, it turns filesystem readonly to prevent further corruptions.
3039 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3040 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3041 window to 0. If it does hit cache miss, and read-pattern is still
3042 stride-io mode, does not reset the stride window, but also does not
3043 increase the stride window length in this case.
3047 Frequency : only for big-endian servers
3048 Description: Check if system is big-endian while mounting fs with extents feature
3049 Details : Mounting a filesystem with extents feature will fail on big-endian
3050 systems since ext3-based ldiskfs is not supported on big-endian
3051 systems. This can be over-riden with "bigendian_extents" mount option.
3053 Severity : enhancement
3055 Description: The root squash functionality
3056 Details : A security feature, which is to prevent users from being able
3057 to mount lustre on their desktop, run as root, and delete
3058 all of the files in the filesystem. The goal is accomplished by
3059 remapping user id (UID) and group id (GID) of the root user to
3060 a UID and GID specified by the system administartor via Lustre
3061 configuration management server (MGS). The functionality also
3062 allows to specify sets of clients for which the remapping does
3067 Description: Excessive recovery window
3068 Details : With AT enabled, the recovery window can be excessively long (6000+
3069 seconds). To address this problem, we no longer use
3070 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3071 timeout no longer depends on the service time, it is set to
3072 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3073 time via pb_service_time.
3077 Description: Watchdog triggered on MDS failover
3078 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3079 the OSTs know that the MDS "UUID" can be reused for the same export
3080 from a different NID, so we do not need to wait for the export to be
3084 Frequency : rare, only if using MMP with Linux RAID
3086 Description: MMP doesn't work with Linux RAID
3087 Details : While using HA for Lustre servers with Linux RAID, it is possible
3088 that MMP will not detect multiple mounts. To make this work we
3089 need to unplug the device queue in RAID when the MMP block is being
3090 written. Also while reading the MMP block, we should read it from
3091 disk and not the cached one.
3093 Severity : enhancement
3095 Description: open file using fid
3096 Details : A file can be opened using just its fid, like
3097 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3100 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3103 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3104 Details : add proper locking for creation of jbd_4k slab cache
3108 Description: MMP check in ext3_remount() fails without displaying any error
3109 Details : When multiple mount protection fails during remount, proper error
3112 Severity : enhancement
3114 Description: Allow stripe size to be up to 4G-64k
3115 Details : Fix math logic to allow large stripe sizes.
3119 Description: add check for >8TB ldiskfs filesystems
3120 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3121 Don't allow >8TB ldiskfs filesystems to be mounted without
3122 force_over_8tb mount option
3127 Descriptoin: OSS complains "too many blocks" and crashes
3128 Details : In filter_commitrw_write():
3129 1. lquota_chkquota();
3130 2. obdo_from_inode();
3131 3. lquota_pending_commit();
3132 "2." has changed uid and gid in this case, then caused
3133 lquota_chkquota() and lquota_pending_commit() to use
3134 different uid and gid. That is the root of the bug.
3138 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3139 Details : Detect and print out the 'stripe count = -1' case properly.
3143 Description: Out or order replies might be lost on replay
3144 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3145 request with tid smaller than one currently being added, add it
3146 to the start, not end of the list.
3148 --------------------------------------------------------------------------------
3150 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3152 * Support for kernels:
3153 2.6.5-7.283 (SLES 9),
3154 2.6.9-55.EL (RHEL 4),
3155 2.6.16.46-0.14 (SLES 10),
3156 2.6.18.8 vanilla (kernel.org)
3157 * Client support for unpatched kernels:
3158 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3159 2.6.16 - 2.6.22 vanilla (kernel.org)
3160 * Due to recently discovered recovery problems, we do not recommend
3161 using patchless RHEL 4 clients with this or any earlier release.
3162 * Recommended e2fsprogs version: 1.39.cfs8
3163 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3164 * Starting with this release, the ldiskfs backing filesystem required
3165 by Lustre is now in its own package, lustre-ldiskfs. This package
3166 should be installed. It is versioned separately from Lustre and
3167 may be released separately in future.
3171 Description: block reactivating mgc import until all deactivates complete
3172 Details : Fix race when failing back MDT/MGS to itself (testing)
3175 Frequency : at statup only
3177 Description: mds_lov_synchronize race leads to various problems
3178 Details : simultaneous MDT->OST connections at startup can cause the
3179 sync to abort, leaving the OSC in a bad state.
3181 Severity : enhancement
3183 Description: add optional extra BUILD_VERSION info
3184 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3185 to override the lustre version.
3188 Frequency : 2.6.18 servers only
3190 Description: ll_kern_mount() doesn't release the module reference
3191 Details : The ldiskfs module reference count never drops down to 0
3192 because ll_kern_mount() doesn't release the module reference.
3197 Description: server LBUG when using old ost_num_threads parameter
3198 Details : Accept the old ost_num_threads parameter but warn that it
3199 is deprecated, and fix an off-by-one error that caused an LBUG.
3204 Description: Transient SCSI error results in persistent IO issue
3205 Details : iobuf->dr_error is not reinitialized to 0 between two
3209 Frequency : sometimes when underlying device returns I/O errors
3211 Description: OSTs not going read-only during write failures
3212 Details : OSTs are not remounted read-only when the journal commit threads
3213 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3214 instead of the ext3 wrappers.
3218 Description: poor connect scaling with increasing client count
3219 Details : Don't run filter_grant_sanity_check for more than 100 exports
3220 to improve scaling for large numbers of clients.
3223 Frequency : SLES10 only
3225 Description: sanity-quota.sh quotacheck failed: rc = -22
3226 Details : Quotas cannot be enabled on SLES10.
3229 Frequency : liblustre clients only
3231 Description: getdirentries does not give error when run on compute nodes
3232 Details : getdirentries does not fail when the size specified as an argument
3233 is too small to contain at least one entry
3235 Severity : enhancement
3237 Description: Add LNET router traceability for debug purposes
3238 Details : If a checksum failure occurs with a router as part of the
3239 IO path, the NID of the last router that forwarded the bulk data
3240 is printed so it can be identified.
3245 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3246 Details : Due to a race condition, liblustre clients were occasionally
3247 evicted incorrectly.
3249 Severity : enhancement
3251 Description: lfs setstripe use optional parameters instead of postional
3254 Severity : enhancement
3256 Description: Nanosecond timestamp support for ldiskfs
3257 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3258 resolution timestamps. There is not yet support for this at
3259 the Lustre filesystem level.
3262 Frequency : during server recovery
3264 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3265 Details : request with rq_no_resend flag not awake l_wait_event if they get
3269 Frequency : nfs export on patchless client
3271 Description: connectathon hang when test nfs export over patchless client
3272 Details : Disconnected dentry cannot be found with lookup, so we do not need
3273 to unhash it or make it invalid
3276 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3277 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3278 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3283 Description: open req refcounting wrong on reconnect
3284 Details : If reconnect happened between getting open reply from server and
3285 call to mdc_set_replay_data in ll_file_open, we will schedule
3286 replay for unreferenced request that we are about to free.
3287 Subsequent close will crash in variety of ways.
3288 Check that request is still eligible for replay in
3289 mdc_set_replay_data().
3294 Description: disable writes to filesystem when reading health_check file
3295 Details : the default for reading the health_check proc file has changed
3296 to NOT do a journal transaction and write to disk, because this
3297 can cause reads of the /proc file to hang and block HA state
3298 checking on a healthy but otherwise heavily loaded system. It
3299 is possible to return to the previous behaviour during configure
3300 with --enable-health-write.
3302 Severity : enhancement
3304 Description: 64-bit inode version
3305 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3306 to the inode. This will be required for version-based recovery.
3311 Description: MDS fails to start if a duplicate client export is detected
3312 Details : in some rare cases it was possible for a client to connect to
3313 an MDS multiple times. Upon recovery the MDS would detect this
3314 and fail during startup. Handle this more gracefully.
3316 Severity : enhancement
3318 Description: Add -o localflock option to simulate old noflock
3320 Details : This will achieve local-only flock/fcntl locks
3326 Description: log_commit_thread vs filter_destroy race leads to crash
3327 Details : Take import reference before releasing llog record semaphore
3332 Description: Wrong request locking in request set processing
3333 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3334 imp_delayed_list, in this place should be used imp_lock.
3337 Frequency : when reconnection
3339 Description: Grant Leak when osc reconnect to OST
3340 Details : When osc reconnect ost, OST(filter) should check whether it
3341 should grant more space to client by comparing fed_grant and
3342 cl_avail_grant, and return the granted space to client instead
3343 of "new granted" space, because client will call osc_init_grant
3344 to update the client grant space info.
3347 Frequency : when client reconnect to OST
3349 Description: Grant Leak when osc do resend and replay bulk write
3350 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3351 bulk write request, because the grant info while be sync between
3352 OSC and OST when reconnect, and we should ignore the grant info
3353 these of resend/replay write req.
3358 Description: Grant space more than available left space sometimes.
3359 Details : When then OST is about to be full, if two bulk writing from
3360 different clients came to OST. Accord the avaliable space of the
3361 OST, the first req should be permitted, and the second one
3362 should be denied by ENOSPC. But if the seconde arrived before
3363 the first one is committed. The OST might wrongly permit second
3364 writing, which will cause grant space > available space.
3367 Frequency : when client is evicted
3369 Description: Grant might be wrongly erased when osc is evicted by OST
3370 Details : when the import is evicted by server, it will fork another
3371 thread ptlrpc_invalidate_import_thread to invalidate the
3372 import, where the grant will be set to 0. While the original
3373 thread will update the grant it got when connecting. So if
3374 the former happened latter, the grant will be wrongly errased
3375 because of this race.
3380 Description: Checking Stale with correct fid
3381 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3382 to check whether it is stale, because sometimes, we want the
3383 enqueue happened anyway, and op_data.fid2 will not be initialized.
3385 Severity : enhancement
3387 Description: update patchless client
3388 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3391 Frequency : only with 2.4 kernel
3393 Description: random memory corruption
3394 Details : size of struct ll_inode_info is to big for union inode.u and this
3395 can be cause of random memory corruption.
3400 Description: Memory leak in recovery
3401 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3402 It should also check obd_fail before fsfilt_start, otherwise if
3403 fsfilt_start return -EROFS,(failover mds during mds recovery).
3404 then the req will return with repmsg->transno = 0 and rc = EROFS.
3405 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3406 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3407 be freed no matter whether fsfilt_commit success or not.
3410 Frequency : only with huge count clients
3412 Description: Prevents from taking the superblock lock in llap_from_page for
3414 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3415 for llap_from_page call in ll_removepage prevents from taking the
3416 superblock lock for a soon died page.
3421 Description: Not check open intent error before release open handle
3422 Details : in some rare cases, the open intent error is not checked before
3423 release open handle, which may cause
3424 ASSERTION(open_req->rq_transno != 0), because it tries to release
3425 the failed open handle.
3430 Description: Set cat log bitmap only after create log success.
3431 Details : in some rare cases, the cat log bitmap is set too early. and it
3432 should be set only after create log success.
3436 Description: Accessing a block bevice can re-enable I/O when Lustre is
3437 tearing down a device.
3438 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3443 Description: service threads may hog cpus when there are a lot of requests
3444 Details : Insert cond_resched to give other threads a chance to use some CPU
3449 Description: the cat log was not initialized in recovery
3450 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3451 unlink log on mds will not be initialized until mds post
3452 recovery. And also in mds post recovery, the unlink log will
3453 initialization will be done asynchronausly, so there will be race
3454 between add unlink log and unlink log initialization.
3458 Description: brw_stats were being printed incorrectly
3459 Details : brw_stats were being printed as log2 but all of them were not
3460 recorded as log2. Also remove some code duplication arising from
3461 filter_tally_{read,write}.
3465 Frequency : rare, only in recovery.
3466 Description: ASSERTION(req->rq_type != LI_POISON) failed
3467 Details : imp_lock should be held while iterating over imp_sending_list for
3468 prevent destroy request after get timeout in ptlrpc_queue_wait.
3472 Description: replay-single.sh test 52 fails
3473 Details : A lock's skiplist need to be cleanup when it being unlinked
3474 from its resource list.
3478 Description: Short directio read returns full requested size rather than
3480 Details : Direct I/O operations should return actual amount of bytes
3481 transferred rather than requested size.
3483 Severity : enhancement
3485 Description: metadata RPC reduction (e.g. for rm performance)
3486 Details : decrease the amount of synchronous RPC between clients and servers
3487 by canceling conflicing lock before the operation on the client
3488 and packing thier handles into the main operation RPC to server.
3490 Severity : enhancement
3492 Description: Async OSC create to avoid the blocking unnecessarily.
3493 Details : If an OST has no remaining object, system will block on the
3494 creation when it needs to create a new object on this OST. Now,
3495 ways use pre-created objects when available, instead of blocking on
3496 an empty osc while others are not empty. If we must block, we
3497 block for the shortest possible period of time.
3501 Description: improve handling recoverable errors
3502 Details : if request processig with error which can be recoverable on server
3503 request should be resend, otherwise page released from cache and
3506 Severity : enhancement
3508 Description: refine locking for avoid write wrong info into lov_objid file
3509 Details : fix possible races with add new target and write/update data in
3512 --------------------------------------------------------------------------------
3514 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3519 Frequency : on some architectures
3521 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3522 Details : Uninitialized flags sometimes cause configuration commands to
3526 Frequency : patchless clients only
3528 Description: missing __iget() symbol export
3529 Details : The __iget() symbol export is missing. To avoid the need for
3530 this on patchless clients the deathrow inode reaper is turned
3531 off, and we depend on the VM to clean up old inodes. This
3532 dependency was during via the fix for bug 12181.
3534 --------------------------------------------------------------------------------
3536 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3538 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3539 INTEROPERATE with older versions automatically. In many cases a
3540 special upgrade step is needed. Please read the
3541 user documentation before upgrading any part of a live system.
3542 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3543 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3544 * WARNING: Lustre configuration and startup changes are required with
3545 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3549 Severity : enhancement
3551 Description: MountConf
3552 Details : Lustre configuration is now managed via mkfs and mount
3553 commands instead of lmc and lconf. New obd types (MGS, MGC)
3554 are added for dynamic configuration management. See
3555 https://mail.clusterfs.com/wikis/lustre/MountConf for
3558 Severity : enhancement
3560 Description: dynamic OST addition
3561 Details : OSTs can now be added to a live filesystem
3563 Severity : enhancement
3565 Description: startup order invariance
3566 Details : MDTs and OSTs can be started in any order. Clients only
3567 require the MDT to complete startup.
3569 Severity : enhancement
3571 Description: parallel, asynchronous orphan cleanup
3572 Details : orphan cleanup is now performed in separate threads for each
3573 OST, allowing parallel non-blocking operation.
3575 Severity : enhancement
3577 Description: optimized stripe assignment
3578 Details : stripe assignments are now made based on ost space available,
3579 ost previous usage, and OSS previous usage, in order to try
3580 to optimize storage space and networking resources.
3582 Severity : enhancement
3584 Description: Permanently set tunables
3585 Details : All writable /proc/fs/lustre tunables can now be permanently
3586 set on a per-server basis, at mkfs time or on a live system.
3588 Severity : enhancement
3590 Description: Lustre message v2
3591 Details : Add lustre message format v2.
3593 Severity : enhancement
3595 Description: client OST exclusion list
3596 Details : Clients can be started with a list of OSTs that should be
3597 declared "inactive" for known non-responsive OSTs.
3601 Description: SPEC SFS validation failure on NFS v2 over lustre.
3602 Details : Changes the blocksize for regular files to be 2x RPC size,
3603 and not depend on stripe size.
3605 Severity : enhancement
3607 Description: Multiple MD RPCs in flight.
3608 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3609 To avoid overly-overloading MDS, introduce a limit on number of
3610 MDS RPCs in flight for a single client and add /proc controls
3611 to adjust this limit.
3613 Severity : enhancement
3615 Description: client read/write statistics
3616 Details : Add client read/write call usage stats for performance
3617 analysis of user processes.
3618 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3619 file access. extents_stats shows chunk size distribution.
3620 extents_stats_per_process show chunk size distribution per
3623 Severity : enhancement
3625 Description: mds statistics
3626 Details : Add detailed mds operations statistics in
3627 /proc/fs/lustre/mds/*/stats.
3631 Description: Failure of copying files with lustre special EAs.
3632 Details : Client side always return success for setxattr call for lustre
3633 special xattr (currently only "trusted.lov").
3638 Description: Refcount LNET uuids
3639 Details : The global LNET uuid list grew linearly with every startup;
3640 refcount repeated list entries instead of always adding to
3643 Severity : enhancement
3645 Description: Dynamic service threads
3646 Details : Within a small range, start extra service threads
3647 automatically when the request queue builds up.
3650 Frequency : mixed-endian client/server environments
3652 Description: mixed-endian crashes
3653 Details : The new msg_v2 system had some failures in mixed-endian
3656 Severity : enhancement
3658 Description: Easy OST removal
3659 Details : OSTs can be permanently deactivated with e.g. 'lctl
3660 conf_param lustre-OST0001.osc.active=0'
3662 Severity : enhancement
3664 Description: MGS proc entries
3665 Details : Added basic proc entries for the MGS showing what filesystems
3668 Severity : enhancement
3670 Description: provide MGS failover
3671 Details : Added config lock reacquisition after MGS server failover.
3673 Severity : enhancement
3675 Description: add Linux 2.4 support
3676 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3680 Description: a large application tries to do I/O to the same resource and dies
3681 in the middle of it.
3682 Details : Check the req->rq_arrival time after the call to
3683 ost_brw_lock_get(), but before we do anything about
3684 processing it & sending the BULK transfer request. This
3685 should help move old stale pending locks off the queue as
3686 quickly as obd_timeout.
3689 Frequency : when an incorrect nid is specified during startup
3691 Description: ptlrpc connect to non-existant node causes kernel crash
3692 Details : LNET can't be re-entered from an event callback, which
3693 happened when we expire a message after the export has been
3694 cleaned up. Instead, hand the zombie cleanup off to another
3697 Severity : enhancement
3699 Description: plain/inodebits lock performance improvement
3700 Details : Grouping plain/inodebits in granted list by their request modes
3701 and bits policy, thus improving the performance of search through
3705 Frequency : only if OST filesystem is corrupted
3707 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3708 Details : for a short time RPCs with bulk IO are in the replay list,
3709 but replay of bulk IOs is unimplemented. If the OST filesystem
3710 is corrupted due to disk cache incoherency and then replay is
3711 started it is possible to trip an assertion. Avoid putting
3712 committed RPCs into the replay list at all to avoid this issue.
3715 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3717 Description: System hang on startup
3718 Details : This bug allowed the liblustre (e.g. catamount) client to
3719 return to the app before handling all startup RPCs. This
3720 could leave the node unresponsive to lustre network traffic
3721 and manifested as a server ptllnd timeout.
3723 Severity : enhancement
3725 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3726 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3727 boolean to control whether to print peer debug info when a
3728 client's RPC times out.
3731 Frequency : only for kernels with patches from Lustre below 1.4.3
3733 Description: Remove old rdonly API
3734 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3737 Frequency : only for devices with external journals
3739 Description: Set external device read-only also
3740 Details : During a commanded failover stop, we set the disk device
3741 read-only while the server shuts down. We now also set any
3742 external journal device read-only at the same time.
3745 Frequency : when upgrading from 1.4 while trying to change parameters
3747 Description: The wrong (new) MDC name was used when setting parameters for
3748 upgraded MDT's. Also allows changing of OSC (and MDC)
3749 parameters if --writeconf is specified at tunefs upgrade time.
3752 Frequency : when setting specific ost indicies
3754 Description: QOS code breaks on skipped indicies
3755 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3756 created with --index need not be sequential.
3758 Severity : enhancement
3760 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3761 Details : The uninit_groups feature works in conjunction with the kernel
3762 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3763 up the pass1 processing of e2fsck. This is a read-only feature
3764 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3765 mount filesystems that have had this feature enabled.
3767 Severity : enhancement
3769 Description: Improve multi-block allocation algorithm to avoid fragmentation
3770 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3771 allocation locality and avoid filesystem fragmentation.
3773 ------------------------------------------------------------------------------
3775 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3777 * Support for kernels:
3778 2.4.21-47.0.1.EL (RHEL 3)
3779 2.6.5-7.283 (SLES 9)
3780 2.6.9-42.0.10.EL (RHEL 4)
3781 2.6.12.6 vanilla (kernel.org)
3782 2.6.16.27-0.9 (SLES 10)
3783 * Recommended e2fsprogs version: 1.39.cfs5
3785 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3789 Frequency : occasional, depends on client load and configuration
3790 Bugzilla : 12181, 12203
3791 Description: data loss for recently-modified files
3793 Details : In some cases it is possible that recently written or created
3794 files may not be written to disk in a timely manner (this should
3795 normally be within 30s unless client IO load is very high).
3796 The problem appears as zero-length files or files that are a
3797 multiple of 1MB in size after a client crash or client eviction
3798 that are missing data at the end of the file.
3800 This problem is more likely to be hit on clients where files are
3801 repeatedly created and unlinked in the same directory, clients
3802 have a large amount of RAM, have many CPUs, the filesystem has
3803 many OSTs, the clients are rebooted frequently, and/or the files
3804 are not accessed by other nodes after being written.
3806 The presence of the problem can be detected by looking at
3807 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3808 smaller than the second (nr_unused) then dirty files will not
3809 be flushed automatically to disk. "sync; sleep 10" should be
3810 run several times on the node before unmounting it to update
3811 Lustre (this is also safe to run on nodes without this problem).
3813 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3814 that can cause this same problem, so customers using that kernel
3815 also need to update the kernel in addition to Lustre. In order
3816 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3818 It is normal that files written just before a client crash (less
3819 than 30s) may not yet have been flushed to disk, even for local
3823 Frequency : frequent on thin XT3 nodes
3825 Description: UUID collision on thin XT3 Linux nodes
3826 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3827 randomly, since we relied on an insufficiently-seeded PRNG.
3832 Description: OSS hangs after "All ost request buffers busy"
3833 Details : A deadlock between quota and journal operations caused OSS
3834 hangs after printing "All ost request buffers busy."
3837 Frequency : always on liblustre builds
3839 Description: Cleanup compiler warnings on liblustre
3842 Frequency : always on liblustre builds on XT3
3844 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3848 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3849 > 32000 subdirectories
3850 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3851 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3852 subdirectory count crosses 32000. This will aid e2fsck to
3853 correctly handle more than 32000 subdirectories.
3856 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3858 Description: System hang on startup
3859 Details : This bug allowed the liblustre (e.g. catamount) client to
3860 return to the app before handling all startup RPCs. This
3861 could leave the node unresponsive to lustre network traffic
3862 and manifested as a server ptllnd timeout.
3864 Severity : enhancement
3866 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3867 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3868 boolean to control whether to print peer debug info when a
3869 client's RPC times out.
3874 Description: make O_SYNC working on 2.6 kernels
3875 Details : 2.6 kernels use different method for mark pages for write,
3876 so need add a code to lustre for O_SYNC work.
3881 Description: Failure to close file and release space on NFS
3882 Details : Put inode details into lock acquired in ll_intent_file_open.
3883 Use mdc_intent_lock in ll_intent_open to properly
3884 detect all kind of errors unhandled by mdc_enqueue.
3889 Description: proc file read during shutdown sometimes raced obd removal,
3891 Details : Add lock to prevent obd access after proc file removal.
3894 Frequency : Only for files larger than 4GB on 32-bit clients.
3896 Description: improperly doing page alignment of locks
3897 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3898 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3903 Description: RPCs being resent when they shouldn't be
3904 Details : Some RPCs that should not be resent are being resent. This
3905 can cause inconsistencies in the RPC state machine. Do not
3906 resend such requests.
3909 Frequency : rare, only with NFS export
3911 Description: Crash on NFS re-export node
3912 Details : under very unusual load conditions an assertion is hit in
3913 ll_intent_file_open()
3916 Frequency : only if OST filesystem is corrupted
3918 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3919 Details : for a short time RPCs with bulk IO are in the replay list,
3920 but replay of bulk IOs is unimplemented. If the OST filesystem
3921 is corrupted due to disk cache incoherency and then replay is
3922 started it is possible to trip an assertion. Avoid putting
3923 committed RPCs into the replay list at all to avoid this issue.
3928 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3929 Details : Large single O_DIRECT read and write calls can fail to allocate
3930 a sufficiently large buffer to process the request. In case of
3931 allocation failure the allocation is retried with a smaller
3932 buffer and broken into smaller requests.
3934 Severity : enhancement
3936 Description: Add -o localflock option to simulate old noflock behaviour.
3937 Details : This will achieve local-only flock/fcntl locks coherentness.
3942 Description: versioning check is incomplete
3943 Details : Checking the version difference of client vs. server, report
3944 error if the gap is too big.
3949 Description: add support PG_writeback bit
3950 Details : add support for PG_writeback bit for Lustre, for more carefull
3951 work with page cache in 2.6 kernel. This also fix some deadlocks
3952 and remove hack for work O_SYNC with 2.6 kernel.
3954 Severity : enhancement
3956 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3957 Details : The uninit_groups feature works in conjunction with the kernel
3958 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3959 up the pass1 processing of e2fsck. This is a read-only feature
3960 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3961 mount filesystems that have had this feature enabled.
3963 Severity : enhancement
3965 Description: Improve multi-block allocation algorithm to avoid fragmentation
3966 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3967 allocation locality and avoid filesystem fragmentation.
3969 ------------------------------------------------------------------------------
3971 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3973 * Support for kernels:
3974 2.6.9-42.0.3.EL (RHEL 4)
3975 2.6.5-7.276 (SLES 9)
3976 2.4.21-47.0.1.EL (RHEL 3)
3977 2.6.12.6 vanilla (kernel.org)
3978 2.6.16.21-0.8 (SLES10)
3979 * Recommended e2fsprogs version: 1.39.cfs2-0
3981 * The backwards-compatible /proc/sys/portals symlink has been removed
3982 in this release. Before upgrading, please ensure that you change
3983 any configuration scripts or /etc/sysctl.conf files that access
3984 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3985 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3986 in advance of the upgrade on any system running Lustre 1.4.6 or
3987 newer, since /proc/sys/lnet was added in that version.
3988 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3992 Frequency : only when quota is used
3994 Description: avoid scanning export list for quota master
3995 Details : Change the algorithms to avoid scanning export list in order
3996 to improve the efficiency.
3999 Frequency : MDS failover only, very rarely
4001 Description: "went back in time" messages on mds failover
4002 Details : The greatest transno may be lost when the current operation
4003 finishes with an error (transno==0) and the client's last_rcvd
4004 record is over-written. Save the greatest transno in the
4005 mds_last_transno for this case.
4008 Frequency : always for specific kernels and striping counts
4010 Description: client may get "Matching packet too big" without ACL support
4011 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4012 when trying to access files in certain configurations. The
4013 clients should in fact be denied when mounting because they do
4014 not understand ACLs.
4017 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4019 Description: many clients connecting with IO in progress causes connect timeouts
4020 Details : Avoid synchronous journal commits to avoid delays caused by many
4021 clients connecting/disconnecting when bulk IO is in progress.
4022 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4023 OST_IO_PORTAL to avoid delays behind potentially many pending
4027 Frequency : occasionally with multiple writers to a single file
4029 Description: shared writes to file may result in wrong size reported by stat()
4030 Details : Allow growing of kms when extent lock is cancelled
4033 Frequency : always with random mmap IO to multi-striped file
4035 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4036 Details : Only if the hole is at the end of OST object so that kms is too
4037 small. Fix is to increase kms accordingly in ll_nopage.
4040 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4042 Description: writes to a missing object would leak memory on the OST
4043 Details : If there is an inconsistency between the MDS and OST filesystems,
4044 such that the MDS references an object that doesn't exist, writes
4045 to that object will leak memory due to incorrect cleanup in the
4046 error handling path, eventually running out of memory on the OST.
4051 Description: Creating too long symlink causes lustre errors
4052 Details : Check symlink and name lengths before sending requests to MDS.
4055 Frequency : only if flock is enabled (not on by default)
4057 Description: posix locks not released on fd closure on 2.6.9+
4058 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4059 kernels, this caused such locks not to be released on fd close and
4060 then assertions on fs unmount about still used locks.
4063 Frequency : MDS failover only, very rarely
4065 Description: clients may get ASSERTION(granted_lock != NULL)
4066 Details : When request was taking a long time, and a client was resending
4067 a getattr by name lock request. The were multiple lock requests
4068 with the same client lock handle and
4069 mds_getattr_name->fixup_handle_for_resent_request found one of the
4070 lock handles but later failed with ASSERTION(granted_lock != NULL).
4075 Description: handle->h_buffer_credits > 0, assertion failure
4076 Details : h_buffer_credits is zero after truncate, causing assertion
4077 failure. This patch extends the transaction or creates a new
4081 Frequency : NFS re-export or patchless client
4082 Bugzilla : 11179, 10796
4083 Description: Crash on NFS re-export node (__d_move)
4084 Details : We do not want to hash the dentry if we don't have a lock.
4085 But if this dentry is later used in d_move, we'd hit uninitialised
4086 list head d_hash, so we just do this to init d_hash field but
4087 leave dentry unhashed.
4090 Frequency : NFS re-export or patchless client
4092 Description: NFS exports has problem with symbolic link
4093 Details : lustre client didn't properly install dentry when re-exported
4094 to NFS or running patchless client.
4097 Frequency : NFS re-export or patchless client
4099 Description: Various nfs/patchless fixes.
4100 Details : fixes reuse disconected alias for lookup process - this fixes
4101 warning "find_exported_dentry: npd != pd",
4102 fix permission error with open files at nfs.
4103 fix apply umask when do revalidate.
4106 Frequency : occasional
4108 Description: Crash on NFS re-export node
4109 Details : calling clear_page() on the wrong pointer triggered oops in
4110 generic_mapping_read().
4113 Frequency : rarely, using O_DIRECT IO
4115 Description: unaligned directio crashes client with LASSERT
4116 Details : check for unaligned buffers before trying any requests.
4119 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4121 Description: stale data returned from RAID cache
4122 Details : If only a small amount of IO is done to the RAID device before
4123 reading it again it is possible to get stale data from the RAID
4124 cache instead of reading it from disk.
4127 Frequency : always for sles10 kernel
4129 Description: sles10 support
4130 Details : ll_follow_link: compile fixes and using of nd_set_link
4131 under newer kernels.
4134 Frequency : depends on arch, kernel and compiler version, always on sles10
4137 Description: recursive or deep enough symlinks cause stack overflow
4138 Details : getting rid of large stack-allocated variable in
4142 Frequency : depends on hardware
4144 Description: lustre write performance loss in the SLES10 kernel
4145 Details : the performance loss is caused by using of write barriers in the
4146 ext3 code. The SLES10 kernel turns barrier support on by
4147 default. The fix is to undo that change for ldiskfs.
4149 ------------------------------------------------------------------------------
4151 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4153 * Support for kernels:
4154 2.6.9-42.0.3EL (RHEL 4)
4155 2.6.5-7.276 (SLES 9)
4156 2.4.21-47.0.1.EL (RHEL 3)
4157 2.6.12.6 vanilla (kernel.org)
4161 Frequency : quota enabled and large files being deleted
4163 Description: releasing more than 4GB of quota at once hangs OST
4164 Details : If a user deletes more than 4GB of files on a single OST it
4165 will cause the OST to spin in an infinite loop. Release
4166 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4171 Description: statfs data retrieved from /proc may be stale or zero
4172 Details : When reading per-device statfs data from /proc, in the
4173 {kbytes,files}_{total,free,avail} files, it may appear
4174 as zero or be out of date.
4177 Frequency : systems with MD RAID1 external journal devices
4179 Description: lconf's call to blkid is confused by RAID1 journal devices
4180 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4181 of returning all block devices that match the journal UUID.
4184 Frequency : always, for aggregate stripe size over 4GB
4186 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4187 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4188 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4191 Frequency : always if "lfs find" used on a local file/directory
4193 Description: "lfs find" segfaults if used on a local file/directory
4194 Details : The case where a directory component was not specified wasn't
4195 handled correctly. Handle this properly.
4198 Frequency : always on ppc64
4200 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4201 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4202 perform endianness conversion.
4205 Frequency : rarely (truncate to non-zero file size after write under load)
4206 Bugzilla : 10730, 10687
4207 Description: Files padded with zeros to next 4K multiple
4208 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4209 it is possible that files that are truncated to a non-zero size
4210 immediately after being written are filled with zero bytes beyond
4211 the truncated size. No file data is lost.
4213 Severity : enhancement
4215 Description: Allow recovery/failover for liblustre clients.
4216 Details : liblustre clients were unaware of failover configurations until
4219 Severity : enhancement
4221 Description: user file locks should fail when not mounting with flock option
4222 Details : Set up an error-returning stub in ll_file_operations.lock field
4223 to prevent incorrect behaviour when client is mounted without
4224 flock option. Also, set up properly f_op->flock field for
4228 Frequency : always on ia64
4230 Description: "lfs df" loops on printing out MDS statfs information
4231 Details : The obd_ioctl_data was not initialized and in some systems
4232 this caused a failure during the ioctl that did not return
4233 an error. Initialize the struct and return an error on failure.
4236 Frequency : SLES 9 only
4238 Description: Error of copying files with lustre special EAs as root
4239 Details : Client side always return success for setxattr call for lustre
4240 special xattr (currently only "trusted.lov").
4243 Frequency : rarely on clusters with both ia64+i386 clients
4245 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4246 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4247 clients are doing concurrent writes to the same file it is
4248 possible that the ia64 clients may LASSERT because the OST
4249 extent locks are not PAGE_SIZE aligned. Ensure that grown
4250 locks are always aligned on the request boundary.
4253 Frequency : specific use, occasional
4255 Description: Overwriting in use executable truncates on-disk binary image
4256 Details : If one node attempts to overwrite an executable in use by
4257 another node, we now correctly return ETXTBSY instead of
4258 truncating the file.
4260 Severity : enhancement
4262 Description: Async OSC create to avoid the blocking unnecessarily.
4263 Details : If an OST has no remaining object, system will block on the creating
4264 when need to create a new object on this OST. Now, ways use
4265 pre-created objects when available, instead of blocking on an
4266 empty osc while others are not empty. If we must block, we block
4267 for the shortest possible period of time.
4272 Description: chmod on Lustre root is propagated to other clients
4273 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4274 invalid by the follow_mount time.
4277 Frequency : liblustre clients only
4279 Description: Race in 'instant cancel' lock handling could lead to such locks
4280 never to be granted in case of SMP MDS
4281 Details : Do not destroy not yet granted but cbpending locks in
4285 Frequency : replay/resend of open
4287 Description: non null lock assetion failure in mds_intent_policy
4288 Details : Trying to replay/resend lockless open requests resulted in
4289 mds_open() returning 0 with no lock. Now it sets a flag if
4290 a lock is going to be returned.
4292 Severity : enhancement
4294 Description: Checksum enhancements
4295 Details : New checksum enhancements allow for resending RPCs that failed
4298 Severity : enhancement
4300 Description: Tunables on number of dirty pages in cacche
4301 Details : Allow to set limit on number of dirty pages cached.
4306 Description: client crash on unmount - lock still has references
4307 Details : In some error handling cases it was possible to leak a lock
4308 reference on a client while accessing a file. This was not
4309 harmful to the client during operation, but would cause the
4310 client to crash when the filesystem is unmounted.
4313 Frequency : specific case, rare
4315 Description: ETXTBSY on mds though file not in use by client
4316 Details : ETXTBSY is no longer incorrectly returned when attempting to
4317 chmod or chown a directory that the user previously tried to
4318 execute or a currently-executing binary.
4321 Frequency : extremely rare except on liblustre-based clients
4323 Description: Lustre space not freed when files are deleted
4324 Details : Clean up open-unlinked files after client eviction. Previously
4325 the unlink was skipped and the files remained as orphans.
4330 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4331 Details : In some cases it was possible to send a blocking callback to
4332 a client doing a glimpse, even though that client didn't get
4333 a lock granted. When the glimpse lock is cancelled on the OST
4334 the freed lock is left on the waiting list and corrupted the list.
4337 Frequency : all core dumps
4339 Description: Broke core dumps to lustre
4340 Details : Negative dentry may be unhashed if parent does not have UPDATE
4341 lock, but some callers, e.g. do_coredump, expect dentry to be
4342 hashed after successful create, hash it in ll_create_it.
4344 ------------------------------------------------------------------------------
4346 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4348 * Support for kernels:
4349 2.6.9-42.0.2.EL (RHEL 4)
4350 2.6.5-7.276 (SLES 9)
4351 2.4.21-40.EL (RHEL 3)
4352 2.6.12.6 vanilla (kernel.org)
4356 Frequency : always on RHEL 3
4358 Description: Number of open files grows over time
4359 Details : The number of open files grows over time, whether or not
4360 Lustre is started. This was due to a filp leak introduced
4361 by one of our kernel patches.
4363 ------------------------------------------------------------------------------
4365 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4367 * Support for kernels:
4368 2.6.9-42.EL (RHEL 4)
4369 2.6.5-7.276 (SLES 9)
4370 2.4.21-40.EL (RHEL 3)
4371 2.6.12.6 vanilla (kernel.org)
4376 Bugzilla : 5719, 9635, 9792, 9684
4377 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4378 Details : If a server is under heavy load and cannot reply to new
4379 connection requests before the client resends the (re)connect,
4380 the connection handling code can behave badly if two service
4381 threads are concurrently handing separate (re)connections from
4382 the same client. Add better locking to the connection handling
4383 code, and ensure that only a single connection will be processed
4384 for a given client UUID, even if the lock is dropped.
4386 Severity : enhancement
4388 Description: add TCP zero-copy support to kernel
4389 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4390 sends if the hardware supports scatter-gather and checksumming.
4391 This allows socklnd to do client-write and server-read more
4392 efficiently and reduce CPU utilization from skbuf copying.
4395 Frequency : only if NFS exporting from client
4397 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4398 Details : If setting attributes on a file created under NFS that had
4399 never been opened it would be possible to oops the client
4400 if the file had no objects.
4403 Frequency : always for liblustre
4405 Description: liblustre client does MDS+OSTs setattr RPC for each write
4406 Details : When doing a write from a liblustre client, the client
4407 incorrectly issued an RPC to the MDS and each OST the file was
4408 striped over in order to update the timestamps. When writing
4409 with small chunks and many clients this could overwhelm the MDS
4410 with RPCs. In all cases it would slow down the write because
4411 these RPCs are unnecessary.
4413 Severity : enhancement
4415 Description: allow number of MDS service threads to be changed at module load
4416 Details : It is now possible to change the number of MDS service threads
4417 running. Adding "options mds mds_num_threads={N}" to the MDS's
4418 /etc/modprobe.conf will set the number of threads for the next
4419 time Lustre is restarted (assuming the "mds" module is also
4420 reloaded at that time). The default number of threads will
4421 stay the same, 32 for most systems.
4426 Description: OST crash if filesystem is unformatted or corrupt
4427 Details : If an OST is started on a device that has never been formatted
4428 or if the filesystem is corrupt and cannot even mount then the
4429 error handling cleanup routines would dereference a NULL pointer.
4434 Description: NULL pointer deref in llap_from_page.
4435 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4436 incorrect) mapping if the page was truncated/reclaimed while it was
4437 searched for. Check for this condition and skip such pages when
4438 doing readahead. Introduce extra check to llap_from_page() to
4439 verify page->mapping->host is non-NULL (so page is not anonymous).
4442 Frequency : Sometimes when using sys_sendfile
4444 Description: "page not covered by a lock" warnings from ll_readpage
4445 Details : sendfile called ll_readpage without right page locks present.
4446 Now we introduced ll_file_sendfile that does necessary locking
4447 around call to generic_file_sendfile() much like we do in
4451 Frequency : with certain MDS communication failures at client mount time
4453 Description: NULL pointer deref after failed client mount
4454 Details : a client connection request may delayed by the network layer
4455 and not be sent until after the PTLRPC layer has timed out the
4456 request. If the client fails the mount immediately it will try
4457 to clean up before the network times out the request. Add a
4458 reference from the request import to the obd device and delay
4459 the cleanup until the network drops the request.
4462 Frequency : occasionally during client (re)connect
4464 Description: assertion failure during client (re)connect
4465 Details : processing a client connection request may be delayed by the
4466 client or server longer than the client connect timeout. This
4467 causes the client to resend the connection request. If the
4468 original connection request is replied in this interval, the
4469 client may trip an assertion failure in ptlrpc_connect_interpret()
4470 which thought it would be the only running connect process.
4473 Frequency : only with obd_echo servers and clients that are rebooted
4475 Description: kernel BUG accessing uninitialized data structure
4476 Details : When running an obd_echo server it did not start the ping_evictor
4477 thread, and when a client was evicted an uninitialized data
4478 structure was accessed. Start the ping_evictor in the RPC
4479 service startup instead of the OBD startup.
4481 Severity : enhancement
4482 Bugzilla : 10193 (patchless)
4483 Description: Remove dependency on various unexported kernel interfaces.
4484 Details : No longer need reparent_to_init, exit_mm, exit_files,
4485 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4488 Frequency : rare (only users of deprecated and unsupported LDAP config)
4490 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4491 Details : LDAP apparently contains 'lustreName' attributes instead of
4492 'name'. A simple remapping of the name is sufficient.
4495 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4497 Description: waiting_locks_callback trips kernel BUG if client is evicted
4498 Details : Running with the dump_on_timeout debug flag turned on makes
4499 it possible that the waiting_locks_callback() can try to dump
4500 the Lustre kernel debug logs from an interrupt handler. Defer
4501 this log dumping to the expired_lock_main() thread.
4503 Severity : enhancement
4505 Description: Support NFS exporting on 2.6 kernels.
4506 Details : Implement non-rawops metadata methods for NFS server to use without
4507 changing NFS server code.
4510 Frequency : very rare (synthetic metadata workload only)
4512 Description: two racing renames might cause an MDS thread to deadlock
4513 Details : Running the "racer" program may cause one MDS thread to rename
4514 a file from being the source of a rename to being the target of
4515 a rename at exactly the same time that another thread is doing
4516 so, and the second thread has already enqueued these locks after
4517 doing a lookup of the target and is trying to relock them in
4518 order. Ensure that we don't try to re-lock the same resource.
4521 Frequency : only very large systems with liblustre clients
4523 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4524 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4525 all OSTs in parallel. This allows the eviction of stale liblustre
4526 clients to proceed much faster than if they were done in series,
4527 and also offers similar improvements for other set_info RPCs.
4532 Description: excessive CPU usage during initial read phase on client
4533 Details : During the initial read phase on a client, it would agressively
4534 retry readahead on the file, consuming too much CPU and impacting
4535 performance (since 1.4.5.8). Improve the readahead algorithm
4536 to avoid this, and also improve some other common cases (read
4537 of small files in particular, where "small" is files smaller than
4538 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4543 Description: MDS crash when receiving packet with unknown intent.
4544 Details : Do not LBUG in unknown intent case, just return -EFAULT
4546 Severity : enhancement
4547 Bugzilla : 9293, 9385
4548 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4549 Details : Do not serialize getattr (non-intent version) and statfs.
4552 Frequency : occasional, when OST network is overloaded/intermittent
4554 Description: client evicted by OST after bulk IO timeout
4555 Details : If a client sends a bulk IO request (read or write) the OST
4556 may evict the client if it is unresposive to its data GET/PUT
4557 request. This is incorrect if the network is overloaded (takes
4558 too long to transfer the RPC data) or dropped the OST GET/PUT
4559 request. There is no need to evict the client at all, since
4560 the pinger and/or lock callbacks will handle this, and the
4561 client can restart the bulk request.
4564 Frequency : Always when mmapping file with no objects
4566 Description: client crashes when mmapping file with no objects
4567 Details : Check that we actually have objects in a file before doing any
4568 operations on objects in ll_vm_open, ll_vm_close and
4574 Description: Request leak when working with deleted CWD
4575 Details : Introduce advanced request refcount tracking for requests
4576 referenced from lustre intent.
4578 Severity : Enhancement
4580 Description: Cache open file handles on client.
4581 Details : MDS now will return special lock along with openhandle, if
4582 requested and client is allowed to hold openhandle, even if unused,
4583 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4584 closing files for every read/write openration.
4586 Severity : Enhancement
4588 Description: Cache open negative dentries on client when possible.
4589 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4590 negative dentries on lock revocation.
4595 Description: Remounting a client read-only wasn't possible with a zconf mount
4596 Details : It wasn't possible to remount a client read-only with llmount.
4598 Severity : enhancement
4599 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4600 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4601 MPICH is included to add Lustre-specific ADIO interfaces.
4602 This is based closely on the UFS ADIO layer and only differs
4603 in file creation, in order to allow the OST striping to be set.
4604 This is user-contributed code and not supported by CFS.
4609 Description: extended inode attributes (immutable, append-only) work improperly
4610 when 2.4 and 2.6 kernels are used on client/server or vice versa
4611 Details : Introduce kernel-independent values for these flags.
4613 Severity : enhancement
4616 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4617 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4618 granularity. Now, Fractional MB granularity is be supported,
4619 this is very useful for low memory system.
4621 Severity : enhancement
4623 Description: Getattr by fid
4624 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4625 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4626 have locks and avoids getattr rpc requests for it on every stat.
4629 Frequency : Always, for filesystems larger than 2TB
4631 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4632 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4633 filesystems larger than 2TB to be mounted with mballoc enabled.
4636 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4638 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4639 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4640 filesystems larger than 2TB is mounted then the kernel will
4641 silently corrupt the start of the filesystem. CONFIG_LBD is
4642 enabled for all CFS-supported kernels, but the possibility of
4643 this happening with a modified kernel config exists.
4645 Severity : enhancement
4647 Description: add client O_DIRECT support for 2.6 kernels
4648 Details : It is now possible to do O_DIRECT reads and writes to files
4649 in the Lustre client mountpoint on 2.6 kernel clients.
4651 Severity : enhancement
4653 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4654 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4655 parallel, not waiting for response from every OST before sending
4656 a rpc to the next OST.
4661 Description: setattr vs write race when updating file timestamps
4662 Details : Client processes that update a file timestamp into the past
4663 right after writing to the file (e.g. tar) it is possible that
4664 the updated file modification time can be reset to the current
4665 time due to a race between processing the setattr and write RPC.
4667 Severity : enhancement
4669 Description: Bring 'lfs find' closer in line with regular Linux find.
4670 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4671 -print0 options and obtains all the needed info through the lustre
4674 Severity : enhancement
4676 Description: support up to 1024 configured devices on one node
4677 Details : change obd_dev array from statically allocated to dynamically
4678 allocated structs as they are first used to reduce memory usage
4683 Description: Flush dirty partially truncated pages during truncate
4684 Details : Immediatelly flush partially truncated pages in filter_setattr,
4685 this way we completely avoid having any pages in page cache on OST
4686 and can retire ugly workarounds during writes to flush such pages.
4691 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4692 Details : protect inode from truncation within vfs_unlink() context
4693 just take a reference before calling vfs_unlink() and release it
4694 when parent's i_sem is free.
4699 Description: last_id value checked outside lock on OST caused LASSERT failure
4700 Details : If there were multiple MDS->OST object precreate requests in
4701 flight, it was possible that the OST's last object id was checked
4702 outside a lock and incorrectly tripped an assertion. Move checks
4703 inside locks, and discard old precreate requests.
4706 Frequency : always, if extents are used on OSTs
4708 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4709 Details : OSTs using the extents format would not zero the high 16 bits of
4710 the index physical block number. This is not a problem for any
4711 OST filesystems smaller than 16TB, and no kernels support ext3
4712 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4713 new/modified files) and can be fixed for existing filesystems
4714 with e2fsprogs-1.39-cfs1.
4719 Description: import connection selection may be incorrect if timer wraps
4720 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4721 import connections to be ignored if the 32-bit jiffies counter
4722 wraps. Use a 64-bit jiffies counter.
4725 Frequency : very large clusters immediately after boot
4727 Description: LNET request buffers exhausted under heavy short-term load
4728 Details : If a large number of client requests are generated on a service
4729 that has previously never seen so many requests it is possible
4730 that the request buffer growth cannot keep up with the spike in
4731 demand. Instead of dropping incoming requests, they are held in
4732 the LND until the RPC service can accept more requests.
4735 Frequency : Sometimes during replay
4737 Description: Assertion failure in ll_local_open after replay.
4738 Details : If replay happened on an open request reply before we were able
4739 to set replay handler, reply will become not swabbed tripping the
4740 assertion in ll_local_open. Now we set the handler right after
4741 recognising of open request
4744 Frequency : very rare
4746 Description: kernel reports "badness in vsnprintf"
4747 Details : Reading from the "recovery_status" /proc file in small chunks
4748 may cause a negative length in lprocfs_obd_rd_recovery_status()
4749 call to vsnprintf() (which is otherwise harmless). Exit early
4750 if there is no more space in the output buffer.
4752 Severity : enhancement
4754 Description: clear OBD RPC statistics by writing to them
4755 Details : It is now possible to clear the OBD RPC statistics by writing
4756 to the "stats" file.
4761 Description: Client mtime is not the same on different clients after utimes
4762 Details : In some cases, the client was using the utimes() syscall on
4763 a file cached on another node. The clients now validate the
4764 ctime from the MDS + OSTs to determine which one is right.
4769 Description: Inability to activate failout mode
4770 Details : lconf script incorrectly assumed that in pythong string's numeric
4771 value is used in comparisons.
4774 Frequency : always with multiple stripes per file
4776 Description: Inefficient object allocation for mutli-stripe files
4777 Details : When selecting which OSTs to stripe files over, for files with
4778 a stripe count that divides evenly into the number of OSTs,
4779 the MDS is always picking the same starting OST for each file.
4780 Return the OST selection heuristic to the original design.
4785 Description: mount failures may take full timeout to return an error
4786 Details : Under some heavy load conditions it is possible that a
4787 failed mount can wait for the full obd_timeout interval,
4788 possibly several minutes, before reporting an error.
4789 Instead return an error as soon as the status is known.
4791 Frequency : quota enabled and large files being deleted
4793 Description: releasing more than 4GB of quota at once hangs OST
4794 Details : If a user deletes more than 4GB of files on a single OST it
4795 will cause the OST to spin in an infinite loop. Release
4796 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4801 Description: statfs data retrieved from /proc may be stale or zero
4802 Details : When reading per-device statfs data from /proc, in the
4803 {kbytes,files}_{total,free,avail} files, it may appear
4804 as zero or be out of date.
4807 Frequency : systems with MD RAID1 external journal devices
4809 Description: lconf's call to blkid is confused by RAID1 journal devices
4810 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4811 of returning all block devices that match the journal UUID.
4814 Frequency : always, for aggregate stripe size over 4GB
4816 Description: assertion fails when trying to use 4GB stripe size
4817 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4818 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4821 Frequency : always on ppc64
4823 Description: the first write on an ext3 filesystem with mballoc got stuck
4824 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4825 perform endianness conversion.
4827 ------------------------------------------------------------------------------
4829 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4831 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4832 INTEROPERATE with older versions automatically. Please read the
4833 user documentation before upgrading any part of a live system.
4834 * WARNING: Lustre networking configuration changes are required with
4835 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4838 * Support for kernels:
4839 2.6.9-22.0.2.EL (RHEL 4)
4840 2.6.5-7.244 (SLES 9)
4841 2.6.12.6 vanilla (kernel.org)
4844 Severity : enhancement
4845 Bugzilla : 7981/8208
4846 Description: Introduced Lustre Networking (LNET)
4847 Details : LNET is new networking infrastructure for Lustre, it includes
4848 a reorganized network configuration mode (see the user
4849 documentation for full details) as well as support for routing
4850 between different network fabrics. Lustre Networking Devices
4851 (LNDs) for the supported network fabrics have also been
4852 created for this new infrastructure.
4854 Severity : enhancement
4855 Description: Introduced Access control lists
4856 Details : clients can set ACLs on files and directories in order to have
4857 more fine-grained permissions than the standard Unix UGO+RWX.
4858 The MDS must be started with the "-o acl" mount option.
4860 Severity : enhancement
4861 Description: Introduced filesystem quotas
4862 Details : Administrators may now establish per-user quotas on the
4865 Severity : enhancement
4867 Description: Configuration change for the XT3
4868 The PTLLND is now used to run Lustre over Portals on the XT3
4869 The configure option(s) --with-cray-portals are no longer used.
4870 Rather --with-portals=<path-to-portals-includes> is used to
4871 enable building on the XT3. In addition to enable XT3 specific
4872 features the option --enable-cray-xt3 must be used.
4877 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4878 Details : A race between allocating a new llcd and re-getting the llcd_lock
4879 allowed another thread to grab newly-allocated llcd.
4881 Severity : enhancement
4883 Description: 2.6 OST async journal commit and locking fix to improve performance
4884 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4885 2.6 kernels are now async as they already were in 2.4 kernels so
4886 that they can commit concurrently with the network bulk transfer.
4887 For block-allocated files the filter allocation semaphore is held
4888 to avoid filesystem fragmentation during allocation. BKL lock
4889 removed for 2.6 xattr operations where it is no longer needed.
4894 Description: lconf incorrectly determined whether two IP networks could talk
4895 Details : In some more complicated routing and multiple-network
4896 configurations, lconf will avoid trying to make a network
4897 connection to a disjoint part of the IP space. It was doing the
4898 math incorrectly for one set of cases.
4903 Description: Fix for potential infinite loop processing records in an llog.
4904 Details : If an llog record is corrupted/zeroed, it is possible to loop
4905 forever in llog_process(). Validate the llog record length
4906 and skip the remainder of the block on error.
4909 Frequency : occasional (liblustre only)
4911 Description: liblustre could not open files whose last component is a symlink
4912 Details : sysio_path_walk() would incorrectly pass the open intent to
4913 intermediate path components.
4916 Frequency : rare (liblustre only with non-standard tuning)
4917 Bugzilla : 7201 (7350)
4918 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4919 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4920 after it has been released, only do this for referenced locks
4922 Severity : enhancement
4924 Description: specifying an (invalid) directory default stripe_size of -1
4925 would reset the directory default striping
4926 Details : stripe_size -1 was used internally to signal directory stripe
4927 removal, now use "all default" to signal dir stripe removal
4928 as a directory striping of "all default" is not useful
4931 Frequency : common for large clusters running liblustre clients
4933 Description: doing an ls when liblustre clients are running is slow
4934 Details : sending a glimpse AST to a liblustre client waits for every AST
4935 to time out, as liblustre clients will not respond. Since they
4936 cannot cache data we refresh the OST lock LVB from disk instead.
4938 Severity : enhancement
4940 Description: doing an ls at the same time as file IO can be slow
4941 Details : enqueue and other "small" requests can be blocked behind many
4942 large IO requests. Create a new OST IO portal for non-IO
4943 requests so they can be processed faster.
4946 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4948 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4949 created file with mode 000
4950 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4951 allow an ftruncate() as a non-root user to a file with mode 000.
4952 Lustre now respects this flag to disable mode checks when
4953 truncating a file owned by the user
4956 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4959 Description: Revoking locks from clients that went dead or catatonic might take
4961 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4962 cancellation of such locks instant on servers without waiting for
4963 any reply from clients. Clients drops these locks when cancel
4964 notification from server is received without replying.
4967 Frequency : liblustre-only, when liblustre client dies or becomes busy
4969 Description: Doing ls on Linux clients can take a long time with active
4971 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4972 granting such locks to it in the first place if possible. Locks
4973 are taken by proxy on the OST during the read or write and
4974 dropped immediately afterward. Add connect flags handling, do
4975 not grant locks to liblustre clients for glimpse ASTs.
4977 Severity : enhancement
4979 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4980 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4981 sequential reads and behaves badly with applications doing random
4982 reads. Improve it by reading ahead at least read region, and
4983 avoiding excessive large RPC for small reads.
4985 Severity : enhancement
4987 Description: Creating more than 1000 files for a single job may cause a load
4988 imbalance on the OSTs if there are also a large number of OSTs.
4989 Details : qos_prep_create() uses an OST index reseed value that is an
4990 even multiple of the number of available OSTs so that if the
4991 reseed happens in the middle of the object allocation it will
4992 still utilize the OSTs as uniformly as possible.
4997 Description: OST or MDS may oops in ping_evictor_main()
4998 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4999 but doesn't restart at beginning of obd_exports_timed list
5002 Severity : enhancement
5004 Description: improve by-nid export eviction on the MDS and OST
5005 Details : allow multiple exports with the same NID to be evicted at one
5006 time without re-searching the exports list.
5009 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5011 Description: MDS may oops in groups_free()
5012 Details : in rare race conditions a newly allocated group_info struct is
5013 freed again, and this can be NULL. The 2.4 compatibility code
5014 for groups_free() checked for a NULL pointer, but 2.6 did not.
5017 Frequency : common for liblustre clients doing little filesystem IO
5018 Bugzilla : 9352, 7313
5019 Description: server may evict liblustre clients accessing contended locks
5020 Details : if a client is granted a lock or receives a completion AST
5021 with a blocking AST already set it would not reply to the AST
5022 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5025 Frequency : lfs setstripe, only systems with more than 160 OSTs
5027 Description: unable to set striping with a starting offset beyond OST 160
5028 Details : llapi_create_file() incorrectly limited the starting stripe
5029 index to the maximum single-file stripe count.
5032 Frequency : LDAP users only
5034 Description: lconf did not handle in-kernel recovery with LDAP properly
5035 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5037 Severity : enhancement
5039 Description: bind OST threads to NUMA nodes to improve performance
5040 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5041 node and do their allocations there to localize memory access
5043 Severity : enhancement
5045 Description: llmount can determine client NID directly from Myrinet (GM)
5046 Details : the client NID code from gmnalnid was moved directly into
5047 llmount, removing the need to use this or specifying the
5048 client NID explicitly when mounting GM clients with zeroconf
5051 Frequency : if client is started with down MDS
5053 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5054 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5055 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5056 ptlrpc_import_delayed_req() because we didn't check if the
5057 request was interrupted, and we also didn't break out of the
5058 event loop if there was no timeout
5063 Description: data loss during non-page-aligned writes to a single file from
5064 both multiple nodes and multiple threads on one node at same time
5065 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5066 inconsistency led to false short-reads, that were cached and later
5067 used by ->prepare_write() to fill in partially written page,
5068 leading to data loss.
5071 Frequency : always, if lconf --abort_recovery used
5073 Description: lconf --abort_recovery fails with 'Operation not supported'
5074 Details : lconf was attempting to abort recovery on the MDT device and not
5077 Severity : enhancement
5079 Description: remove cleanup logs
5080 Details : replace lconf-generated cleanup logs with lustre internal
5081 cleanup routines. Eliminates the need for client-cleanup and
5084 Severity : enhancement
5086 Description: add support for EAs (user and system) on lustre filesystems
5087 Details : it is now possible to store extended attributes in the Lustre
5088 client filesystem, and with the user_xattr mount option it
5089 is possible to allow users to store EAs on their files also
5091 Severity : enhancement
5093 Description: Add possibility (config option) to show minimal available OST free
5095 Details : When compiled with --enable-mindf configure option, statfs(2)
5096 (and so, df) will return least minimal free space available from
5097 all OSTs as amount of free space on FS, instead of summary of
5098 free spaces of all OSTs.
5100 Severity : enhancement
5102 Description: do not expand extent locks acquired on OST-side
5103 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5104 by server: they are not cached anyway.
5107 Frequency : when mmap is used/binaries executed from Lustre
5109 Description: Unmmap pages before throwing them away from read cache.
5110 Details : llap_shrink cache now attempts to unmap pages before discarding
5111 them (if unmapping failed - do not discard). SLES9 kernel has
5112 extra checks that trigger if this unmapping is not done first.
5117 Description: lconf didn't resolve symlinks before checking to see whether a
5118 given mountpoint was already in use
5121 Frequency : when migrating failover services
5122 Bugzilla : 6395, 9514
5123 Description: When migrating a subset of services from a node (e.g. failback
5124 from a failover service node) the remaining services would
5125 time out and evict clients.
5126 Details : lconf --force (implied by --failover) sets the global obd_timeout
5127 to 5 seconds in order to quickly disconnect, but this caused
5128 other RPCs to time out too quickly. Do not change the global
5129 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5131 Severity : enhancement
5132 Frequency : if MDS is started with down OST
5133 Bugzilla : 9439,5706
5134 Description: Allow startup/shutdown of an MDS without depending on the
5135 availability of the OSTs.
5136 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5137 Add appropriate locking and lov-osc refcounts for safe
5138 cleaning. Add osc abort_inflight calls in case the
5139 synchronize never started.
5142 Frequency : occasional (Cray XT3 only)
5144 Description: root not authorized to access files in CRAY_PORTALS environment
5145 Details : The client process capabilities were not honoured on the MDS in
5146 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5147 been accessed by an authorized user then root was able to access
5148 the file on the local client also. The root user capabilities
5149 are now allowed on the MDS, as this environment has secure UID.
5152 Frequency : occasional
5154 Description: ldiskfs "too long searching" message happens too often
5155 Details : A debugging message (otherwise harmless) prints too often on
5156 the OST console. This has been reduced to only happen when
5157 there are fragmentation problems on the filesystem.
5162 Description: Division by zero in statfs when all OSCs are inactive
5163 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5164 lov_statfs divides by value returned by lov_get_stripecnt().
5168 Bugzilla : 9489, 3273
5169 Description: First write from each client to each OST was only 4kB in size,
5170 to initialize client writeback cache, which caused sub-optimal
5171 RPCs and poor layout on disk for the first written file.
5172 Details : Clients now request an initial cache grant at (re)connect time
5173 and so that they can start streaming writes to the cache right
5174 away and always do full-sized RPCs if there is enough data.
5175 If the OST is rebooted the client also re-establishes its grant
5176 so that client cached writes will be honoured under the grant.
5181 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5182 Details : Now I/O RPCs go to different portal number and (presumably) fast
5183 lock requests (and glimses) and other RPCs get their own service
5184 threads pool that should be able to service those RPCs
5187 Severity : enhancement
5189 Description: Ability to exchange lustre version between client and servers and
5190 issue warnings at client side if client is too old. Also for
5191 liblustre clients there is ability to refuse connection of too old
5193 Details : New 'version' field is added to connect data structure that is
5194 filled with version info. That info is later checked by server and
5198 Frequency : rare, liblustre only.
5199 Bugzilla : 9296, 9581
5200 Description: Two simultaneous writes from liblustre at offset within same page
5201 might proceed at the same time overwriting eachother with stale
5203 Details : I/O lock withing llu_file_prwv was released too early, before data
5204 actually was hitting the wire. Extended lock-holding time until
5205 server acknowledges receiving data.
5208 Frequency : extremely rare. Never observed in practice.
5210 Description: avoid generating lustre_handle cookie of 0.
5211 Details : class_handle_hash() generates handle cookies by incrementing
5212 global counter, and can hit 0 occasionaly (this is unlikely, but
5213 not impossible, because initial value of cookie counter is
5214 selected randonly). Value of 0 is used as a sentinel meaning
5215 "unassigned handle" --- avoid it. Also coalesce two critical
5216 sections in this function into one.
5218 Severity : enhancement
5220 Description: allow liblustre clients to delegate truncate locking to OST
5221 Details : To avoid overhead of locking, liblustre client instructs OST to
5222 take extent lock in ost_punch() on client's behalf. New connection
5223 flag is added to handle backward compatibility.
5225 Severity : enhancement
5226 Bugzilla : 4928, 7341, 9758
5227 Description: allow number of OST service threads to be specified
5228 Details : a module parameter allows the number of OST service threads
5229 to be specified via "options ost ost_num_threads={N}" in the
5230 OSS's /etc/modules.conf or /etc/modprobe.conf.
5234 Bugzilla : 6146, 9635, 9895
5235 Description: servers crash with bad pointer in target_handle_connect()
5236 Details : In rare cases when a client is reconnecting it was possible that
5237 the connection request was the last reference for that export.
5238 We would temporarily drop the export reference and get a new
5239 one, but this may have been the last reference and the export
5240 was just destroyed. Get new reference before dropping old one.
5242 Severity : enhancement
5243 Frequency : if client is started with failover MDS
5245 Description: Allow multiple MDS hostnames in the mount command
5246 Details : Try to read the configuration from all specified MDS
5247 hostnames during a client mount in case the "primary"
5250 Severity : enhancement
5252 Description: Stop sending data to evicted clients as soon as possible.
5253 Details : Check if the client we are about to send or are sending data to
5254 was evicted already. (Check is done every second of waiting,
5255 for which l_wait_event interface was extended to allow checking
5256 of exit condition at specified intervals).
5259 Frequency : rare, normally only when NFS exporting is done from client
5261 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5263 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5266 Frequency : always, if srand() is called before liblustre initialization
5268 Description: Liblustre uses system PRNG disturbing its usage by user application
5269 Details : Introduce internal to lustre fast and high-quality PRNG for
5270 lustre usage and make liblustre and some other places in generic
5271 lustre code to use it.
5273 Severity : enhancement
5274 Bugzilla : 9477, 9557, 9870
5275 Description: Verify that the MDS configuration logs are updated when xml is
5276 Details : Check if the .xml configuration logs are newer than the config
5277 logs stored on the MDS and report an error if this is the case.
5278 Request --write-conf, or allow starting with --old_conf.
5280 Severity : enhancement
5282 Description: Handle symlinks in the path when checking if Lustre is mounted.
5283 Details : Resolve intermediate symlinks when checking if a client has
5284 mounted a filesystem to avoid duplicate client mounts.
5289 Description: lconf can hit an error exception but still return success.
5290 Details : The lconf command catches the Command error exception at the top
5291 level script context and will exit with the associated exit
5292 status, but doesn't ensure that this exit status is non-zero.
5297 Description: failure of ptlrpc thread startup can cause oops
5298 Details : Starting a ptlrpc service thread can fail if there are a large
5299 number of threads or the server memory is very fragmented.
5300 Handle this without oopsing.
5303 Frequency : always, only if liblustre and non-default acceptor port was used
5305 Description: liblustre cannot connect to servers with non-default acceptor port
5306 Details : tcpnal_set_default_params() was not called and was therefore
5307 ignoring the environment varaible TCPNAL_PORT, as well as other
5308 TCPNAL_ environment variables
5313 Description: two objects could be created on the same OST for a single file
5314 Details : If an OST is down, in some cases it was possible to create two
5315 objects on a single OST for a single file. No problems other
5316 than potential performance impact and spurious error messages.
5320 Bugzilla : 5681, 9562
5321 Description: Client may oops in ll_unhash_aliases
5322 Details : Client dcache may become inconsistent in race condition.
5323 In some cases "getcwd" can fail if the current directory is
5329 Description: Inode refcounting problems in NFS export code
5330 Details : link_raw functions used to call d_instantiate without obtaining
5331 extra inode reference first.
5335 Bugzilla : 9942, 9903
5336 Description: Referencing freed requests leading to crash, memleaks with NFS.
5337 Details : We used to require that call to ll_revalidate_it was always
5338 followed by ll_lookup_it. Also with revalidate_special() it is
5339 possible to call ll_revalidate_it() twice for the same dentry
5340 even if first occurence returned success. This fix changes semantic
5341 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5342 reference on a request referred from the intent.
5343 ll_intent_release() then releases such a request.
5346 Frequency : rare, normally benchmark loads only
5348 Description: unlinked inodes were kept in memory on the client
5349 Details : If a client is repeatedly creating and unlinking files it
5350 can accumulate a lot of stale inodes in the inode slab cache.
5351 If there is no other client load running this can cause the
5352 client node to run out of memory. Instead flush old inodes
5353 from client cache that have the same inode number as a new inode.
5356 Frequency : SLES9 2.6.5 kernel and long filenames only
5357 Bugzilla : 9969, 10379
5358 Description: utime reports stale NFS file handle
5359 Details : SLES9 uses out-of-dentry names in some cases, which confused
5360 the lustre dentry revalidation. Change it to always use the
5364 Frequency : rare, unless heavy write-truncate concurrency is continuous
5365 Bugzilla : 4180, 6984, 7171, 9963, 9331
5366 Description: OST becomes very slow and/or deadlocked during object unlink
5367 Details : filter_destroy() was holding onto the parent directory lock
5368 while truncating+unlinking objects. For very large objects this
5369 may block other threads for a long time and slow overall OST
5370 responsiveness. It may also be possible to get a lock ordering
5371 deadlock in this case, or run out of journal credits because of
5372 the combined truncate+unlink. Solution is to do object truncate
5373 first in one transaction without parent lock, and then do the
5374 final unlink in a new transaction with the parent lock. This
5375 reduces the lock hold time dramatically.
5378 Frequency : rare, 2.4 kernels only
5380 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5381 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5382 before unmounting their filesystem in order to avoid deadlock.
5383 The kernel_locked() function in 2.4 kernels only checks whether
5384 the kernel lock is held, not whether it is this process that is
5385 holding it as 2.6 kernels do.
5390 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5391 Details : The client ptlrpc code may be trying to reconnect to a down
5392 server before a previous connection attempt has timed out.
5393 Increase the reconnect interval to be longer than the connection
5394 timeout interval to avoid sending duplicate connections to
5398 Frequency : echo_client brw_test command
5400 Description: fix echo_client to work with OST preallocated code
5401 Details : OST preallocation code (5137) didn't take echo_client IO path
5402 into account: echo_client calls filter methods outside of any
5403 OST thread and, hence, there is no per-thread preallocated
5404 pages and buffers to use. Solution: hijack pga pages for IO. As
5405 a byproduct, this avoids unnecessary data copying.
5409 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5410 Description: Client can oops in mdc_commit_close() after open replay
5411 Details : It was possible for the MDS to return an open request with no
5412 transaction number in mds_finish_transno() if the client was
5413 evicted, but without actually returning an error. Clients
5414 would later try to replay that open and may trip an assertion
5415 Simplify the client close codepath, and always return an error
5416 from the MDS in case the open is not successful.
5419 Frequency : rare, 2.6 OSTs only
5421 Description: OST may deadlock under high load on fragmented files
5422 Details : If there was a heavy load and highly-fragmented OST filesystems
5423 it was possible to have all the OST threads deadlock waiting on
5424 allocation of biovecs, because the biovecs were not released
5425 until the entire RPC IO was completed. Instead, release biovecs
5426 as soon as they are complete to ensure forward IO progress.
5428 Severity : enhancement
5430 Description: Support for specifying external journal device at mount
5431 Details : If an OST or MDS device is formatted with an external journal
5432 device, this device major/minor is stored in the ext3 superblock
5433 and may not be valid for failover. Allow detecting and
5434 specifying the external journal at mount time.
5439 Description: Mounting an MDS with pending unlinked files may cause oops
5440 Details : target_finish_recovery() calls mds_postrecov() which returned
5441 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5442 considers this an error and immediately begins cleaning up the
5443 lov, just after starting the mds_lov process
5445 Severity : enhancement
5447 Description: Implement 'lfs df' to report actual free space on per-OST basis
5448 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5449 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5450 usage of objects; '-h' to report in human readable format.
5452 ------------------------------------------------------------------------------
5454 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5461 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5462 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5463 at startup during orphan unlinking. Instead, skip buddy update
5464 before setup, as it will be regenerated after recovery is complete.
5469 Description: If an OST is inactive, its locks might reference stale inodes.
5470 Details : lov_change_cbdata() must iterate over all namespaces, even if
5471 they are inactive to clear inode references from the lock.
5473 Severity : enhancement
5474 Frequency : occasional, if non-standard max_dirty_mb used
5476 Description: Client will block write RPCs if not enough grant
5477 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5478 then the client will block writes while waiting for another RPC
5479 to complete instead of consuming its dirty limit. With change
5480 we get improved performance when max_dirty_mb is small.
5482 Severity : enhancement
5483 Bugzilla : 3389, 6253
5484 Description: Add support for supplementary groups on the MDS.
5485 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5486 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5487 lookups for user supplementary groups into a cache.
5491 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5492 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5493 for liblustre O_EXCL is enforced by MDS.
5497 Description: readdir never returns NULL in liblustre.
5498 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5499 dentry in struct dirent.
5503 Frequency : liblustre only, depends on application IO pattern
5504 Description: liblustre clients evicted if not contacting servers
5505 Details : Don't put liblustre clients into the ping_evictor list, so
5506 they will not be evicted by the pinger ever.
5508 Severity : enhancement
5510 Description: Add ability to evict clients by NID from MDS.
5511 Details : By echoing "nid:$NID" string into
5512 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5513 $NID would be instantly evicted from this MDS and from all active
5514 OSTs connected to it.
5518 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5519 Details : lookup_it_finish() used to query file size from OSTs that was not
5524 Description: service threads change working directory to that of init
5525 Details : Starting lustre service threads may pin the working directory
5526 of the parent thread, making that filesystem busy. Threads
5527 now change to the working directory of init to avoid this.
5531 Frequency : during shutdown only
5532 Description: shutdown with a failed MDS or OST can cause unmount to hang
5533 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5538 Frequency : 2.6 only
5539 Description: chmod/chown may include an extra supplementary group
5540 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5541 supplementary group and if none is specified this is used.
5544 Bugzilla : 5479 (6816)
5546 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5547 Details : If lookup is in progress on a file that is unlinked we might try
5548 to revalidate the inode and fail in revalidate after lookup is
5549 complete and ll_file_open() enqueues the open again but
5550 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5553 Frequency : always, if lconf --abort_recovery used
5555 Description: lconf --abort_recovery fails with 'Operation not supported'
5556 Details : lconf was attempting to abort recovery on the MDT device and not
5559 ------------------------------------------------------------------------------
5561 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5566 Frequency : rare (only unsupported configurations with a node running as an
5568 Bugzilla : 6514, 5137
5569 Description: Mounting a Lustre file system on a node running as an OST could
5571 Details : OSTs now preallocates memory needed to write out data at
5572 startup, instead of when needed, to avoid having to
5573 allocate memory in possibly low memory situations.
5574 Specifically, if the file system is mounted on on OST,
5575 memory pressure could force it to try to write out data,
5576 which it needed to allocate memory to do. Due to the low
5577 memory, it would be unable to do so and the node would
5578 become unresponsive.
5580 Severity : enhancement
5582 Description: Addition of lconf --service command line option
5583 Details : lconf now accepts a '--service <arg>' option, which is
5584 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5586 Severity : enhancement
5588 Description: Failover mode is now the default for OSTs.
5589 Details : By default, OSTs will now run in failover mode. To return to
5590 the old behaviour, add '--failout' to the lmc line for OSTs.
5592 Severity : enhancement
5594 Description: Health checks are now provided for MDS and OSTs
5595 Details : Additional detailed health check information on MSD and OSTs
5596 is now provided through the procfs health_check value.
5599 Frequency : occasional, depends on IO load
5601 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5602 after numerous create/delete cycles
5603 Details : The ext3 inode allocation policy would not allocate new inodes
5604 very well on the OSTs because there are no new directories
5605 being created. Instead we look for groups with free space if
5606 the parent directories are nearly full.
5611 Description: Network or server problems during mount may cause partially
5612 mounted clients instead of returning an error.
5613 Details : The config llog parsing code may overwrite the error return
5614 code during mount error handling, returning success instead
5620 Description: MDS can fail to allocate large reply buffers
5621 Details : After long uptimes the MDS can fail to allocate large reply
5622 buffers (e.g. zconf client mount config records) due to memory
5623 fragmentation or consumption by the buffer cache. Preallocate
5624 some large reply buffers so that these replies can be sent even
5625 under memory pressure.
5629 Frequency : rare (liblustre)
5630 Description: fsx running with liblustre complained that using truncate() to
5631 extend the file doesn't work. This patch corrects that issue.
5632 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5633 ATTR_SIZE and lsm use in llu_setattr_raw.
5637 Frequency : rare, only 2.6 kernels
5638 Description: Unusual file access patterns on the MDS may result in inode
5639 data being lost in very rare circumstances.
5640 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5641 code in the 2.6 kernel caused the inode and/or EA data not to
5642 be read from disk, causing single-file corruption.
5646 Frequency : rare, only 2.6 filesystems using extents
5647 Description: Heavy concurrent write and delete load may cause data corruption.
5648 Details : It was possible under high-load situations to have an extent
5649 metadata block in the block device cache from a just-unlinked
5650 file overwrite a newly-allocated data block. We now unmap any
5651 metadata buffers that alias just-allocated data blocks.
5655 Frequency : filesystems with default stripe_count larger than 77
5656 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5657 Details : lconf specifies an inode size of 4096 bytes when the default
5658 stripe_count is larger than 77. This conflicts with the default
5659 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5660 this case to avoid pinning too much memory for large EAs.
5662 ------------------------------------------------------------------------------
5664 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5669 Frequency : rare (extremely heavy IO load with hundreds of clients)
5671 Description: Client is evicted, gets IO error writing to file
5672 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5673 caused clients to be evicted for AST timeouts. The fixes in
5674 bug 5192 mean we no longer need to have such short AST timeouts
5675 so ldlm_timeout has been increased.
5678 Frequency : occasional during --force or --failover shutdown under load
5679 Bugzilla : 5949, 4834
5680 Description: Server oops/LBUG if stopped with --force or --failover under load
5681 Details : a collection of import/export refcount and cleanup ordering
5682 issues fixed for safer force cleanup
5685 Frequency : only filesystems larger than 120 OSTs
5686 Bugzilla : 5990, 6223
5687 Description: lfs getstripe would oops on a very large filesystem
5688 Details : lov_getconfig used kfree on vmalloc'd memory
5691 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5692 Bugzilla : 6242, 6243
5693 Description: reading from files that had been truncated to a non-zero size
5694 but never opened returned no data
5695 Details : ll_file_read() reads zeros from no-object files to EOF
5700 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5701 Details : There's little harm in aborting MDS/OSS recovery and letting it
5702 try again, so I removed the LASSERT and return an error instead.
5704 Severity : enhancement
5706 Description: New debugging infrastructure for tracking down data corruption
5707 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5708 (b) cover more of the client-side code path, and (c) try to narrow
5709 down where problems occurred
5713 Bugzilla : 3819, 4364, 4397, 6313
5714 Description: Racing close and eviction MDS could cause assertion in mds_close
5715 Details : It was possible to get multiple mfd references during close and
5716 client eviction, leading to one thread referencing a freed mfd.
5718 Severity: : enhancement
5719 Bugzilla : 3262, 6359
5720 Description: Attempts to reconnect to servers are now more aggressive.
5721 Details : This builds on the enhanced upcall-less recovery that was added
5722 in 1.4.2. When trying to reconnect to servers, clients will
5723 now try each server in the failover group every 10 seconds. By
5724 default, clients would previously try one server every 25 seconds.
5729 Description: After recovery, certain operations trigger a failed
5730 assertion on a client.
5731 Details : Failing over an mds, using lconf -d --failover, while a
5732 client was doing a readdir() call would cause the client to
5733 LBUG after recovery completed and the readdir() was resent.
5735 Severity : enhancement
5737 Description: Default groups are now added by lconf
5738 Details : You can now run lconf --group <servicename> without having to
5739 manually add groups with lmc.
5742 Frequency : occasional
5744 Description: Nodes with an elan id of 0 trigger a failed assertion
5747 Frequency : always when accessing e.g. tty/console device nodes
5749 Description: tty and some other devices nodes cannot be used on lustre
5750 Details : file's private_data field is used by device data and lustre
5751 values in there got lost. New field was added to struct file to
5752 store fs-specific private data.
5755 Frequency : when exporting Lustre via NFS
5757 Description: NFSD failed occasionally when looking up a path component
5758 Details : NFSD is looking up ".." which was broken in ext3 directories
5759 that had grown large enough to become hashed.
5762 Frequency : Clusters with multiple interfaces not on the same subnet
5764 Description: Nodes will repeatedly try to reconnect to an interface which it
5765 cannot reach and report an error to the log.
5766 Details : Extra peer list entries will be created by lconf with some peers
5767 unreachable. lconf now validates the peer before adding it.
5770 Frequency : Only if a default stripe is set on the filesystem root.
5772 Description: Setting a default stripe on the filesystem root prevented the
5773 filesystem from being remounted.
5774 Details : The client was sending extra request flags in the root getattr
5775 request and did not allocate a reply buffer for the dir EA.
5778 Frequency : occasional, higher if lots of files are accessed by one client
5779 Bugzilla : 6159, 6097
5780 Description: Client trips assertion regarding lsm mismatch/magic
5781 Details : While revalidating inodes the VFS looks up inodes with ifind()
5782 and in rare cases can find an inode that is being freed.
5783 The ll_test_inode() code will free the lsm during ifind()
5784 when it finds an existing inode and then the VFS later attaches
5785 this free lsm to a new inode.
5789 Bugzilla : 6422, 7030
5790 Description: MDS deadlock between mkdir and client eviction
5791 Details : Creating a new file via mkdir or mknod (starting a transaction
5792 and getting the ns lock) can deadlock with client eviction
5793 (gets ns lock and trying to finish a synchronous transaction).
5796 Frequency : occasional
5797 Description: While starting a server, the fsfilt_ext3 module could not be
5799 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5800 kernels. Previously, lconf would still use the ext3 name
5801 when trying to load modules. Now, it will correctly use
5802 ext3 on 2.4 and ldiskfs on 2.6.
5804 Severity : enhancement
5805 Description: The default stripe count has been changed to 1
5806 Details : The interpretation of the default stripe count (0, to lfs
5807 or lmc) has been changed to mean striping across a single
5808 OST, rather than all available. For general usage we have
5809 found a stripe count of 1 or 2 works best.
5811 Severity : enhancement
5812 Description: Add support for compiling against Cray portals.
5813 Details : Conditional compiling for some areas that are different
5817 Frequency : occasional
5818 Bugzilla : 6409, 6834
5819 Description: Creating files with an explicit stripe count may lead to
5820 a failed assertion on the MDS
5821 Details : If some OSTs are full or unavailable, creating files may
5822 trigger a failed assertion on the MDS. Now, Lustre will
5823 try to use other servers or return an error to the
5827 Frequency : occasional
5829 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5830 caused lock timeout during readahead (since 1.4.2).
5831 Details : Processes doing readahead might match a lock that hasn't been
5832 granted yet if there are overlapping and conflicting lock
5833 requests. The readahead process waits on ungranted lock
5834 (original lock is CBPENDING), while OST waits for that process
5835 to cancel CBPENDING read lock and eventually evicts client.
5837 Severity : enhancement
5839 Description: Initial enabling of flock support for clients
5840 Details : Implements fcntl advisory locking and file status functions.
5841 This feature is provided as an optional mount flag (default
5842 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5843 locking are implemented yet, and those that are are not guaranteed
5844 to be completely correct in production environments.
5845 mount -t lustre -o [flock|noflock] ...
5848 Frequency : occasional
5850 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5851 assertion in the ext3 JBD (journaling) layer.
5852 Details : The b_committed_data struct is protected by the big kernel lock
5853 in 2.4 kernels, serializing journal_commit_transaction() and
5854 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5855 access to this struct. In 2.6 kernels there is finer grained
5856 locking to improve SMP performance of the JBD layer.
5860 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5865 Description: merge and cleanup kernel patches.
5868 -----------------------------------------------------------------------------
5870 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5872 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5873 versions of Lustre. Please update all servers and clients to
5874 version 1.4.2 or later at the same time. You must also run
5875 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5876 to update the configuration logs.
5878 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5879 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5880 - swab llog records of type '0' so we get proper header size/idx (5861)
5881 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5882 - fix rename of one directory over another leaking an inode (5953)
5883 - avoid SetPageDirty on 2.6 (5981)
5884 - don't re-add just-being-destroyed locks to the waiting list (5653)
5885 - when creating new directories, inherit the parent's custom
5886 striping settings if present parent (3048)
5887 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5888 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5889 - don't hold client locks on temporary worklist from l_lru (5666)
5890 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5891 - automatically evict dead clients (5921)
5892 - Update file size properly in create+truncate+fstat case (6196)
5893 - Do not unhash mountpoint dentries, do not allow removal of
5895 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5896 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5897 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5898 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5899 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5900 - Correctly handle reads of files with no objects (6243)
5901 - lctl recover will also mark a device active if deactivate used (5933)
5903 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5904 - allow --write-conf on an MDS with different nettype than client (5619)
5905 - don't write config llogs to MDS for mounts not from that MDS (5617)
5906 - lconf should create multiple TCP connections from a client (5201)
5907 - init scripts are now turned off by default; run chkconfig --on
5908 lustre and chkconfig --on lustrefs to use them
5909 - upcalls are no longer needed for clients to recover to failover
5911 - add --abort-recovery option to lconf to abort recovery on device
5913 - add support for an arbitrary number of OSTs (3026)
5914 - Quota support protocol changes.
5915 - forward compatibility changes to wire structs (6007)
5916 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5917 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5918 - improved "lustre status" script
5919 - initialize blocksize for non-regular files (6062)
5920 - added --disable-server and --disable-client configure options (5782)
5921 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5922 - Vanilla 2.4.29 support
5923 - increase maximum number of obd devices to 520 (6242)
5924 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5925 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5926 SLES 9 2.6.5 kernels
5927 - limit stripes per file to 160 (the maximum EA size) (6093)
5929 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5932 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5933 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5934 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5935 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5936 - small ext3 extents cleanups and fixes (5733)
5937 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5938 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5939 - avoid needless client->OST connect, fix handle mismatch (5317)
5940 - fix DLM error path that led to out-of-sync client, long delays (5779)
5941 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5942 - fix several locking issues related to i_size (5492,5624,5654,5672)
5943 - don't move pending lock onto export if it is already evicted (5683)
5944 - fix kernel oops when creating .foo in unlinked directory (5548)
5945 - fix deadlock in obdfilter statistics vs. object create (5811)
5946 - use time_{before,after} to avoid timer jiffies wrap (5882)
5947 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5948 - Do not leak request if server was not able to process it (5154)
5949 - If mds_open unable to find parent dir, make that negative lookup(5154)
5950 - don't create new directories with extent-mapping (5909, 5936)
5952 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5953 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5954 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5955 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5956 - support for RHEL4 kernel on i686 (5773)
5957 - provide error messages when incompatible logs are encountered (5898)
5959 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5960 * version 1.4.0.10 (1.4.1 release candidate 1)
5962 - don't keep a lock reference when lock is not granted (4238)
5963 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5964 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5965 - drop import inflight refcount on signal_completed_replay error (5255)
5966 - unlock page after async write error during send (3677)
5967 - handle missing objects in filter_preprw_read properly (5265)
5968 - no transno return for symlink open, don't save no-trasno open (3440)
5969 - don't try to complete elan receive that already failed (4012)
5970 - free RPC server reply state on error (5406)
5971 - clean up thread from ptlrpc_start_thread() on error (5160)
5972 - readahead could read extra page into cache that wasn't ejected (5388)
5973 - prevent races in class_attach/setup/cleanup/detach (5260)
5974 - don't dereference de->d_inode after l_dput of de (5458)
5975 - use "int" for stripe value returned from lock_to_stripe (5544)
5976 - mballoc allocation and error-checking fixes in 2.6 (5504)
5977 - block device patches to fix I/O request sizes in 2.6 (5482)
5978 - look up hostnames for IB nals (5602)
5979 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5980 - don't start multiple acceptors for the same port (5277)
5981 - fix incorrect LASSERT in mds_getattr_name (5635)
5982 - export a proc file for general "ping" checking (5628)
5983 - fix "lfs check" to not block when the MDS is down (5628)
5985 - service request history (4965)
5986 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5987 - create an "evict_client" /proc entry on OSTs, like the MDS has
5988 - fix mount usage message, return errors per mount(8) (5168)
5989 - change grep [] to grep "[]" in tests so they work in more UMLs
5990 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5991 - remove ancient LOV_MAGIC_V0 EA support (5047)
5992 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5993 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5994 - allow readahead allocations to fail when low on memory (5383)
5995 - mmap locking landed again, after considerable improvement (2828)
5996 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5998 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6001 - send OST transaction number in read/write reply to free req (4966)
6002 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6003 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6004 - the watchdog thread now runs as interruptible (5246)
6005 - flock/lockf fixes (but it's still disabled, pending 5135)
6006 - don't use EXT3 constants in llite code (5094)
6007 - memory shortage at startup could cause assertion (5176)
6009 - reorganization of lov code
6010 - single portals codebase
6012 - add extents/mballoc support (5025)
6013 - direct I/O reads in the obdfilter (4048)
6014 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6016 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6019 - send OST transaction number in read/write reply to free req (4966)
6020 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6021 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6022 - the watchdog thread now runs as interruptible (5246)
6023 - handle missing objects in filter_preprw_read properly (5265)
6024 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6025 - drop import inflight refcount on signal_completed_replay error (5255)
6026 - unlock page after async write error during send (3677)
6027 - return original error code on reconstructed replies (3761)
6028 - no transno return for symlink open, don't save no-trasno open (3440)
6030 - add pid to ldlm debugging output (4922)
6031 - bump the watchdog timeouts -- we can't handle 30sec yet
6032 - extra debugging for orphan dentry/inode bug (5259)
6034 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6037 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6038 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6039 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6040 - fix NULL dereference in /proc/sys/portals/routes (4827)
6041 - allow failed mdc_close() operations to be interrupted (4561)
6042 - stop precreate on OST before MDS would time out on it (4778)
6043 - don't send partial-page writes before EOF from client (4410)
6044 - discard client grant for sub-page writes on large-page clients (4520)
6045 - don't free dentries not owned by NFS code, check generation (4806)
6046 - fix lsm leak if mds_create_objects() fails (4801)
6047 - limit debug_daemon file size, always print CERROR messages (4789)
6048 - use transno after validating reply (3892)
6049 - process timed out requests if import state changes (3754)
6050 - update mtime on OST during writes, return in glimpse (4829)
6051 - add mkfsoptions to LDAP (4679)
6052 - use ->max_readahead method instead of zapping global ra (5039)
6053 - don't interrupt __l_wait_event() during strace
6055 - add software watchdogs to catch hung threads quickly (4941)
6056 - make lustrefs init script start after nfs is mounted
6057 - fix CWARN/ERROR duplication (4930)
6058 - return async write errors to application if possible (2248)
6059 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6060 - print NAL number in %x format (4645)
6061 - update barely-supported suse-2.4.21-171 series (4842)
6062 - support for sles 9 %post scripts
6063 - support for building 2.6 kernel-source packages
6064 - support for sles km_* packages
6066 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6069 - ignore -ENOENT errors in osc_destroy (3639)
6070 - notify osc create thread that OSC is being cleaned up (4600)
6071 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6072 - reconstruct ost_handle() like mds_handle() (4657)
6073 - create a new thread to do import eviction to avoid deadlock (3969)
6074 - let lconf resolve symlinked-to devices (4629)
6075 - don't unlink "objects" from directory with default EA (4554)
6076 - hold socknal file ref over connect in case target is down (4394)
6077 - allow more than 32000 subdirectories in a single directory (3244)
6078 - fix blocks count for O_DIRECT writes (3751)
6079 - OST returns ENOSPC from object create when no space left (4539)
6080 - don't send truncate RPC if file size isn't changing (4410)
6081 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6082 - bind to privileged port in socknal and tcpnal (3689)
6084 - rate limit CERROR/CWARN console message to avoid overload (4519)
6085 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6086 - basic mmap support (3918)
6087 - kernel patch series update from b1_4 (4711)
6089 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6092 - avoid crash during MDS cleanup with OST shut down (2775)
6093 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6094 - don't use bad inodes on the MDS (3744)
6095 - dynamic object preallocation to improve recovery speed (4236)
6096 - don't hold spinlock over lock dumping or change debug flags (4401)
6097 - don't zero obd_dev when it is force cleaned (3651)
6098 - print grants to console if they go negative (4431)
6099 - "lctl deactivate" will stop automatic recovery attempts (3406)
6100 - look for existing locks in ldlm_handle_enqueue() (3764)
6101 - don't resolve lock handle twice in recovery avoiding race (4401)
6102 - revalidate should check working dir is a directory (4134)
6104 - don't always mark "slow" obdfilter messages as errors (4418)
6106 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6109 - don't close LustreDB during write_conf until it is done (3860)
6110 - fix typo in lconf for_each_profile (3821)
6111 - allow dumping logs from multiple threads at one time (3820)
6112 - don't allow multiple threads in OSC recovery (3812)
6113 - fix debug_size parameters (3864)
6114 - fix mds_postrecov to initialize import for llog ctxt (3121)
6115 - replace config semaphore with spinlock (3306)
6116 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6117 - don't allow enqueue to complete on a destroyed export (3822)
6118 - down write_lock before checking llog header bitmap (3825)
6119 - recover from lock replay timeout (3764)
6120 - up llog sem before sending rpc (3652)
6121 - reduce ns lock hold times when setting kms (3267)
6122 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6123 - fix NULL deref and obd_dev leak on setup error (3312)
6124 - replace some LBUG about llog ops with error handling (3841)
6125 - don't match INVALID dentries from d_lookup and spin (3784)
6126 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6127 - fix invalid assertion in ptlrpc_set_wait (3880)
6129 - add libwrap support for the TCP acceptor (3996)
6130 - add /proc/sys/portals/routes for non-root route listing (3994)
6131 - allow setting MDS UUID in .xml (2580)
6132 - print the stack of a process that LBUGs (4228)
6134 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6137 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6138 - make sure to unset replay flag from failed open requests (3440)
6139 - if default stripe count is 0, use OST count for inode size (3636)
6140 - update parent mtime/ctime on client for create/unlink (2611)
6141 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6142 - free recovery state on server during a forced cleanup (3571)
6143 - unregister_reply for resent reqs (3063)
6144 - loop back devices mounting and status check on 2.6 (3563)
6145 - fix resource-creation race that can provoke i_size == 0 (3513)
6146 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6147 - more debugging for page-accounting assertion (3746)
6148 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6149 - avoid deadlock after precreation failure (3758)
6150 - fix race and lock order deadlock in orphan handling (3450, 3750)
6151 - add validity checks when grabbing inodes from l_ast_data (3599)
6153 - add /proc/.../recovery_status to obdfilter (3428)
6154 - lightweight CDEBUG infrastructure, debug daemon (3668)
6155 - change default OSC RPC parameters to be better on small clusters
6156 - turn off OST read cache for files smaller than 32MB
6157 - install man pages and include them in rpms (3100)
6158 - add new init script for (un)mounting lustre filesystems (2593)
6159 - run chkconfig in %post for init scripts (3701)
6160 - drop scimac NAL (unmaintained)
6162 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6165 - clean kiobufs before and after use (3485)
6166 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6167 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6168 - change init script to fail more clearly if not run as root (1528)
6169 - allow clients to reconnect during replay (1742)
6170 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6171 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6172 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6173 - fix cond_resched (3554)
6174 - only evict unfinished clients after recovery (3515)
6175 - allow bulk resend, prevent data loss (3570)
6176 - dynamic ptlrpc request buffer allocation (2102)
6177 - don't allow unlinking open directory if it isn't empty (2904)
6178 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6179 - remove extraneous obd dereference causing LASSERT failure (3334)
6180 - don't use get_cycles() when creating temp. files on the mds (3156)
6181 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6182 - handle EEXIST for set-stripe, set proper directory name (3336)
6184 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6185 - fix ksocknal_fmb_callback() error messages (2918)
6187 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6190 - don't copy lvb into (possibly NULL) reply on error (2983)
6191 - don't deref dentry after dput, don't free lvb on error (2922)
6192 - use the kms to determine writeback rpc length (2947)
6193 - increment oti_logcookies when osc is inactive (2948)
6194 - update client's i_blocks count via lvb messages (2543)
6195 - handle intent open/close of special files properly (1557)
6196 - mount MDS with errors=remount-ro, like obdfilter (2009)
6197 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6198 - don't use cancelling-locks' kms values (2947)
6199 - use highest lock extent for kms, not last one (2925)
6200 - don't dereference ERR_PTR() dentry in error handling path (3107)
6201 - fix thread race in portals_debug_dumplog() (3122)
6202 - create lprocfs device entries at setup instead of at attach (1519)
6203 - common AST error handler, don't evict client on completion race (3145)
6204 - zero nameidata in detach_mnt in 2.6 (3118)
6205 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6206 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6207 - initialize RPC timeout timer earlier for 2.6 (3219)
6208 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6209 - print nal/nid for unknown nid (3258)
6210 - additional checks for oscc recovery before doing precreate (3284)
6211 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6212 - don't crash in mdc_close for bad permissions on open (3285)
6213 - zero i_rdev for non-device files (3147)
6214 - clear page->private before handing to FS, better assertion (3119)
6215 - tune the read pipeline (3236)
6216 - fix incorrect decref of invalidated dentry (2350)
6217 - provide read-ahead stats and refine rpc in flight stats (3328)
6218 - don't hold journal transaction open across create RPC (3313)
6219 - update atime on MDS at close time (3265)
6220 - close LDAP connection when recovering to avoid server load (3315)
6221 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6222 - don't leak open file on MDS after open resend (3325)
6223 - serialize filter_precreate and filter_destroy_precreated (3329)
6224 - loop device shouldn't call sync_dev() for nul device (3092)
6225 - clear page cache after eviction (2766)
6226 - resynchronize MDS->OST in background (2824)
6227 - refuse to mount the same filesystem twice on same mountpoint (3394)
6228 - allow llmount to create routes for mounting behind routers (3320)
6229 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6230 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6231 - fix rare problem with rename on htree directories (3417)
6233 - allow default OST striping configuration per directory (1414)
6234 - fix compilation for qswnal for 2.6 kernels (3125)
6235 - increase maximum number of MDS request buffers for large systems
6236 - change liblustreapi to be useful for external progs like lfsck (3098)
6237 - increase local configuration timeout for slow disks (3353)
6238 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6240 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6243 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6244 - don't overwrite extent policy data in reply if lock was blocked (2901)
6245 - drop filter export grants atomically with removal from device (2663)
6246 - del obd_self_export from work_list in class_disconnect_exports (2908)
6247 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6248 - swab reply message in mdc_close, other PPC fixes (2464)
6249 - fix destroying of named logs (2325)
6250 - overwrite old logs when running lconf --write_conf (2264)
6251 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6252 - fix race in target_handle_connect (2898)
6253 - mds_reint_create() should take same inode create lock (2926)
6254 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6255 - don't close files for self_export to avoid uninitialized obd (2936)
6256 - allow MDS with the same name as client node (2939)
6257 - hold dentry reference for closed log files for unlink (2325)
6258 - reserve space for all logs during transactions (2059)
6259 - don't evict page beyond end of stripe extent (2925)
6260 - don't oops on a deleted current working directory (2399)
6261 - handle hard links to targets without a parent properly (2517)
6262 - don't dereference NULL lock when racing during eviction (2867)
6263 - don't grow lock extents when lots of conflicting locks (2919)
6265 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6268 - account for cache space usage on clients to avoid data loss (974)
6269 - lfsck support in lustre kernel code (2349)
6270 - reduce journal credits needed for BRW writes (2370)
6271 - orphan handling to avoid losing space on client/server crashes
6272 - ptlrpcd can be blocked, stopping ALL progress (2477)
6273 - use lock value blocks to assist in proper KMS, faster stat (1021)
6274 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6275 - recovery for initial connections (2355)
6276 - fixes for mds_cleanup_orphans (1934)
6277 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6278 - block all file creations until orphan recovery completes (1901)
6279 - client remove rq_connection from request struct (2423)
6280 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6281 - recovery timer race (2670)
6282 - mdc_close recovey bug (2532)
6283 - ptlrpc cleanup bug (2710)
6284 - mds timeout on local locks (2588)
6285 - namespace lock held during RPCs (2431)
6286 - handle interrupted sync write properly (2503)
6287 - don't try to handle a message that hasn't been replied to (2699)
6288 - client assert failure during cleanup after abort recovery (2701)
6289 - leak mdc device after failed mount (2712)
6290 - ptlrpc_check_set allows timedout requests to complete (2714)
6291 - wait for inflight reqs when ptlrpcd finishes (2710)
6292 - make sure unregistered services are removed from the srv_list
6293 - reset bulk XID's when resending them (caught by 1138 test)
6294 - unregister_bulk after timeout
6295 - fix lconf error (2694)
6296 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6297 - readahead locks pages, leaves pending causing memory pressure (2673)
6298 - increase OST request buffers to 4096 on large machines (2729)
6299 - fix up permission of existing directories in simple_mkdir (2661)
6300 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6301 - don't assign transno to errored transactions (2742)
6302 - don't delete objects on OST if given a bogus objid from MDS (2751)
6303 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6304 - if rq_no_resend, then timeout request after recovery (2432)
6305 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6306 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6307 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6308 - MDS recovery shouldn't delete objects that it has given out (2730)
6309 - if enqueue arrives after completion, don't clobber LVB (2819)
6310 - don't unlock pages twice when trigger_group_io returns error (2814)
6311 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6312 - don't write pages to disk if there was an error (1450)
6313 - don't ping imports that have recovery disabled (2676)
6314 - take buffered bytes into account when balancing socknal conn (2817)
6315 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6316 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6317 - remove little-endian swabbing of llog records (1987)
6318 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6319 - retry reposting request buffers if they fail (1191)
6320 - grow extent at grant time to avoid granting a revoked lock (2809)
6321 - lock revoke doesn't evict page if covered by a second lock (2765)
6322 - disable VM readahead to avoid reading outside lock extents (2805)
6324 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6325 - updated kernel patches for hp-2.4.20 kernel (2681)
6327 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6330 - fix truncated write corruption (2366)
6331 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6333 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6334 - ensure len(uuid) < 37 in lmc (1171)
6335 - fix ia64 OOPS in llog_test (2255)
6336 - zero end of page at obdfilter for partial page writes (2648)
6337 - don't leave stale dentries around after renames (bug 2428)
6338 - fix timeouts when evicting a client with a single lock held (2642)
6339 - set deadline for the initial HELLO message to drain (2634)
6340 - print out dotted-quad IP addresses in the socknal (2302)
6342 - additional debugging for MDS client eviction problem (2443)
6343 - fix mkfsoptions support for osts (2603, 2604)
6345 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6348 - add series for the vanilla 2.6.0 kernel
6349 - add series for the vanilla 2.4.24 kernel
6350 - add series for a cray x86/64 UL kernel drop
6351 - fix xattr patches for the vanilla 2.4.19 series
6353 - generate true UUIDs in lmc (1171)
6354 - have portals stack dumping break in UML (2466)
6355 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6356 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6357 - wake callback waiting threads on client eviction (2460)
6358 - Add --ptldebug and --subsystem to lmc (1719)
6359 - update assertion to allow safe interrupt allocation
6360 - set rq_no_resend for cancel requests (2432)
6361 - recalculate ptlrpcd timeout after resend (2494)
6362 - call vfs_rmdir when removing pending directories (2368)
6363 - fix renaming a file to itself (2429)
6364 - lmc creates a default one-stripe lov (2454)
6365 - expand procfs space to handle large clusters (2326)
6366 - increase UML stack to avoid overflow
6367 - update lconf's list of debug and subsystem masks
6368 - fix lfs find --obd (2510)
6369 - /proc tunable for disabling filter read caching (2591)
6370 - stop rpm packages from altering slapd.conf (2301)
6371 - disable nagle in the socknal under 0conf (2578)
6372 - choose mds inode size based on stripe count (2572)
6373 - fix kernel-source rpm problems (2516)
6375 - add --disable-doc to avoid pdf generation (2421)
6376 - update documentation, tests, type-os, comments
6377 - avoid format warnings on ia64
6378 - remove the TOE NAL
6379 - tiny code cleanups by removing unused fields
6381 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6384 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6385 - remove the most blatant lies from BUILDING file (2371)
6386 - change default debug level to reasonable production setting
6387 - reduce client side cache size to reduce cache flush time
6388 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6389 - make TCP zerocopy and pinger support enabled by default (2476)
6390 - sync writes completed after process exits caused crashes (2319)
6391 - maintain correct mount count on the MDS (2356)
6392 - backout 1557, because 2316 wasn't really fixed
6393 - better file I/O statistics gathering in /proc
6394 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6395 - another kernel patch to fix zero-copy TCP function export
6396 - don't take duplicate lock when processing re-sent getattr (2420)
6397 - lctl uses obd_self_export instead of creating new conn (2353)
6398 - MDS/OST recovery case which requires object creation asserted (2425)
6399 - move lfs from /usr/sbin to /usr/bin in packages
6400 - fix race between mds_client_add and mds_client_free (2417)
6401 - use kmalloc instead of slabs in portals (2430)
6402 - don't create duplicate records when a failover MDS is present (2442)
6403 - remove unnecessary mount age check (2332)
6404 - don't remove directory inodes from locks prematurely (2451)
6405 - don't break if MDS service name is the same as hostname (2103)
6406 - fix races in client write RPC generation when cache full (2482)
6408 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6411 - remove now-unused request->rq_obd (278)
6412 - if an allocation fails, print out how much memory we've used (1933)
6413 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6414 - add the "configurable stack size" patch to most series files (1256)
6415 - ability to write large log records, for 100+ OST configs (2306)
6416 - fix NULL deref when filter_prep fails (2314)
6417 - fix operator precedence error in filter_sync
6418 - dynamic allocation of socknal TX descriptors (2315)
6419 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6420 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6421 - ensure that CWARN messages in Portals always get to the syslog
6422 - __init/__exit are not for prototype decls (ldlm_init/exit)
6423 - x86-64 compile warning fixes
6424 - fix gateway LMC keyword conflict (2318)
6425 - fix MDS lock inversions in getattr/reint paths (1844)
6426 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6427 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6428 - DLM race condition prevented some lock evictions (2328)
6429 - ENOMEM detection and retry on socknal sends (2230)
6430 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6431 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6432 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6433 - better error messages when a client is rejected during recovery (1505)
6434 - avoid cancelling locks which were never granted, after failure (2330)
6435 - fix i_sem/journal inversion in mds_client_add (2333)
6436 - fix truncate/getattr lock cycle deadlock (2334)
6437 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6438 - fix two rare exit paths which could leak an l_lock() ref (2321)
6439 - fencepost error in MDS/OST orphan recovery (2226)
6440 - make log record alignment 8 bytes (1988)
6441 - lstripe now fails when requested offset > ost_count (2237)
6442 - ensure that all kernel series have a complete list.h (1607)
6443 - fix crashes in special-file operations (2316)
6444 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6446 - allow configurable automake binary, for testing new versions
6447 - small update to the lfs documentation
6449 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6451 * fix negative export reference count in fsfilt_sync (2312)
6453 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6454 * release candidate 0.9.1
6456 - orphans are moved into the PENDING directory for possible recovery
6457 - replayed opens now open by fid for orphan/rename safety (1042)
6458 - last close of an orphan inode generates a transno (683)
6459 - chdir() and mount() now pin the directory entry (1020)
6460 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6461 - discard very old requests without processing them (1502)
6462 - remove some common, well-understood CERRORs (1505)
6463 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6464 - clear "grant" flags in OST replies until OST grant code lands (1644)
6465 - fix read performance by not clobbering i_blksize on client (1598)
6466 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6467 - make LRU size a /proc tunable, clears locks when reduced (707)
6468 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6469 - server locks take references on exports now (1558)
6470 - build fixes for 2.4.20-rh trees (1663)
6471 - return an error from lov_create if all OSCs are inactive (1751)
6472 - fix import levels when a reconnect happens without a timeout (1597)
6473 - exit early from mds_open if we get a lookup error (1749)
6474 - partial page read at EOF wouldn't wait for disk before sending (1642)
6475 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6476 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6477 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6478 - write/truncate lock inversion (1639)
6479 - Don't auto-load obdclass, portals modules during cleanup (1495)
6480 - fix timestamps from jumping to "now" (1763)
6481 - extra journal assertions (1648)
6482 - add an extra multiunlink test (1771)
6483 - fix read_record/write_record API (1776)
6484 - fix leak of offset_extent, possible incorrect i_size later (1772)
6485 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6486 - Debugging for the kqswnal_get_idle_tx problems (1820)
6487 - Allow recovery to be attempted multiple times (1536)
6488 - Write out MDS last_rcvd file after it is first created (1600)
6489 - Fix tx_descriptor leak in failed transmit situations (1827)
6490 - ext3 journaling fixes for assertion failure after IO error (1871)
6491 - class_export_put() on freed export after completion AST error (1896)
6492 - Fix revalidate looping in VFS (1322)
6493 - Don't access a freed export during MDS_REINT timeout (1521)
6494 - Add open-unlink recovery support on the MDS (1673,1764)
6495 - Return an error if no MDS data was read from last_rcvd (1946)
6496 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6497 - Don't setup a disk device that doesn't match exported UUID (317)
6498 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6499 - avoid committing NULL handle in force close
6500 - local.sh is now a one-stripe LOV configuration
6501 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6502 - fix ext3 htree duplicate directory entry corruption (1516)
6503 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6504 - update to new LOV EA format (2097)
6505 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6507 2003-06-15 Phil Schwan <phil@clusterfs.com>
6510 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6511 - per-import/export recovery handling (958, 931, 959)
6512 - multiple last-rcvd slots, for serving multiple FSes (949)
6513 - connections are again shared between multiple imp/exports (963, 964)
6514 - "umount -f" would hang if any requests needed to be sent (393, 978)
6515 - avoid pinning large req buffer by copying for queued messages (989)
6516 - add "uuid" to "lctl device" command to help upcalls (991)
6517 - "open" RPCs with transnos would confuse recovery counters (1037)
6518 - do proper endian conversion of all wire messages (288, 340, 891)
6519 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6520 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6521 - fix ptlrpc_connection leak in target_handle_connect (1174)
6522 - fix import refcounting bug in OST and MDS cleanup (1134)
6523 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6524 - fix very unlikely obd_types race condition (501)
6525 - remove osc_open hack for echo_client (1187)
6526 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6527 - a failure in read_inode2 leads to deadlock (1139)
6528 - cancel ack-locks as soon as transaction is committed (1072)
6529 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6530 - make sure to commitrw after any preprw to avoid deadlock (1162)
6531 - failing to execute a file in a lustre FS would lock inode (1203)
6532 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6533 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6534 - obd_uuid2tgt would walk off the end of the list (1255)
6535 - on IA64 the osc would give portals incorrect bulk size (1258)
6536 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6537 - fix lock inversion caused by new llite matching code (1282)
6538 - limit the number of dirty pages on a client to 10MB (1286)
6539 - timed out locks were not being corrected cancelled (1289)
6540 - fix O_DIRECT above 4GB on IA-32 (1292)
6541 * major user-visible changes
6542 - fail out/fail over policy now controlled by the upcall (993)
6544 - add OBD_PING to check server availability and failure (954)
6545 - lustre messages are now sent in sending host order (288, 340, 891)
6546 - add eadatalen to MDS getattr reply (340)
6547 - OST read replies may contain second buffer, with per-page status (593)
6549 2003-03-11 Phil Schwan <phil@clusterfs.com>
6552 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6553 - failed open()s could cause deadlock; fixed (867, 869)
6554 - stop cancelling OST locks when files are closed (481)
6555 - overlapping XID spaces caused network corruption (851, 853)
6556 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6557 - setattr_raw added, to do single-RPC, server-side setattrs
6558 - lmc/lconf syntax change for OST UUIDs
6559 - fix crashy race condition between ptlrpc_free_req and osc_close
6560 - don't use request in mdc_enqueue if we hit a timeout (889)
6561 - don't set the inode i_size for regular files from the MDS (896)
6562 - handle out of order completion AST (842)
6563 - don't LBUG if a lock request times out after receiving AST (913)
6564 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6565 - if a bad lock AST arrives, send an error instead of dropping entirely
6566 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6567 - fix leak in bulk IO when only partially completed (899, 900, 926)
6568 - fix O_DIRECT for ia64 (55)
6569 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6570 - C-z after timeout could hang a process forever; fixed (977)
6572 - client-side I/O cache (678, 924, 929, 941, 970)
6574 - READPAGE and SETATTRs which don't take server-side locks get
6577 2003-02-11 Phil Schwan <phil@clusterfs.com>
6580 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6581 - Fix fsfilt_extN_readpage() to read a full page of directory
6582 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6583 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6584 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6585 - Most intent operations were removed, in favour of a new RPC mode
6586 that does a single RPC to the server and bypasses most of the VFS
6587 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6588 - Aggressively cancel local locks on DLM servers
6589 - mds_reint_unlink sends EA to the client if it's the last nlink.
6590 client uses that EA to unlink OST objects.
6591 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6592 - recursive symlinks were fixed (439)
6593 - fixed NULL deref in DEBUG_REQ
6594 - filter_update_lastobjid no longer calls sync, which annoyed extN
6595 - fixed multi-client small-writes to a single file problem (445)
6596 - fixed mtime updates during file writes (607)
6597 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6598 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6599 - fixed Linux OST size reporting problem (444, 656)
6600 - OST now updates object mtime with writes or setattr (607, 619)
6601 - client verifies file size before zeroing page past EOF (445)
6602 - OST now writes last allocated objid to disk with allocation (108)
6603 - LOV on echo now works (409)
6605 - mds_reint_unlink sends a new buffer, with the EA included. this
6606 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6607 set if a regular file was being unlinked, and it was the last link
6608 - use PtlGet from the target for bulk writes (315)
6609 - OST now updates object mtime with writes or setattr (607, 619)
6610 - LDLM now has a grant-time callback to revalidate locked items, if
6612 - Many MDS operations were reorganized to combat race conditions
6614 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6615 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6617 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6620 - Fully reactivate OST imports after reconnection (512, others)
6621 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6622 - More graceful error handling for truncating on dead OST (515)
6623 - Don't error out unless we're actually accessing dead stripes (474)
6624 - Fix garbage sizes when stripes are missing (410)
6625 - LRU counters were broken, causing constant lock purge (433, 432)
6626 - garbage on read from stripes with failed OSTs (441)
6627 - mark OSCs as active before reconnecting during recovery (438)
6628 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6629 - lfind did not preserve OST order in output (443)
6630 - symlinks cause hung clients, incorrect data (439)
6631 - stop dereferencing request after dropping refcount (457)
6632 - don't LASSERT(spin_is_locked) on non-SMP (455)
6633 - fixes for many rename() bugs
6634 - fstat didn't correctly synchronize attributes (399)
6635 - server must handle lock cancellation during blocking AST prep (487)
6636 - bulk descriptors were free()d too soon (511)
6637 - fix paths in lconf, which would load incorrect modules (451, 507)
6638 - fix confusing lconf 'host not found' error message (386)
6639 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6640 - fix race condition in mdc_blocking_ast() for inode access (526)
6641 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6642 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6643 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6644 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6645 - fix lstripe in several ways (kernel side) (527)
6646 - fix request leak in ldlm_cli_enqueue (262)
6647 - incorrect OSC was marked inactive after OST failure
6648 - call mds_fs_cleanup before unmounting filesystem (524)
6649 - fix races between taking ns_lock and ldlm_lock_change_resource
6650 - fix races updating LOV export open file list
6651 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6652 - fix recovery NULL deref in ldlm_cli_cancel_unused
6653 - fix some DLM races by using new hash table for lock handles (419)
6654 - permit the client to specify desired inodes, at replay
6655 - duplicate requests when we queue them for replay reintegration
6656 - fix last_rcvd offset calculation
6657 - sync after each recovered transaction, so we always make progress
6658 - never, not always, ERESTART requests without transnos
6659 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6661 - skip replay if the MDS says that the client is already connected
6662 - don't check for a recovery-enabled export to match lctl's UUID
6663 - don't INC_USE_COUNT for phantom exports
6664 - don't crash when cleaning up phantom exports (567)
6665 - don't double-finish or set replay data for errored mdc_open requests
6666 - abort requests when they time out, so we don't get old replies
6667 - send/receive replies for AST messages again
6668 - if the client says that it doesn't have the lock, cancel it on the
6670 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6671 mark it as destroyed, it will all work out when decref is called
6672 - fix module use counts (22, 581)
6674 - ASTs now expect a reply (server cancels lock on error reply)
6676 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6679 - fix many simultaneous client startup (392)
6680 - fix dentry->d_it clobbering
6681 - credentials weren't being shipped for readdir/getattr operations
6682 - remove invalid assertions triggered during some concurrent MD
6684 - proper Lustre versions added (336, 389)
6685 - fix memory leak for create error case (398)
6686 - fix LOV locking bug that would get cli/srv out of sync
6687 - fix echo client over LOV (409)
6688 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6689 - fix double-O_EXCL intent crash (424)
6690 - avoid sending multiple lock CANCELs (352)
6692 - MDS can do multi-client recovery (modulo bugs in new code)
6694 - many updates, edits, cleanups
6696 2002-11-18 Phil Schwan <phil@clusterfs.com>
6699 - fix null d_it dereference (346)
6700 - fix full OST/dbench hang (333)
6701 - fix permission problem with file removal (286)
6702 - fix removal of OSCs from LOV when they fail
6703 - fix NULL deref during bulk timeout (214)
6704 - fix problems related to multiple filesystems on one MDS (241)
6705 - fixed serious subtle metadata locking bugs
6706 - free locks on clients when inodes are removed due to memory
6708 - fix inode pointer in lock data (285)
6709 - partial support for multiple MDS on a single host (241)
6710 - data locks weren't cancelled at clear_inode time (290, 311)
6711 - intent locks could lead to unbounded lock growth (205)
6712 - added a maximum lock count, an LRU list, and a flusher
6713 - fix multiple rename (365)
6714 - properly abstracted the echo client
6715 - OSC locked 1 byte too many; fixed
6716 - rewrote brw callback code:
6717 - fixed recovery bugs related to LOVs (306)
6718 - fixed too-many-pages-in-one-write crash (191)
6719 - fixed (again) crash in sync_io_timeout (214)
6720 - probably fixed callback-related race (385)
6722 - Add capability to MDS protocol
6723 - LDLM cancellations and callbacks on different portals
6725 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6728 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6729 - defer open object destruction to close time (601981, 138)
6730 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6731 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6732 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6733 - don't call obd_setattr in ll_file_release for destroyed objects
6734 * protocol change to lustre_msg: move |version| and add |flags|
6735 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6736 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6737 * added replay of create, unlink, link and rename operations during
6738 MDS failover; recovery should be much more robust now
6739 * remove failed OSCs from LOVs (only lov_create uses this so far)
6740 * the lustre-HOWTO was brought (more) up to date (582544)
6742 2002-10-23 Phil Schwan <phil@clusterfs.com>
6745 - in-use dentries weren't being reused properly (617851)
6746 - prevent multiple LDLM setup (599178)
6747 - fix LOV size calculations for truncate (617853)
6748 - fix client handling of MDS intent errors (POSIX)
6749 - fix permission bug in lovstripe.c test (624321)
6750 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6751 - truncate past end of file could corrupt data
6752 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6753 - a race in recovery could return ETIMEDOUT to apps (623947)
6754 - building outside the source directory was fixed
6755 * the lustre-HOWTO was brought (more) up to date (582544)
6756 * major progress was made on recovery functionality
6758 2002-10-10 Phil Schwan <phil@clusterfs.com>
6761 - recovery deadlock fix
6762 - rm -rf causes LBUG fix (617817)
6763 - file open by multiple tasks fix (618962)
6764 - directory permissions bugs (602707 and 620007)
6765 - journal_stop fixed with locking (611313)
6766 - O_APPEND failures resolved (618273, perhaps 614459)
6767 - lconf PATH fix (619770)
6768 - IA64 build fix (621450)
6769 - RPC buffer sizes scale with amount of memory
6771 2002-10-01 Phil Schwan <phil@clusterfs.com>
6774 - locks would be cancelled without throwing away data pages,
6775 resulting in inconsistent data (605627)
6776 - inode attributes were not always being refreshed (605627, 612449)
6777 - lconf now continues to cleanup after lctl reports an error
6778 - MDS now enforces user permissions (602707)
6779 - lprocfs cleanup fixed, but not yet enabled (614157)
6780 - fixed infinite server hang, should a client not respond to an AST
6781 - avoid going into recovery if user calls readlink() with a buffer
6782 that's too small (613941)
6783 - AST RPCs no longer require replies (614867) -- this may be changed
6784 - don't crash server if client sends an IOV that's too big (611336)
6785 - fixed lock conversion deadlock (611892)
6786 - fixed the following of symlinks (614622)
6787 * recovery: the server can remove locks from a client that dies, other
6788 clients can make progress
6789 * more extN patch fixes
6790 * compile-time configurable ptlrpc buffer allocations
6792 - collaborative read cache document
6793 - Lustre Lite Performance CDR document-in-progress
6795 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6798 - fix typo in patch-2.4.18
6800 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6803 - clear ptlrpc request each time in handle_incoming_request()
6804 - unlink of files now destroys the object on the OST
6806 2002-09-19 Peter Braam <braam@clusterfs.com>
6808 * add hard link support
6809 * change obdfile creation method
6810 * kernel patch changed
6812 2002-09-19 Peter Braam <braam@clusterfs.com>
6815 - stack overflow bug in extN fixed
6817 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6819 * documentation updates
6820 - add man pages for config tools
6821 - update tests/README to describe testing with new config tools
6822 - finish metadata API descriptions
6823 * bug fixes and cleanups
6824 - statfs workaround for 16TB limit
6825 - LOV stripe allocation improved, can stripe on subset of OSTs
6826 - LOV file size/IO offset was wrong for files > 4GB in size
6827 - object EA data was being dropped, caused files to be unreadable
6828 - memory overflow with non-LOV OST caused memory corruption
6829 - fixed regression tests to work with new config tools, obdfilter
6830 - fixed bug when directory size became larger than 1 block
6831 - fixed bug (for single client case) when PWD was deleted
6832 - invalidate local directory pages when doing intent-based ops
6833 - avoid LDLM oops when lock callback contained bad data
6835 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6837 * documentation updates
6838 * bug fixes and cleanups
6839 - configuration tools
6842 - 64-bit compile warnings
6843 - 64-bit internal statfs data
6845 * test_brw on persistent OST devices
6847 * lprocfs (disabled)
6849 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6851 * documentation updates
6852 * bug fixes and cleanups
6853 * configuration tools
6855 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6858 * many small fixes to 0_5_4
6859 * io/network handling
6860 * thinkos in MDS operations
6862 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6865 * crucial basic fixes to 0.5.3
6866 * IOR, Iozone work over Elan
6869 2002-08-07 Phil Schwan <phil@clusterfs.com>
6870 * version 0_5_3, our first alpha
6871 * we use the new Portals iovs
6872 * documentation updates
6873 * bug fixes and cleanups
6874 * small changes in the DLM wire protocol
6876 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6877 * version 0_5_1 with some initial stability,
6878 * locking on MD and file I/O.
6879 * documentation updates
6880 * several bug fixes since 0.5.0
6881 * small changes in wire protocol
6883 2002-07-18 Phil Schwan <phil@clusterfs.com>
6885 * delivered as Lustre Light Alpha
6886 * fixed a crash after handling invalid MDS requests
6887 * fixed directory pages for architectures with non-4k pages sizes
6889 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6890 * release version v0_4_4
6891 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6893 * Fixes a number of interruption problems with OST operations.
6894 * Update documentation for portals header changes
6895 * Move all wire protocol structs/defines to lustre_idl.h
6896 * Fixes symlink length bug.
6897 * Add tcpdump to repository.
6899 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6900 * release version v0_4_3
6901 * Fixes statfs for inodes on extN.
6902 * Fixes bug in runtests which would delete /etc/hosts.
6903 * Use 64-bit object IDs wherever possible (not into VFS though)
6904 Remove ost_get_info, which is unused by lustre, and out of date.
6906 2002-07-03 Peter Braam <braam@clusterfs.com>
6907 * release version v0_4_2 Fixes a lookup error (type not passed)
6908 * move forward to head of Portals
6909 * move forward to latest Lustre kernel
6911 2002-06-25 Peter Braam <braam@clusterfs.com>
6912 * release version v0_4_1. Hopefully stable on single node use.