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 3.10.0-957.27.2.el7 (RHEL7.6)
7 3.10.0-1062.18.1.el7 (RHEL7.7)
8 3.10.0-1127.19.1.el7 (RHEL7.8)
9 3.10.0-1160.2.1.el7 (RHEL7.9)
10 * Server known to build on the following kernels (others may also work):
11 2.6.32-696.18.7.el6 (RHEL6.9)
12 3.10.0-862.14.4.el7 (RHEL7.5)
13 4.4.120-92.70 (SLES12 SP2)
14 4.4.180-94.97 (SLES12 SP3)
15 4.4.0-142 (Ubuntu 16.04)
16 4.15.0-45 (Ubuntu 18.04)
17 vanilla linux 4.6.7 (ZFS only)
18 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
19 * Client primary unpatched kernels built and tested during release cycle:
20 3.10.0-957.27.2.el7 (RHEL7.6)
21 3.10.0-1062.18.1.el7 (RHEL7.7)
22 3.10.0-1127.19.1.el7 (RHEL7.8)
23 3.10.0-1160.2.1.el7 (RHEL7.9)
24 4.18.0-80.11.2.el8 (RHEL8.0)
25 4.4.180-94.97 (SLES12 SP3)
26 4.12.14-95.54 (SLES12 SP4)
27 4.12.14-122.20 (SLES12 SP5)
28 4.15.0-45 (Ubuntu 18.04)
29 * Client known to build on unpatched kernels (others may also work):
30 2.6.32-696.18.7.el6 (RHEL6.9)
31 2.6.32-754.6.3.el6 (RHEL6.10)
32 3.10.0-862.14.4.el7 (RHEL7.5)
33 4.14.0-49.13.1.el7a (RHEL7.5)
34 4.14.0-115.2.2.el7a (RHEL7.6)
35 4.18.0-147.3.1.el8 (RHEL8.1)
36 4.18.0-193.28.1.el8 (RHEL8.2)
37 4.18.0-240.1.1.el8 (RHEL8.3)
38 4.4.120-92.70 (SLES12 SP2)
39 4.4.0-142 (Ubuntu 16.04)
40 vanilla linux 4.15.0 (ZFS only)
41 * Recommended e2fsprogs version: 1.45.6.wc1 or newer
42 * Recommended ZFS / SPL version: 0.7.13
43 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
44 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
45 stack overflow. For more information, please refer to bugzilla 17630.
46 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
47 the re-exporting nodes to be mounted with "32bitapi" mount option
48 * Note for Ubuntu 14.04.5:
49 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
50 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
54 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
55 for currently supported client and server kernel versions.
56 * Server primary kernels built and tested during release cycle:
57 3.10.0-957.27.2.el7 (RHEL7.6)
58 3.10.0-1062.18.1.el7 (RHEL7.7)
59 3.10.0-1127.8.2.el7 (RHEL7.8)
60 * Server known to build on the following kernels (others may also work):
61 2.6.32-696.18.7.el6 (RHEL6.9)
62 3.10.0-862.14.4.el7 (RHEL7.5)
63 4.4.120-92.70 (SLES12 SP2)
64 4.4.180-94.97 (SLES12 SP3)
65 4.4.0-142 (Ubuntu 16.04)
66 4.15.0-45 (Ubuntu 18.04)
67 vanilla linux 4.6.7 (ZFS only)
68 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
69 * Client primary unpatched kernels built and tested during release cycle:
70 3.10.0-957.27.2.el7 (RHEL7.6)
71 3.10.0-1062.18.1.el7 (RHEL7.7)
72 3.10.0-1127.8.2.el7 (RHEL7.8)
73 4.18.0-80.11.2.el8 (RHEL8.0)
74 4.4.180-94.97 (SLES12 SP3)
75 4.12.14-95.45 (SLES12 SP4)
76 4.12.14-122.20 (SLES12 SP5)
77 4.15.0-45 (Ubuntu 18.04)
78 * Client known to build on unpatched kernels (others may also work):
79 2.6.32-696.18.7.el6 (RHEL6.9)
80 2.6.32-754.6.3.el6 (RHEL6.10)
81 3.10.0-862.14.4.el7 (RHEL7.5)
82 4.14.0-49.13.1.el7a (RHEL7.5)
83 4.14.0-115.2.2.el7a (RHEL7.6)
84 4.18.0-147.8.1.el8 (RHEL8.1)
85 4.18.0-193.1.2.el8 (RHEL8.2)
86 4.4.120-92.70 (SLES12 SP2)
87 4.4.0-142 (Ubuntu 16.04)
88 vanilla linux 4.15.0 (ZFS only)
89 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
90 * Recommended ZFS / SPL version: 0.7.13
91 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
92 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
93 stack overflow. For more information, please refer to bugzilla 17630.
94 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
95 the re-exporting nodes to be mounted with "32bitapi" mount option
96 * Note for Ubuntu 14.04.5:
97 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
98 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
102 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
103 for currently supported client and server kernel versions.
104 * Server primary kernels built and tested during release cycle:
105 3.10.0-957.27.2.el7 (RHEL7.6)
106 3.10.0-1062.9.1.el7 (RHEL7.7)
107 * Server known to build on the following kernels (others may also work):
108 2.6.32-696.18.7.el6 (RHEL6.9)
109 3.10.0-862.14.4.el7 (RHEL7.5)
110 4.4.120-92.70 (SLES12 SP2)
111 4.4.180-94.97 (SLES12 SP3)
112 4.4.0-142 (Ubuntu 16.04)
113 4.15.0-45 (Ubuntu 18.04)
114 vanilla linux 4.6.7 (ZFS only)
115 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
116 * Client primary unpatched kernels built and tested during release cycle:
117 3.10.0-957.27.2.el7 (RHEL7.6)
118 3.10.0-1062.9.1.el7 (RHEL7.7)
119 4.18.0-80.11.2.el8 (RHEL8.0)
120 4.4.180-94.97 (SLES12 SP3)
121 4.12.14-95.29 (SLES12 SP4)
122 4.15.0-45 (Ubuntu 18.04)
123 * Client known to build on unpatched kernels (others may also work):
124 2.6.32-696.18.7.el6 (RHEL6.9)
125 2.6.32-754.6.3.el6 (RHEL6.10)
126 3.10.0-862.14.4.el7 (RHEL7.5)
127 4.14.0-49.13.1.el7a (RHEL7.5)
128 4.14.0-115.2.2.el7a (RHEL7.6)
129 4.18.0-147.3.1.el8 (RHEL8.1)
130 4.4.120-92.70 (SLES12 SP2)
131 4.4.0-142 (Ubuntu 16.04)
132 vanilla linux 4.15.0 (ZFS only)
133 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
134 * Recommended ZFS / SPL version: 0.7.13
135 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
136 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
137 stack overflow. For more information, please refer to bugzilla 17630.
138 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
139 the re-exporting nodes to be mounted with "32bitapi" mount option
140 * Note for Ubuntu 14.04.5:
141 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
142 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
146 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
147 for currently supported client and server kernel versions.
148 * Server primary kernels built and tested during release cycle:
149 3.10.0-957.27.2.el7 (RHEL7.6)
150 3.10.0-1062.1.1.el7 (RHEL7.7)
151 * Server known to build on the following kernels (others may also work):
152 2.6.32-696.18.7.el6 (RHEL6.9)
153 3.10.0-862.14.4.el7 (RHEL7.5)
154 4.4.120-92.70 (SLES12 SP2)
155 4.4.180-94.97 (SLES12 SP3)
156 4.4.0-142 (Ubuntu 16.04)
157 4.15.0-45 (Ubuntu 18.04)
158 vanilla linux 4.6.7 (ZFS only)
159 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
160 * Client primary unpatched kernels built and tested during release cycle:
161 3.10.0-957.27.2.el7 (RHEL7.6)
162 3.10.0-1062.1.1.el7 (RHEL7.7)
163 4.18.0-80.7.1.el8 (RHEL8.0)
164 4.4.180-94.97 (SLES12 SP3)
165 4.12.14-95.29 (SLES12 SP4)
166 4.15.0-45 (Ubuntu 18.04)
167 * Client known to build on unpatched kernels (others may also work):
168 2.6.32-696.18.7.el6 (RHEL6.9)
169 2.6.32-754.6.3.el6 (RHEL6.10)
170 3.10.0-862.14.4.el7 (RHEL7.5)
171 4.14.0-49.13.1.el7a (RHEL7.5)
172 4.14.0-115.2.2.el7a (RHEL7.6)
173 4.4.120-92.70 (SLES12 SP2)
174 4.4.0-142 (Ubuntu 16.04)
175 vanilla linux 4.15.0 (ZFS only)
176 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
177 * Recommended ZFS / SPL version: 0.7.13
178 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
179 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
180 stack overflow. For more information, please refer to bugzilla 17630.
181 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
182 the re-exporting nodes to be mounted with "32bitapi" mount option
183 * Note for Ubuntu 14.04.5:
184 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
185 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
189 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
190 for currently supported client and server kernel versions.
191 * Server known to build on the following kernels (others may also work):
192 2.6.32-696.18.7.el6 (RHEL6.9)
193 3.10.0-862.14.4.el7 (RHEL7.5)
194 3.10.0-957.10.1.el7 (RHEL7.6)
195 4.4.120-92.70 (SLES12 SP2)
196 4.4.162-94.69 (SLES12 SP3)
197 4.4.0-142 (Ubuntu 16.04)
198 4.15.0-45 (Ubuntu 18.04)
199 vanilla linux 4.6.7 (ZFS only)
200 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
201 * Client known to build on unpatched kernels (others may also work):
202 2.6.32-696.18.7.el6 (RHEL6.9)
203 2.6.32-754.6.3.el6 (RHEL6.10)
204 3.10.0-862.14.4.el7 (RHEL7.5)
205 3.10.0-957.10.1.el7 (RHEL7.6)
206 4.14.0-49.13.1.el7a (RHEL7.5)
207 4.14.0-115.2.2.el7a (RHEL7.6)
208 4.4.120-92.70 (SLES12 SP2)
209 4.4.162-94.69 (SLES12 SP3)
210 4.12.14-95.13 (SLES12 SP4)
211 4.4.0-142 (Ubuntu 16.04)
212 4.15.0-45 (Ubuntu 18.04)
213 vanilla linux 4.15.0 (ZFS only)
214 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
215 * Recommended ZFS / SPL version: 0.7.13
216 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
217 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
218 stack overflow. For more information, please refer to bugzilla 17630.
219 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
220 the re-exporting nodes to be mounted with "32bitapi" mount option
221 * Note for Ubuntu 14.04.5:
222 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
223 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
228 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
229 for currently supported client and server kernel versions.
230 * Server known to build on the following kernels (others may also work):
231 2.6.32-696.18.7.el6 (RHEL6.9)
232 3.10.0-862.14.4.el7 (RHEL7.5)
233 3.10.0-957.10.1.el7 (RHEL7.6)
234 4.4.120-92.70 (SLES12 SP2)
235 4.4.162-94.69 (SLES12 SP3)
236 4.4.0-142 (Ubuntu 16.04)
237 4.15.0-45 (Ubuntu 18.04)
238 vanilla linux 4.6.7 (ZFS only)
239 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
240 * Client known to build on unpatched kernels (others may also work):
241 2.6.32-696.18.7.el6 (RHEL6.9)
242 2.6.32-754.6.3.el6 (RHEL6.10)
243 3.10.0-862.14.4.el7 (RHEL7.5)
244 3.10.0-957.10.1.el7 (RHEL7.6)
245 4.14.0-49.13.1.el7a (RHEL7.5)
246 4.14.0-115.2.2.el7a (RHEL7.6)
247 4.4.120-92.70 (SLES12 SP2)
248 4.4.162-94.69 (SLES12 SP3)
249 4.12.14-95.13 (SLES12 SP4)
250 4.4.0-142 (Ubuntu 16.04)
251 4.15.0-45 (Ubuntu 18.04)
252 vanilla linux 4.15.0 (ZFS only)
253 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
254 * Recommended ZFS / SPL version: 0.7.13
255 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
256 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
257 stack overflow. For more information, please refer to bugzilla 17630.
258 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
259 the re-exporting nodes to be mounted with "32bitapi" mount option
260 * Note for Ubuntu 14.04.5:
261 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
262 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
266 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
267 for currently supported client and server kernel versions.
268 * Server known to build on the following kernels (others may also work):
269 2.6.32-696.18.7.el6 (RHEL6.9)
270 3.10.0-862.14.4.el7 (RHEL7.5)
271 3.10.0-957.el7 (RHEL7.6)
272 4.4.120-92.70 (SLES12 SP2)
273 4.4.162-94.69 (SLES12 SP3)
274 4.4.0-131 (Ubuntu 16.04)
275 4.15.0-32 (Ubuntu 18.04)
276 vanilla linux 4.6.7 (ZFS only)
277 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
278 * Client known to build on unpatched kernels (others may also work):
279 2.6.32-696.18.7.el6 (RHEL6.9)
280 2.6.32-754.6.3.el6 (RHEL6.10)
281 3.10.0-862.14.4.el7 (RHEL7.5)
282 3.10.0-957.el7 (RHEL7.6)
283 4.14.0-49.13.1.el7a (RHEL7.5)
284 4.14.0-115.2.2.el7a (RHEL7.6)
285 4.4.120-92.70 (SLES12 SP2)
286 4.4.162-94.69 (SLES12 SP3)
287 4.4.0-131 (Ubuntu 16.04)
288 4.15.0-32 (Ubuntu 18.04)
289 vanilla linux 4.15.0 (ZFS only)
290 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
291 * Recommended ZFS / SPL version: 0.7.9
292 * Tested with ZFS / SPL version: 0.7.9
293 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
294 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
295 stack overflow. For more information, please refer to bugzilla 17630.
296 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
297 the re-exporting nodes to be mounted with "32bitapi" mount option
298 * Note for Ubuntu 14.04.5:
299 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
300 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
302 --------------------------------------------------------------------------------
305 2018-04-03 Intel Corporation
307 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
308 for currently supported client and server kernel versions.
309 * Server known to build on patched kernels:
310 2.6.32-431.29.2.el6 (RHEL6.5)
311 2.6.32-504.30.3.el6 (RHEL6.6)
312 2.6.32-573.26.1.el6 (RHEL6.7)
313 2.6.32-642.15.1.el6 (RHEL6.8)
314 2.6.32-696.18.7.el6 (RHEL6.9)
315 3.10.0-693.21.1.el7 (RHEL7.4)
316 3.0.101-0.47.71 (SLES11 SP3)
317 3.0.101-107 (SLES11 SP4)
318 3.12.74-60.64.40 (SLES12 SP1)
319 4.4.114-92.64 (SLES12 SP2)
320 4.4.114-94.11 (SLES12 SP3)
321 3.13.0-101 (Ubuntu 14.04, ZFS only)
322 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
323 4.4.0-116 (Ubuntu 16.04)
324 vanilla linux 4.6.7 (ZFS only)
325 * Client known to build on unpatched kernels:
326 2.6.32-431.29.2.el6 (RHEL6.5)
327 2.6.32-504.30.3.el6 (RHEL6.6)
328 2.6.32-573.26.1.el6 (RHEL6.7)
329 2.6.32-642.15.1.el6 (RHEL6.8)
330 2.6.32-696.18.7.el6 (RHEL6.9)
331 3.10.0-693.21.1.el7 (RHEL7.4)
332 3.0.101-0.47.71 (SLES11 SP3)
333 3.0.101-107 (SLES11 SP4)
334 3.12.74-60.64.40 (SLES12 SP1)
335 4.4.114-92.64 (SLES12 SP2)
336 4.4.114-94.11 (SLES12 SP3)
337 3.13.0-101 (Ubuntu 14.04)
338 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
339 4.4.0-116 (Ubuntu 16.04)
341 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
342 * Recommended ZFS / SPL version: 0.7.6
343 * Tested with ZFS / SPL version: 0.7.6
344 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
345 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
346 stack overflow. For more information, please refer to bugzilla 17630.
347 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
348 the re-exporting nodes to be mounted with "32bitapi" mount option
349 * Note for Ubuntu 14.04.5:
350 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
351 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
353 --------------------------------------------------------------------------------
355 2017-07-05 Intel Corporation
357 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
358 for currently supported client and server kernel versions.
359 * Server known to build on patched kernels:
360 2.6.32-431.29.2.el6 (RHEL6.5)
361 2.6.32-504.30.3.el6 (RHEL6.6)
362 2.6.32-573.26.1.el6 (RHEL6.7)
363 2.6.32-642.15.1.el6 (RHEL6.8)
364 2.6.32-696.3.1.el6 (RHEL6.9)
365 3.10.0-514.21.1.el7 (RHEL7.3)
366 3.0.101-0.47.71 (SLES11 SP3)
367 3.0.101-100 (SLES11 SP4)
368 3.12.74-60.64.40 (SLES12 SP1)
369 4.4.59-92.17 (SLES12 SP2)
370 vanilla linux 4.6.7 (ZFS only)
371 * Client known to build on unpatched kernels:
372 2.6.32-431.29.2.el6 (RHEL6.5)
373 2.6.32-504.30.3.el6 (RHEL6.6)
374 2.6.32-573.26.1.el6 (RHEL6.7)
375 2.6.32-642.15.1.el6 (RHEL6.8)
376 2.6.32-696.3.1.el6 (RHEL6.9)
377 3.10.0-514.21.1.el7 (RHEL7.3)
378 3.0.101-0.47.71 (SLES11 SP3)
379 3.0.101-100 (SLES11 SP4)
380 3.12.74-60.64.40 (SLES12 SP1)
381 4.4.59-92.17 (SLES12 SP2)
383 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
384 * Recommended ZFS / SPL version: 0.6.5.9
385 * Tested with ZFS / SPL version: 0.6.5.9
386 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
387 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
388 stack overflow. For more information, please refer to bugzilla 17630.
389 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
390 the re-exporting nodes to be mounted with "32bitapi" mount option
392 --------------------------------------------------------------------------------
394 12-07-2016 Intel Corporation
396 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
397 for currently supported client and server kernel versions.
398 * Server known to build on patched kernels:
399 2.6.32-431.29.2.el6 (RHEL6.5)
400 2.6.32-504.30.3.el6 (RHEL6.6)
401 2.6.32-573.26.1.el6 (RHEL6.7)
402 2.6.32-642.11.1.el6 (RHEL6.8)
403 3.10.0-514.el7 (RHEL7.3)
404 3.0.101-0.47.71 (SLES11 SP3)
405 3.0.101-80 (SLES11 SP4)
406 3.12.62-60.64.8 (SLES12 SP1)
407 vanilla linux 4.5.7 (ZFS only)
408 * Client known to build on unpatched kernels:
409 2.6.32-431.29.2.el6 (RHEL6.5)
410 2.6.32-504.30.3.el6 (RHEL6.6)
411 2.6.32-573.26.1.el6 (RHEL6.7)
412 2.6.32-642.11.1.el6 (RHEL6.8)
413 3.10.0-514.el7 (RHEL7.3)
414 3.0.101-0.47.71 (SLES11 SP3)
415 3.0.101-80 (SLES11 SP4)
416 3.12.62-60.64.8 (SLES12 SP1)
418 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
419 * Recommended ZFS / SPL version: 0.6.5.8
420 * Tested with ZFS / SPL version: 0.6.5.8
421 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
422 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
423 stack overflow. For more information, please refer to bugzilla 17630.
424 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
425 the re-exporting nodes to be mounted with "32bitapi" mount option
427 --------------------------------------------------------------------------------
429 02-29-2016 Intel Corporation
431 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
432 for currently supported client and server kernel versions.
433 * Server known to build on patched kernels:
434 2.6.32-431.29.2.el6 (RHEL6.5)
435 2.6.32-504.30.3.el6 (RHEL6.6)
436 2.6.32-573.12.1.el6 (RHEL6.7)
437 3.10.0-327.3.1.el7 (RHEL7.2)
438 3.0.101-0.47.71 (SLES11 SP3)
439 3.0.101-68 (SLES11 SP4)
440 vanilla linux 4.2.1 (ZFS only)
441 * Client known to build on unpatched kernels:
442 2.6.32-431.29.2.el6 (RHEL6.5)
443 2.6.32-504.30.3.el6 (RHEL6.6)
444 2.6.32-573.12.1.el6 (RHEL6.7)
445 3.10.0-327.3.1.el7 (RHEL7.2)
446 3.0.101-0.47.71 (SLES11 SP3)
447 3.0.101-68 (SLES11 SP4)
450 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
451 * Recommended ZFS / SPL version: 0.6.4.2
452 * Tested with ZFS / SPL version: 0.6.4.2
453 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
454 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
455 stack overflow. For more information, please refer to bugzilla 17630.
456 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
457 the re-exporting nodes to be mounted with "32bitapi" mount option
459 --------------------------------------------------------------------------------
461 03-10-2015 Intel Corporation
463 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
464 for currently supported client and server kernel versions.
465 * Server known to build on patched kernels:
466 2.6.32-431.29.2.el6 (RHEL6.5)
467 2.6.32-504.8.1.el6 (RHEL6.6)
468 3.0.101-0.46 (SLES11 SP3)
469 * Client known to build on unpatched kernels:
470 2.6.32-431.29.2.el6 (RHEL6.5)
471 2.6.32-504.8.1.el6 (RHEL6.6)
472 3.10.0-123.20.1.el7 (RHEL7)
473 3.0.101-0.46 (SLES11 SP3)
474 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
475 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
476 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
477 stack overflow. For more information, please refer to bugzilla 17630.
478 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
479 the re-exporting nodes to be mounted with "32bitapi" mount option
481 Severity : enhancement
483 Description: control OST-index in IDIF via ROCOMPAT flag.
484 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
485 last_rcvd file. For new formatted OST device, it will be auto set;
486 for the case of upgrading from old OST device, you can enable it
487 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
488 enabled, for new created OST-object, its IDIF-in-LMA will contain
489 the OST-index; for the existing OST-object, the OSD will convert
490 old format IDIF as new format IDIF with OST-index stored in the
491 LMA EA when accessing such OST-object or via OI scrub. Once such
492 flag is enabled, it cannot be reverted back, so the system cannot
493 be downgraded to the orignal incompatible version.
495 --------------------------------------------------------------------------------
497 07-30-2014 Intel Corporation
499 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
500 for currently supported client and server kernel versions.
501 * Server known to build on patched kernels:
502 2.6.32-431.20.3.el6 (RHEL6.5)
503 3.0.101-0.31 (SLES11 SP3)
504 * Client known to build on unpatched kernels:
505 2.6.32-431.20.3.el6 (RHEL6.5)
506 3.0.101-0.31 (SLES11 SP3)
509 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
510 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
511 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
512 stack overflow. For more information, please refer to bugzilla 17630.
513 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
514 the re-exporting nodes to be mounted with "32bitapi" mount option
516 --------------------------------------------------------------------------------
518 10-31-2013 Intel Corporation
520 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
521 * Server builds for kernels:
522 2.6.32-358.18.1.el6 (RHEL6.4)
523 * Client builds for unpatched kernels:
524 2.6.32-358.18.1.el6 (RHEL6.4)
525 3.0.92-0.8 (SLES11 SP3)
527 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
528 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
529 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
530 stack overflow. For more information, please refer to bugzilla 17630.
531 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
532 the re-exporting nodes to be mounted with "32bitapi" mount option
534 --------------------------------------------------------------------------------
536 05-30-2013 Intel Corporation
538 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
539 * Server builds for kernels:
540 2.6.32-358.6.2.el6 (RHEL6.4)
541 * Client builds for unpatched kernels:
542 2.6.32-358.6.2.el6 (RHEL6.4)
543 3.0.74-0.6.6 (SLES11 SP2)
545 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
546 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
547 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
548 stack overflow. For more information, please refer to bugzilla 17630.
549 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
550 the re-exporting nodes to be mounted with "32bitapi" mount option
552 --------------------------------------------------------------------------------
554 9-30-2012 Whamcloud, Inc.
556 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
557 * Server support for kernels:
558 2.6.32-279.5.1.el6 (RHEL6)
559 * Client support for unpatched kernels:
560 2.6.32-279.5.1.el6 (RHEL6)
562 2.6.18-238.19.1.el5 (RHEL5)
563 2.6.16.60-0.69.1 (SLES10)
564 * Recommended e2fsprogs version: 1.41.90.wc4
565 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
566 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
567 stack overflow. For more information, please refer to bugzilla 17630.
568 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
569 the re-exporting nodes to be mounted with "32bitapi" mount option
571 Severity : enhancement
573 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
575 Severity : enhancement
577 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
579 Severity : enhancement
581 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
583 Severity : enhancement
585 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
587 Severity : enhancement
589 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
591 --------------------------------------------------------------------------------
593 04-30-2012 Whamcloud, Inc.
595 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
596 * Server support for kernels:
597 2.6.32-220.4.2.el6 (RHEL6)
598 * Client support for unpatched kernels:
599 2.6.32-220.4.2.el6 (RHEL6)
601 2.6.18-238.19.1.el5 (RHEL5)
602 2.6.16.60-0.69.1 (SLES10)
603 * Recommended e2fsprogs version: 1.41.90.wc4
604 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
605 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
606 stack overflow. For more information, please refer to bugzilla 17630.
607 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
608 the re-exporting nodes to be mounted with "32bitapi" mount option
610 Severity : enhancement
612 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
614 Severity : enhancement
616 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
618 Severity : enhancement
620 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
622 Severity : enhancement
624 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
626 --------------------------------------------------------------------------------
628 09-30-2011 Whamcloud, Inc.
630 * Server support for kernels:
631 2.6.18-238.19.1.el5 (RHEL5)
632 2.6.32-131.6.1.el6 (RHEL6)
633 * Client support for unpatched kernels:
634 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
635 2.6.18-238.19.1.el5 (RHEL5)
636 2.6.32-131.6.1.el6 (RHEL6)
637 2.6.29.4-167.fc11 (FC11)
638 2.6.16.60-0.69.1 (SLES10)
640 * Recommended e2fsprogs version: 1.41.90.wc3
641 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
642 Lustre filesystem with 4K stack may cause a stack overflow. For more
643 information, please refer to bugzilla 17630.
644 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
645 mounted with "32bitapi" mount option
647 Severity : enhancement
648 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
650 Severity : enhancement
651 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
653 Severity : enhancement
654 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
657 Frequency : only with SLES10
659 Description: Use OFED "KMP" provided by Novell
660 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
661 Lustre is now built against this package. That means you need to
662 install the ofed-kmp package from Novell for the patchless client.
663 Note that the ofed-kmp that Novell ships may not exactly match the
664 kernel version but should still be compatible.
666 Severity : enhancement
668 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
669 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
671 --------------------------------------------------------------------------------
673 2010-07-15 Oracle, Inc.
675 * Server support for kernels:
676 2.6.18-164.11.1.el5 (RHEL 5)
677 2.6.18-164.11.1.0.1.el5 (OEL 5)
678 * Client support for unpatched kernels:
679 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
680 2.6.18-164.11.1.el5 (RHEL 5),
681 2.6.18-164.11.1.0.1.el5 (OEL 5)
682 2.6.16.60-0.42.8 (SLES 10),
684 2.6.29.4-167.fc11 (FC11)
685 * Recommended e2fsprogs version: 1.41.10-sun2
687 Severity : enhancement
689 Description: Distribute lfs_migrate script with lustre
690 Details : The lfs_migrate mentioned in the manual is now included in the
691 lustre tree in lustre/scripts and will distributed in lustre RPMs.
693 Severity : enhancement
695 Description: Failover nid list added to import proc file.
697 Severity : enhancement
699 Description: "som_preview" mount option required to use SOM
700 Details : The clients will need to use som_preview mount option
704 Description: Lustre does not do 1MB IOs to HW RAID
705 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
707 Severity : enhancement
709 Description: lfs quota output is not very convenient for awk/sed-parsing
710 Details : Some positions in lfs quota output table could be empty or
711 non-empty which made it hard to parse it with scripts, now
712 a dash is put instead of space where there is not supposed
717 Description: Update to OFED 1.5.1
718 Details : OFED 1.5.1 will be used in environments that don't use
719 vendor supplied OFED.
723 Description: Remove lnet. prefix from lctl params display
724 Details : Also renamed lnet.memused to lnet_memused to
729 Description: Use conf_param -d to remove permanant settings.
733 Description: Add list_param -R
734 Details : lctl list_param now supports recursive mode.
737 Bugzilla : 21829, 22312
738 Description: Fix llobdstat.
739 Details : 1. Now updating the write_bytes counter
740 2. Port script fixes from 1.8.
744 Description: decrease the usage of memory on clients.
745 Details : 1. On clients, recycle dentries and inodes unused.
746 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
751 Description: llobdstat fix and enhancement
752 Details : add a counter to set a limit to how many samples will be returned
753 fix a wildcard in the path to limit to obdfilter stats only
757 Description: lfs check only allowed for root
758 Details : Code cleanups and added a regresion test of non-root use of
763 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
764 Details : Fix hr->hr_index overload.
768 Description: Fix FIEMAP functionality.
772 Description: FC11 patchless client support added
776 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
777 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
781 Description: New acc-sm script and various test-framework enhancments.
785 Description: Previously landed patch was reverted.
789 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
793 Description: Add ldiskfs maxdirsize mount option.
797 Description: Fix mkfs.lustre support for 16TB LUNs.
801 Description: Print mmp_check_interval
802 Details : print mmp_check_interval and make it possible to abort
803 mount operation in case it takes too long.
807 Description: Add cascading_rw to lustre mpi tests.
811 Description: Store SOM data in EA.
812 Details : Also several SOM related cleanups.
816 Description: Add refcount to md_open_data
820 Description: Use refcount on lov_request_set to prevent memory corruption.
824 Description: Statahead: use parent dir's i_mutex for synchronization.
828 Description: loadgen improvements.
832 Description: Enable peer health detection in ksocklnd and ko2iblnd.
836 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
840 Description: Fix race on obd_devcie:md_stats
844 Description: Fix SLES 11 client builds.
845 Details : Resolved uninitalized pointer gcc warnings.
849 Description: Allow non-root access for "lfs check".
850 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
854 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
856 Severity : enhancement
858 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
859 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
863 Description: Sanity test_27p fails.
864 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
865 Sanity tests 27 were cleaned up and improved.
869 Description: Unify req format on client/servers
870 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
871 buffers and swabbers approach..
876 Description: drop join file stuff.
878 Severity : enhancement
880 Description: Add 2.6.27 and SLES11 (patchless client) support
884 Description: Fix in ptlrpc_expire_one_request() to print the signed time
887 Severity : enhancement
889 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
892 Severity : enhancement
894 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
896 Severity : enhancement
898 Description: correctly handle big reply message.
899 Details : send LNet event if reply is bigger than buffer and adjust this buffer
904 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
906 Severity : enhancement
908 Description: Add support for OEL5.
912 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
914 Severity : enhancement
916 Description: "weak-modules" support
917 Details : Implement "weak-modules" support which enables kernel modules
918 to be used with any kernel that implements the same kABI. In
919 order to achieve this modules are now installed in
920 /lib/modules/$(uname -r)/updates/kernel on all distributions.
924 Description: client could not reconnect to OST because of an active request.
925 Details : The client now retries to reconnect to the same server, if a
926 connect request failed with EBUSY or -EAGAIN. Test case 35b
927 conf-sanity.sh has been created.
931 Description: Don't allow make backward step on assignin osc next id.
932 Details : race between allocation next id and ll_sync thread can be cause
933 of set wrong osc next id and can be kill valid ost objects.
937 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
941 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
945 Description: Error handling in osc_statfs_interpret() has been improved.
946 Details : Check in osc_statfs_interpret() for EBADR.
950 Description: Do not send statfs() requests to OSTs disabled by administrator.
951 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
953 Severity : enhancement
955 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
957 Severity : enhancement
959 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
961 Severity : enhancement
963 Description: Enhancement for lfs(1) command to use numeric uid/gid.
967 Description: Improved cleanup procedure for conf-sanity test_46a.
971 Description: Do not start bulk transfer, if client tried to reconnect.
972 Details : Do not start bulk transfer, if client tried to reconnect,
973 but failed due to active rpcs. Check for exp_abort_active_req
974 has been implemented.
978 Description: Prevent inconsistences between linux and lustre mount structures.
979 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
980 Make the sleep interruptible.
982 Severity : enhancement
984 Description: provide server to client comms path
985 Details : server to client communications path via new LDLM_SET_INFO rpc
989 Description: truncate starts GFP_FS allocation under transaction causing deadlock
990 Details : ldiskfs_truncate calls grab_cache_page which may start page
991 allocation under an open transaction. This may lead to
992 calling prune_icache with consequent lustre reentrance.
996 Description: send correctly lsm on open replay
997 Details : MDS is trust to LSM size on replay open, but client can set wrong size
1002 Description: enable client interface failover
1003 Details : When a child reconnects from another NID, properly update export
1004 nid hash position and ldlm reverse import.
1006 Severity : enhancement
1008 Description: Communicate OST degraded/readonly state via statfs to MDS
1009 Details : Flags in the statfs returned from OSTs indicate whether the
1010 OST is in a degraded RAID state, or if the filesystem has
1011 turned read-only after a filesystem error is detected.
1016 Description: don't panic if EPROTO was hit when reading symlink
1017 Details : correctly handling request reference in error cases.
1019 Severity : enhancement
1021 Description: Add LustreNetLink, a kernel-userspace communcation path.
1023 Severity : enhancement
1025 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1030 Description: access to llog context before init.
1031 Details : move handling CATALOGS file at osc layer and forbid access to llog
1032 context before init.
1035 Frequency : rare, only when client is writing to OST on local node
1037 Description: Avoid deadlock for local client writes
1038 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1039 memory freeing context. This allows OST threads to set the
1040 PF_MEMALLOC flag on task structures in order to allocate memory
1041 from reserved pools and complete IO.
1042 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1043 so that the OST threads generate memory pressure and allow
1044 inactive pages to be reclaimed.
1046 Severity : enhancement
1048 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1053 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1054 Details : move ldlm namespace creation in setup phase to avoid grab
1055 _lprocfs_lock with cli_sem held.
1059 Description: Temporarily disable grant shrink.
1060 Details : Disable the feature for debugging.
1063 Frequency : only when formatting test filesystems
1065 Description: Unable to run several mkfs.lustre on loop devices at the same time
1066 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1067 formatting. The solution is to proper handle the error.
1069 Severity : enhancement
1071 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1073 Severity : enhancement
1075 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1080 Description: resolve race between obd_disconnect and class_disconnect_exports
1081 Details : if obd_disconnect will be called to already disconnected export he
1082 forget release one reference and osc module can't unloaded.
1084 Severity : enhancement
1086 Description: Allow tuning service thread via /proc
1087 Details : For each service a new
1088 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1089 created that can be used to set min/max thread counts, and get the
1090 current number of running threads.
1095 Descriptoin: don't return error if have partially created objects for file.
1096 Details : lov_update_create_set uses set->set_success as index for created
1097 objects, so if some requests will be failed, they will have hole
1098 at end of array and qos_shrink_lsm for allocate correct lsm.
1100 Severity : enhancement
1102 Description: Update OFED support to 1.4.1
1107 Description: Reduce small size read RPC
1108 Details : Set read-ahead limit for every file and only do read-ahead when
1109 available read-ahead pages are bigger than 1M to avoid small size
1112 Severity : enhancement
1114 Description: move AT tunable parameters for more consistent usage
1115 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1117 Severity : enhancement
1119 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1120 Details : allow skip disconnected ost for send statfs request and hide error
1126 Description: fix racy locking of mballoc block bitmaps causing BUG
1127 Details : The locking of the mballoc buddy bitmap and the in-memory
1128 block bitmap was using two different spin locks in some
1129 cases. This made it possible to incorrectly access the
1130 mballoc bitmap while another process was modifying it,
1131 causing a sanity assertion to fail. While no on-disk corruption
1132 was reported, there was some risk of this happening.
1135 Frequency : rare, on llog test 6
1137 Descriptoin: don't allow connect to already connected import
1138 Details : allowing connect to already connected import is hide connecting problem.
1143 Descriptoin: allow kill process which wait statahead result
1144 Details : for some reasons 'ls' can stick in waiting result from statahead,
1145 in this case need way for kill this process.
1147 Severity : enhancement
1149 Description: Add state history info file, enhance import info file
1150 Details : Track import connection state changes in a new osc/mdc proc file;
1151 add overview-type data to the osc/mdc import proc file.
1153 Severity : enhancement
1155 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1157 Severity : enhancement
1159 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1164 Descriptoin: OST grant too much space to client even there are not enough space.
1165 Details : Client will shrink its grant cache to OST if there are no write
1166 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1167 this grant cache if there are already not enough avaible space
1168 (left_space < total_clients * 32M).
1171 Frequency : start MDS on uncleanly shutdowned MDS device
1173 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1174 Details : stay in waiting mds<>ost recovery finished produce random bugs
1175 due race between two ll_sync thread for one lov target. send
1176 ACTIVATE event only if connect realy finished and import have
1180 Frequency : rare, connect and disconnect target at same time
1182 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1183 Details : don't call obd_disconnect under lov_lock. this long time
1184 operation and can block ptlrpcd which answer to connect request.
1189 Descriptoin: don't lose wakeup for imp_recovery_waitq
1190 Details : recover_import_no_retry or invalidate_import and import_close can
1191 both sleep on imp_recovery_waitq, but we was send only one wakeup
1195 Frequency : always with long access acl
1197 Descriptoin: mds can't pack reply with long acl.
1198 Details : mds don't control size of acl but they limited by reint/getattr
1201 Severity : enhancement
1203 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1205 Severity : enhancement
1207 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1210 Frequency : start MDS on uncleanly shutdowned MDS device
1212 Descriptoin: aborting recovery hang on MDS
1213 Details : don't throttle destroy RPCs for the MDT.
1216 Frequency : on remount
1218 Description: external journal device not working after the remount
1219 Details : clear dev_rdonly flag for external journal devices in
1225 Description: shutdown vs evict race
1226 Details : client_disconnect_export vs connect request race.
1227 if client will evicted at this time - we start invalidate
1228 thread without referece to import and import can be freed
1232 Frequency : rare, need acl's on inode.
1234 Description: client can't handle ost additional correctly
1235 Details : if ost was added after client connected to mds client can have
1236 hit lnet_try_match_md ... to big messages to wide striped files.
1237 in this case need teach client to handle config events about add
1238 lov target and update client max ea size at that event.
1240 Severity : enhancement
1242 Description: Changelogs
1243 Details : Changelogs are a lightweight mechanism to track filesystem
1244 metadata and namespace changes. The changelog is recorded
1245 permanently on the MDTs, and is periodically "consumed" / purged
1246 when records are no longer needed.
1248 Severity : enhancement
1250 Description: compact fld format with extents
1251 Details : Store range of seq rather than every seq in FLD. Seq
1252 controller update FLD rather than clients. In Case of CMD, mdt0
1253 has FLD, all other metadata server act as non persistent proxy
1254 for FLD queries and cache fld entries in fld cache.
1259 Description: don't skip ost target if they assigned to file
1260 Details : Drop slow OSCs if we can, but not for requested start idx.
1261 This means "if OSC is slow and it is not the requested
1262 start OST, then it can be skipped, otherwise skip it only
1263 if it is inactive/recovering/out-of-space.
1267 Description: more cleanup in mds_lov
1268 Details : not send LOV EA under replay, we can't know about they size at this
1269 time. Don't allow client connect to mds before any ost connected,
1270 for avoid problems with LOV EA size and returning EIO to client.
1272 Severity : enhancement
1274 Description: Interoperability at server side (Disk interoperability)
1276 Severity : enhancement
1278 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1280 Severity : enhancement
1282 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1284 Severity : enhancement
1286 Description: New client IO stack (CLIO).
1288 Severity : enhancement
1290 Description: Commit on sharing. Eliminate inter-client dependencies between
1291 uncommitted transactions by doing transaction commits.
1292 Thereby clients may recovery independently.
1295 Frequency : Create a symlink file with a very long name
1297 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1298 Details : If there is no extra space in the request for early cancels,
1299 ldlm_req_handles_avail() returns 0 instead of a negative value.
1301 Severity : enhancement
1303 Description: Add /proc entry for import status
1304 Details : The mdc, osc, and mgc import directories now have
1305 an import directory that contains useful import data for debugging
1306 connection problems.
1308 Severity : enhancement
1310 Description: Re-disable certain /proc logging
1311 Details : Enable and disable client's offset_stats, extents_stats and
1312 extents_stats_per_process stats logging on the fly.
1315 Frequency : Only on FC kernels 2.6.22+
1317 Description: oops in statahead
1318 Details : Do not drop reference count for the dentry from VFS when lookup,
1319 VFS will do that by itself.
1321 Severity : enhancement
1323 Description: Generic /proc file permissions
1324 Details : Set /Proc file permissions in a more generic way to enable non-
1325 root users operate on some /proc files.
1329 Description: Hitting mdc_commit_close() ASSERTION
1330 Details : Properly handle request reference release in
1331 ll_release_openhandle().
1335 Description: quota recovery deadlock during mds failover
1336 Details : This patch includes att18982, att18236, att18237 in bz14840.
1338 1. fix osts hang when mds does failover with quotaon
1339 2. prevent watchdog storm when osts threads wait for the
1344 Frequency : only patchless client
1345 Description: add workaround for race between add/remove dentry from hash
1347 Severity : enhancement
1349 Description: Allow OST glimpses to return PW locks
1353 Description: LBUG when llog conf file is full
1354 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1358 Description: Prevent import from entering FULL state when server in recovery
1362 Description: service mount cannot take device name with ":"
1363 Details : Only when device name contains ":/" will mount treat it as
1369 Description: replace ptlrpcd with the statahead thread to interpret the async
1370 statahead RPC callback
1374 Frequency : on recovery
1375 Description: I/O failures after umount during fail back
1376 Details : if client reconnected to restarted server we need join to recovery
1377 instead of find server handler is changed and process self
1378 eviction with cancel all locks.
1380 Severity : enhancement
1382 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1384 Severity : enhancement
1386 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1388 Severity : enhancement
1390 Description: Upcall on Lustre log has been dumped
1391 Details : Allow for a user mode script to be called once a Lustre log has
1392 been dumped. It passes the filename of the dumped log to the
1393 script, the location of the script can be specified via
1394 /proc/sys/lnet/debug_log_upcall.
1399 Description: avoid idr_remove called for id which is not allocated.
1400 Details : Move assigment s_dev for clustered nfs to end of initialization,
1401 to avoid problem with error handling.
1406 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1407 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1408 into CONN_USED_HASH and this prodice warning when put connection
1409 again in unused hash.
1412 Severity : enhancement
1414 Description: File striping can now be set to use an arbitrary pool of OSTs.
1416 Severity : enhancement
1418 Description: Export bytes_read/bytes_write count on OSC/OST.
1422 Description: Early reply size mismatch, MGC loses connection
1423 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1424 the connect flags are properly negotiated.
1429 Description: quotas are not honored with O_DIRECT
1430 Details : all writes with the flag O_DIRECT will use grants which leads to
1431 this problem. Now using OBD_BRW_SYNC to guard this.
1435 Description: add quota statistics
1436 Details : 1. sort out quota proc entries and proc code.
1437 2. add quota statistics
1439 Severity : enhancement
1441 Description: enable quota support for HEAD.
1445 Description: Properly propagate oinfo flags from lov to osc for statfs
1446 Details : restore missing copy oi_flags to lov requests.
1448 Severity : enhancement
1450 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1451 : l_getgroups(8), lst(8), routerstat(8)
1452 Details : included man pages for llobdstat(8), llstat(8),
1453 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1455 Severity : enhancement
1457 Description: configure's --enable-quota should check the
1458 : kernel .config for CONFIG_QUOTA
1459 Details : configure is terminated if --enable-quota is passed but
1460 : no quota support is in kernel
1464 Description: Remove portals compatibility
1465 Details : Remove portals compatibility, not interoperable with releases
1470 Description: Resolve device initialization race
1471 Details : Prevent proc handler from accessing devices added to the
1472 obd_devs array but yet be intialized.
1474 Severity : enhancement
1476 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1478 Severity : enhancement
1480 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1485 Description: Using wrong pointer in osc_brw_prep_request
1486 Details : Access to array[-1] can produce panic if kernel compiled with
1487 CONFIG_PAGE_ALLOC enabled
1491 Description: Client runs out of low memory
1492 Details : Consider only lowmem when counting initial number of llap pages
1496 Description: *optional* service tags registration
1497 Details : if the "service tags" package is installed on a Lustre node
1498 When the filesystem is mounted, a local-node service tag will
1499 be created. See http://inventory.sun.com/ for more information
1500 about the Service Tags asset management system.
1504 Description: Kernel BUG tries to release flock
1505 Details : Lustre does not destroy flock lock before last reference goes
1506 away. So always drop flock locks when client is evicted and
1507 perform unlock regardless of successfulness of speaking to MDS.
1511 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1515 Description: sanity test 65a fails if stripecount of -1 is set
1516 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1521 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1522 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1525 Severity : enhancement
1527 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1531 Description: do not process already freed flock
1532 Details : flock can possibly be freed by another thread before it reaches
1533 to ldlm_flock_completion_ast.
1537 Description: LBUG during stress test
1538 Details : Need properly lock accesses the flock deadlock detection list.
1542 Description: oops in page fault handler
1543 Details : kernel page fault handler can return two special 'pages' in error
1544 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1548 Description: timeout with invalidate import.
1549 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1550 should be handled by ptlrpcd. This produce long age waiting and
1551 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1553 Severity : enhancement
1555 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1559 Description: enable MGS and MDT services start separately
1560 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1561 a co-located MGS without starting the MGS, which is a complement
1562 to 'nosvc' mount option.
1566 Frequency : after recovery
1567 Description: precreate to many object's after del orphan.
1568 Details : del orphan st in oscc last_id == next_id and this triger growing
1569 count of precreated objects. Set flag LOW to skip increase count
1570 of precreated objects.
1574 Frequency : rare, on clear nid stats
1575 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1576 Details : when clean nid stats sometimes try destroy live entry,
1577 and this produce panic in free.
1581 Description: Stack overflow during MDS log replay
1582 ease stack pressure by using a thread dealing llog_process.
1586 Description: wait until IO finished before start new when do lock cancel.
1587 Details : VM protocol want old IO finished before start new, in this case
1588 need wait until PG_writeback is cleared until check dirty flag and
1589 call writepages in lock cancel callback.
1591 Severity : enhancement
1593 Description: using special macro for print time and cleanup in includes.
1597 Description: mds_mfd_close() ASSERTION(rc == 0)
1598 Details : In mds_mfd_close(), we need protect inode's writecount change
1599 within its orphan write semaphore to prevent possible races.
1603 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1607 Frequency : rare, on shutdown ost
1608 Description: don't hit live lock with umount ost.
1609 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1610 use shrink_dcache_sb instead.
1614 Description: don't panic with use echo client
1615 Details : echo client pass NULL as client nid pointer and this produce null
1616 pointer dereference.
1620 Description: fix build on ppc32
1621 Details : compile code with -m64 flag produce wrong object file for ppc32.
1625 Description: add message levels for liblustreapi
1629 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1630 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1631 will result in an -EAGAIN and in the exit call path we call
1632 alloc_rr() which will with increasing aggressiveness attempt to
1633 aquire precreated objects on the minimum number of required OSCs.
1637 Description: Use old size assignment to avoid deadlock
1638 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1639 the scheduling while holding a spinlock. We do not need locking
1640 for size in ll_update_inode() because size is only updated from
1641 the MDS for directories or files without objects, so there is no
1642 other place to do the update, and concurrent access to such inodes
1643 are protected by the inode lock.
1647 Description: resolve "_IOWR redefined" build error on SLES10
1651 Description: dump the memory debugging after all modules are unloaded to
1652 suppress false negative in conf_sanity test 39
1654 Severity : enhancement
1656 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1661 Description: fix SLES kernel versioning
1662 Details : the kernel version for our SLES 10 kernel did not include a "-"
1663 before the "smp" at the end. while this was not a problem in
1664 general, it did mean that software trying to use the kernel
1665 version to try to detect a vendor specific kernel would fail.
1666 this was most evident by the OFED build scripts.
1670 Description: Don't update lov_desc members until making sure they are valid
1671 Details : When updating lov_desc members via proc fs, need fix their
1672 validities before doing the real update.
1676 Description: don't put request into delay list while invalidate in flight.
1677 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1678 import in flight. this produce timeout for invalidate and sometimes
1679 can cause stale data.
1683 Frequency : on ppc only
1684 Description: not convert ost objects for directory because it's not exist.
1685 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1689 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1691 Severity : enhancement
1693 Description: Update to support 2.6.22.14 vanilla kernel.
1697 Frequency : rare, on recovery
1698 Description: read procfs can produce deadlock in some situation
1699 Details : Holding lprocfs lock which send rpc can produce block for destroy
1700 obd objects and this also block reconnect with -EALREADY. This
1701 isn't fix all lprocfs bugs - but make it rare.
1703 Severity : enhancement
1705 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1708 Frequency : frequent on X2 node
1710 Description: mdc_set_open_replay_data LBUG
1711 Details : Set replay data for requests that are eligible for replay.
1715 Description: lustre_mgs: operation 101 on unconnected MGS
1716 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1717 MGC, and late on MGC cannot successfully connect to MGS and a lot
1718 of the error messages complaining that MGS is not connected.
1722 Frequency : on network error
1723 Description: panic with double free request if network error
1724 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1725 but it's true only for synchronus enqueue, for async enqueue
1726 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1729 Severity : enhancement
1731 Description: client-side metadata stat-ahead during readdir(directory readahead)
1732 Details : perform client-side metadata stat-ahead when the client detects
1733 readdir and sequential stat of dir entries therein
1736 Frequency : on start mds
1738 Description: Implement get_info(last_id) in obdfilter.
1741 Frequency : occasional
1743 Description: Correctly check stale fid, not start epoch if ost not support SOM
1744 Details : open with flag O_CREATE need set old fid in op_fid3 because
1745 op_fid2 was overwritten with new generated fid, but mds can answer
1746 with one of these two fids and both is not stale. Setattr
1747 incorrectly started an epoch and assume will be called
1748 done_writing, but without SOM done_writing ever being called.
1751 Frequency : rare, depends on device drivers and load
1753 Description: MDS or OSS nodes crash due to stack overflow
1754 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1755 In some cases, in conjunction with device drivers that use a lot
1756 of stack the MDS (or possibly OSS) service threads could overflow
1757 the stack. One change which was identified to consume additional
1758 stack has been reworked to avoid the extra stack usage.
1761 Frequency : occasional
1763 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1764 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1765 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1766 oscc_has_objects and see OSC has no objests and call
1767 oscc_internal_create to resend create request.
1769 Severity : enhancement
1771 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1773 Severity : enhancement
1775 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1778 Frequency : very rare
1780 Description: Fix replay if there is an un-replied request and open
1781 Details : In some cases, older replay request will revert the
1782 mcd->mcd_last_xid on MDS which is used to record the client's
1783 latest sent request.
1785 Severity : enhancement
1787 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1789 Severity : enhancement
1791 Description: Add rhel5 support to HEAD.
1793 Serverity : enhancement
1795 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1800 Description: Don't allow skipping OSTs if index has been specified.
1801 Details : Don't allow skipping OSTs if index has been specified, make
1802 locking in internal create lots better.
1806 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1807 Details : access to bitfield in structure is always rounded to long
1808 and this produce problem with not atomic change any bit.
1812 Description: Lustre make rpms failed.
1813 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1814 builds Lustre rpms from tarball.
1817 Frequency : rare on shutdown ost
1819 Description: If llog cancel was not send before clean_exports phase, this can
1820 produce deadlock in llog code.
1821 Details : If llog thread has last reference to obd and call class_import_put
1822 this produce deadlock because llog_cleanup_commit_master wait when
1823 last llog_commit_thread exited, but this never success because was
1824 called from llog_commit_thread.
1828 Description: allow userland application know is lost one of stripes.
1829 Details : fill lvb_blocks with error code on ost and return it to
1830 application if error flag found.
1834 Description: NULL lov_tgts causing MDS oops
1835 Details : more safe checks for NULL lov_tgts for avoid oops.
1837 Severity : enhancement
1839 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1843 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1844 Details : If an OST is down the MDS will hang indefinitely in
1845 obd_statfs_rqset() waiting for the statfs data. While for
1846 MDS QOS usage of statfs, it should not stuck in waiting.
1848 Severity : enhancement
1850 Description: remote_acl support
1851 Details : Support ACL-based permission check for remote user.
1852 Support setfacl/getfacl for remote user with the utils
1853 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1854 the system "{s,g}etfacl" utils.
1856 Severity : enhancement
1858 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1860 Severity : enhancement
1862 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1866 Description: make mgs_setparam() handle fsname containing dash
1867 Details : fsname containing a dash does not work with lctl conf_param
1869 Severity : enhancement
1871 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1873 Severity : enhancement
1875 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1879 Frequency : rare, at shutdown
1880 Description: access already free / zero obd_namespace.
1881 Details : if client_disconnect_export was called without force flag set,
1882 and exist connect request in flight, this can produce access to
1883 NULL pointer (or already free pointer) when connect_interpret
1884 store ocd flags in obd_namespace.
1888 Frequency : only at startup
1889 Description: not alloc memory with spinlock held.
1890 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1891 if any spinlock held.
1893 Severity : enhancement
1895 Description: make lustre randomly fail allocating memory
1896 Details : Make lustre randomly failed allocating memory for testing purpose.
1898 Severity : enhancement
1900 Description: lost problems with lov objid file
1901 Details : Fixes some scability and access to not inited memory problems
1902 in work with lov objdid file.
1907 Description: lfs find does not continue on file error
1908 Details : Continue other files processing when a file/dir is absent.
1912 Description: Inconsistent usage of lustre_pack_reply()
1913 Details : Standardize the usage of lustre_pack_reply() such that it
1914 always generate a CERROR on failure.
1917 Frequency : occasional
1919 Description: MDS hang or stay in waiting lock
1920 Details : If client receive lock with CBPENDING flag ldlm need send lock
1921 cancel as separate rpc, to avoid situation when cancel request
1922 can't processed due all i/o threads stay in wait lock.
1926 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1930 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1934 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1938 Description: Update to RHEL4 latest kernel.
1940 Severity : enhancement
1942 Description: Build SLES10 patchless client fails
1943 Details : The configure was broken by run ./configure with
1944 --with-linux-obj=.... argument for patchless client. When the
1945 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1946 can't search header adequately. Use absolute path such as
1947 -I($LINUX)/include instead.
1951 Description: interrupt oig_wait produce painc on resend.
1952 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1953 private set, and this produce situation when rq_ptlrpcd_data not
1954 copyed to new allocated request and triggered LBUG on assert
1955 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1956 ptlrpcd set if request is changed and can be safety changed to use
1961 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1962 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1963 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1964 these macros and replace them with their difinitions instead.
1968 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1969 Details : Update which_patch & target file for SLES10 latest kernel.
1972 Description: Remove client patches from SLES 10 kernel.
1973 Details : This causes SLES 10 clients to behave as patchless clients
1974 even on a Lustre-patched (server) kernel.
1976 Severity : enhancement
1978 Description: self-adjustable client's lru lists
1979 Details : use adaptive algorithm for managing client cached locks lru
1980 lists according to current server load, other client's work
1981 pattern, memory activities, etc. Both, server and client
1982 side namespaces provide number of proc tunables for controlling
1985 Severity : enhancement
1987 Description: light-weight GSS support
1988 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1989 and reduce performance overhead.
1991 Severity : enhancement
1993 Description: Linux keyring support
1994 Details : Support using service of Linux keyring for Lustre GSS internal
1995 context refresh/cache mechanism.
1999 Description: Fix errors in lfs documentation
2000 Details : Fixes man pages
2004 Description: don't use GFP_* in generic Lustre code.
2005 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2009 Description: obdclass is limited by single OBD_ALLOC(idarray)
2010 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2014 Description: warnings with build patchless client with vanila 2.6.19 and up
2015 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2016 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2020 Frequency : Only for SLES
2021 Description: sanity_quota fail test_1
2022 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2023 file, which makes TSTID[2] inunique.
2027 Frequency : Only for SLES9
2028 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2029 Details : Update target/ChangeLog/which_patch .
2033 Frequency : Only for test_10 in sanity-quota.sh
2034 Description: a bug in quota test code
2035 Details : checking a test flag in a wrong place
2039 Frequency : Only for test_13 in sanity-quota.sh
2040 Description: a bug in quota test script
2041 Details : shouldn't check overfull allocated quota
2045 Description: jbd statistics
2046 Details : Port older jbd statistics patch for sles10
2048 Severity : enhancement
2050 Description: Add printing inode info into message about error in writepage.
2055 Description: reply_lock_interpret crash due to race with it and lock cancel.
2056 Details : Do not replay locks that are being cancelled. Do not reference
2057 locks by their address during replay, just by their handle.
2061 Frequency : When flocks are used.
2062 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2063 Details : Flock locks might destroy just granted lock if it could be merged
2064 with another existing flock, this is done in completion handler,
2065 so teach ldlm_cli_enquque_fini that this is a valid case for
2071 Description: Oops in read and write path when failing to allocate lock.
2072 Details : Check if lock allocation failed and return error back.
2076 Description: lstripe command fails for valid OST index
2077 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2078 instead of lov->desc.ld_active_tgt_count.
2082 Description: sanity.sh failed test 103
2083 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2088 Description: df doesn't work properly if diskfs blocksize != 4K
2089 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2093 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2094 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2096 Severity : enhancement
2098 Description: Adaptive timeouts
2099 Details : RPC timeouts adapt to changing server load and network
2100 conditions to reduce resend attempts and improve recovery time.
2104 Description: llapi_file_create() does not allow some changes
2105 Details : add llapi_file_open() that allows specifying the mode and
2106 open flags, and also returns an open file handle.
2110 Description: merge and cleanup kernel patches.
2111 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2115 Description: Add journal checksum support.(Kernel part)
2116 Details : The journal checksum feature adds two new flags i.e
2117 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2118 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2119 indicates that the commit block contains the checksum for
2120 the blocks described by the descriptor blocks. Now commit
2121 record can be sent to disk without waiting for descriptor
2122 blocks to be written to disk. This behavior is controlled
2123 using JBD2_FEATURE_ASYNC_COMMIT flag.
2127 Description: OSS needs mutliple precreate threads
2128 Details : Add ability to start more than one create thread per OSS.
2132 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2133 Details : execute lfs setstripe on client
2137 Description: mds_obd_create error creating tmp object
2138 Details : When the user sets quota on root, llog will be affected and can't
2139 create files and write files.
2142 Frequency : Always on ia64 patchless client, and possibly others.
2144 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2145 Details : This allows the patchless client to be loaded on architectures
2146 without this export.
2150 Description: RedHat Update kernel for RHEL5
2151 Details : Add the kernel config file for RHEL5.
2155 Description: RedHat Update kernel for RHEL5
2156 Details : Modify the kernel config file more closer RHEL5.
2160 Description: Build failure against Centos5 (RHEL5)
2161 Details : Define PAGE_SIZE when it isn't present.
2165 Description: test_7 on sanity-quota.sh can't be used on separate machines
2166 Details : add facet to handle it
2170 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2171 Details : with NFS, the anon dentry's parent was set to itself in
2172 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2173 corresponding dentry other than use rec->ur_name.
2175 Severity : enhancement
2177 Description: lfs setstripe enhancement
2178 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2182 Description: enable data checksumming by default
2183 Details : enable checksum by default, allow --disable-checksum
2184 configure option and "-o nochecksum" mount option. Checksums
2185 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2189 Description: lustre support for RHEL5
2190 Details : Add support for RHEL5.
2194 Description: Client eviction due to failover config
2195 Details : after a connection loss, the lustre client should attempt to
2196 reconnect to the last active server first before trying the
2197 other potential connections.
2201 Description: when mds and osts use different quota unit(32bit and 64bit),
2202 quota will be released repeatly.
2203 Details : void sending multiple quota reqs to mds, which will keep the
2204 status between the reqs.
2208 Description: osts not allocated evenly to files
2209 Details : change the condition to increase offset_idx
2212 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2214 Description: Data corruption for OSTs that are formatted larger than 2TB
2216 Details : When generating the bio request for lustre file writes the
2217 sector number would overflow a temporary variable before being
2218 used for the IO. The data reads correctly from Lustre (which
2219 will overflow in a similar manner) but other file data or
2220 filesystem metadata may be corrupted in some cases.
2224 Description: Tune the kernel for good SCSI performance.
2225 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2226 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2231 Description: rewrite ext2-derived code in obdclass/uuid.c
2232 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2233 from scratch preserving functionality.
2237 Description: "lfs find -obd UUID" prints directories
2238 Details : "lfs find -obd UUID" will return all directory names instead
2239 of just file names. It is incorrect because the directories
2240 do not reside on the OSTs.
2244 Description: use i_size_read and i_size_write in 2.6 port
2245 Details : replace inode->i_size access with i_size_read/write()
2248 Frequency : when removing large files
2250 Description: scheduling issue during removal of large Lustre files
2251 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2252 It causes scheduling issues when removing large files (17TB in the
2256 Frequency : only with liblustre clients on XT3
2258 Description: evictions taking too long
2259 Details : allow llrd to evict clients directly on OSTs
2262 Frequency : only on ppc
2264 Description: /proc/fs/lustre/devices broken on ppc
2265 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2266 We should make sure the type of 'index' is loff_t and then cast
2267 explicitly as needed below. Do not assign an explicitly cast
2271 Frequency : only for rhel5
2273 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2274 Details : Modify the target file & which_kernel.
2276 Severity : enhancement
2278 Description: omit set fsid for export NFS
2279 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2280 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2284 Frequency : Always, for kernels after 2.6.16
2285 Description: Fix warning idr_remove called for id=.. which is not allocated.
2286 Details : Last kernels save old s_dev before kill super and not allow
2287 to restore from callback - restore it before call kill_anon_super.
2291 Description: buffer overruns could theoretically occur
2292 Details : llapi_semantic_traverse() modifies the "path" argument by
2293 appending values to the end of the origin string, and a
2294 overrun may occur. Adding buffer overrun check in liblustreapi.
2298 Description: Fix error on 'ls .' at the top of the Lustre mount.
2299 Details : Don't revalidate dentry if it is a root dentry.
2303 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2304 Details : Modify vm-tunables-rhel4.patch.
2308 Description: Kernel config for 2.6.18-vanilla.
2309 Details : Modify targets/2.6-vanilla.target.in.
2310 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2311 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2312 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2313 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2317 Description: improve handling recoverable errors
2318 Details : if request processig with error which can be recoverable on server
2319 request should be resend, otherwise page released from cache and
2324 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2325 Details : Modify target file & which_patch.
2326 A flaw was found in the IA32 system call emulation provided
2327 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2328 value could be stored in the %RAX register, which could trigger an
2329 out-of-bounds system call table access. An untrusted local user
2330 could exploit this flaw to run code in the kernel
2331 (ie a root privilege escalation). (CVE-2007-4573).
2335 Description: O_DIRECT bypasses client statistics.
2336 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2337 read_ahead_stats were updated. copied the stats section from
2338 osc_send_oap_rpc() into async_internal().
2342 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2346 Description: change order of libsysio includes
2347 Details : '#include sysio.h' should always come before '#include xtio.h'
2351 Description: handle "serious error: objid * already exists" more gracefully
2352 Details : If LAST_ID value on disk is smaller than the objects existing in
2353 the O/0/d* directories, it indicates disk corruption and causes an
2354 LBUG(). If the object is 0-length, then we should use the existing
2355 object. This will help to avoid a full fsck in most cases.
2357 Severity : enhancement
2359 Description: adapt the lustre_config script to support the upgrade case
2360 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2361 server targets to 1.6 in parallel.
2365 Description: To avoid grant space > avaible space when the disk is almost
2366 full. Without this patch you might see the error "grant XXXX >
2367 available" or some LBUG about grant, when the disk is almost
2369 Details : In filter_check_grant, for non_grant cache write, we should
2370 check the left space by if (*left > ungranted + bytes), instead
2371 of (*left > ungranted), because only we are sure the left space
2372 is enough for another "bytes", then the ungrant space should be
2373 increase. In client, we should update cl_avail_grant only there
2374 is OBD_MD_FLGRANT in the reply.
2378 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2379 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2380 described in CVE-2007-4573. This problem could allow untrusted
2381 local users to gain root access.
2384 Frequency : when using O_DIRECT and quotas
2386 Description: Incorrect file ownership on O_DIRECT output files
2387 Details : block usage reported by 'lfs quota' does not take into account
2388 files that have been written with O_DIRECT.
2393 Description: touch file failed when fs is not full
2394 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2395 otherwise we can get ENOSP while fs is not full.
2399 Description: parallel lock callbacks
2400 Details : Instead of sending blocking and completion callbacks as separated
2401 requests, adding them to a set and sending in parallel.
2404 Frequency : only for Cray XT3
2405 Bugzilla : 12829/13455
2406 Description: Changing primary group doesn't change the group lustre assigns to
2408 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2409 overridden with the primary group provided by the group upcall,
2410 whereas the supplied fsgid can be trusted if it is in the list of
2411 supplementary groups returned by the group upcall.
2413 Severity : enhancement
2415 Description: Allow masking D_WARNING, D_ERROR messages from console
2416 Details : Console messages can now be disabled via lnet.printk.
2420 Description: User code with malformed file open parameter crashes client node
2421 Details : Before packing join_file req, all the related reference should be
2422 checked carefully in case some malformed flags cause fake
2423 join_file req on client.
2427 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2428 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2429 mode is equal to the granted mode. However, it should also check
2430 that LDLM_FL_CP_REQD is not set.
2434 Description: Heavy nfs access might result in deadlocks
2435 Details : After ELC code landed, it is now improper to enqueue any mds
2436 locks under och_sem, because enqueue might want to decide to
2437 cancel open locks for same inode we are holding och_sem for.
2441 Description: Client eviction while running blogbench
2442 Details : A lot of unlink operations with concurrent I/O can lead to a
2443 deadlock causing evictions. To address the problem, the number of
2444 oustanding OST_DESTROY requests is now throttled to
2445 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2446 callbacks are processed in priority.
2450 Description: enable ACLs on MDS by default
2451 Details : ACLs must be enabled on MDS by default.
2454 Frequency : PPC/PPC64 only
2456 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2457 Details : fix duplicated definitions between asm-ppc64/types.h and
2458 lustre_types.h on PPC.
2461 Frequency : PPC/PPC64 only
2463 Description: asm-ppc/segment.h does not exist
2464 Details : fix compile issue on PPC.
2468 Description: better handle error messages in extents code
2471 Frequency : RHEL4 only
2473 Description: mkfs is very slow on IA64/RHEL4
2474 Details : A performance regression has been discovered in the MPT Fusion
2475 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2476 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2477 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2479 Severity : enhancement
2481 Description: SNMP support enhancement
2482 Details : Adding total number of sampled request for an MDS node in snmp
2485 Severity : enhancement
2487 Description: Optimize ldlm waiting list processing for PR extent locks
2488 Details : When processing waiting list for read extent lock and meeting read
2489 lock that is same or wider to it that is not contended, skip
2490 processing rest of the list and immediatelly return current
2491 status of conflictness, since we are guaranteed there are no
2492 conflicting locks in the rest of the list.
2496 Description: Time out and refuse to reconnect
2497 Details : When the failover node is the primary node, it is possible
2498 to have two identical connections in imp_conn_list. We must
2499 compare not conn's pointers but NIDs, otherwise we can defeat
2500 connection throttling.
2504 Description: port llog fixes from b1_6 into HEAD
2505 Details : Port llog reference couting and some llog cleanups from b1_6
2506 (bug 10800) into HEAD, for protect from panic and access to already
2507 free llog structures.
2511 Description: Detect stride IO mode in read-ahead
2512 Details : When a client does stride read, read-ahead should detect that and
2513 read-ahead pages according to the detected stride pattern.
2517 Description: data checksumming impacts single node performance
2518 Details : add support for several checksum algorithm. Currently, only CRC32
2519 and Adler-32 are supported. The checksum type can be changed on
2520 the fly via /proc/fs/lustre/osc/*/checksum_type.
2524 Description: use adler32 for page checksums
2525 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2530 Description: build for x2 fails
2531 Details : fix compile issue on Cray systems.
2535 Description: Properly match for duplicate locks
2536 Details : Due to different lock order from skiplists code, we need to
2537 traverse entire list for now
2540 Frequency : only on PPC/SLES10
2542 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2543 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2544 be defined. Add a hack in mkfs_lustre.c to work around this compile
2549 Description: LASSERT on MDS when client holding flock lock dies
2550 Details : ldlm pool logic depends on number of granted locks equal to
2551 number of released locks which is not true for flock locks, so
2552 just exclude such locks from consideration.
2556 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2557 Details : Use fsfilt_sync() for both the whole filesystem sync and
2558 individual file sync to eliminate dangerous inode locking
2559 with I_LOCK that can lead to a deadlock.
2563 Description: performance in 1.6.3
2564 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2565 sure that 1MB requests are not fragmented by the block layer.
2567 Severity : enhancement
2569 Description: organize the server-side client stats on per-nid basis
2570 Details : Change the structure of stats under obdfilter and mds to
2578 The "uuid"s file would list the uuids of _active_ exports.
2579 And the clear entry is to clear all stats and stale nids.
2581 Severity : enhancement
2583 Description: eliminate client locks in face of contention
2584 Details : file contention detection and lockless i/o implementation
2585 for contended files.
2589 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2592 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2594 Description: MDS may lose file striping (and hence file data) in some cases
2595 Details : If there are additional extended attributes stored on the MDS,
2596 in particular ACLs, SELinux, or user attributes (if user_xattr
2597 is specified for the client mount options) then there is a risk
2598 of attribute loss. Additionally, the Lustre file striping
2599 needs to be larger than default (e.g. striped over all OSTs),
2600 and an additional attribute must be stored initially in the
2601 inode and then increase in size enough to be moved to the
2602 external attribute block (e.g. ACL growing in size) for file
2607 Description: skiplist implementation simplification
2608 Details : skiplists are used to group compatible locks on granted list
2609 that was implemented as tracking first and last lock of each
2610 lock group the patch changes that to using doubly linked lists
2614 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2615 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2616 because LASSERT on both the data supplied by a client, and the
2617 data on disk is dangerous and incorrect.
2619 Severity : enhancement
2621 Description: Slow truncate/writes to huge files at high offsets.
2622 Details : Directly associate cached pages to lock that protect those pages,
2623 this allows us to quickly find what pages to write and remove
2624 once lock callback is received.
2628 Description: more ldlm soft lockups
2629 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2630 starve other threads from the resource lock for a long time in
2631 case of long waiting queue, so change the debug level from
2632 D_OTHER to the less frequently used D_INFO.
2634 Severity : enhancement
2636 Description: add -gid, -group, -uid, -user options to lfs find
2640 Description: Hung threads in invalidate_inode_pages2_range
2641 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2642 once one is completed. As a result, some RPCs are stuck in the
2643 queue and are never sent.
2647 Description: filter threads hungs on waiting journal commit
2648 Details : Cleanup filter group llog code, then only filter group llog will
2649 be only created in the MDS/OST syncing process.
2653 Description: Procfs and llog threads access destoryed import sometimes.
2654 Details : Sync the import destoryed process with procfs and llog threads by
2655 the import refcount and semaphore.
2657 Severity : enhancement
2659 Description: openlock cache of b1_6 port to HEAD
2664 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2665 Details : If insertion of an extent fails, then discard the inode
2666 preallocation and free data blocks else it can lead to duplicate
2671 Description: don't always update ctime in ext3_xattr_set_handle()
2672 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2673 In some cases the ctime should not be updated, for example for
2674 2.0->1.8 compatibility it is necessary to delete an xattr and it
2675 should not update the ctime.
2679 Bugzilla : 15713/16362
2680 Description: Assertion in iopen_connect_dentry in 1.6.3
2681 Details : looking up an inode via iopen with the wrong generation number can
2682 populate the dcache with a disconneced dentry while the inode
2683 number is in the process of being reallocated. This causes an
2684 assertion failure in iopen since the inode's dentry list contains
2685 both a connected and disconnected dentry.
2689 Description: assertion failure in ldlm_handle2lock()
2690 Details : fix a race between class_handle_unhash() and class_handle2object()
2691 introduced in lustre 1.6.5 by bug 13622.
2696 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2697 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2702 Description: oops in sd_iostats_seq_show()
2703 Details : unloading/reloading the scsi low level driver triggers a kernel
2704 bug when trying to access the sd iostat file.
2709 Description: Kernel panics during QLogic driver reload
2710 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2711 patch, causing memory corruption.
2716 Description: journal_dev option does not work in b1_6
2717 Details : pass mount option during pre-mount.
2719 Severity : enhancement
2721 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2722 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2723 extent information of a file. It can be used to map logical blocks
2724 in a file to physical blocks in the block device.
2728 Description: LDLM soft lockups - improvement
2729 Details : It is be possible to send the lock handle along with each read
2730 or write request because the client is already doing a lock match
2731 itself so there isn't any reason the OST should have to re-do that
2735 Frequency : only on Cray X2
2737 Description: X2 build failures
2738 Details : fix build failures on Cray X2.
2742 Description: xid & resent requests
2743 Details : Initialize RPC XID from clock at startup (randomly if clock is
2746 Severity : enhancement
2748 Description: Add lustre_start utility to start or stop multiple Lustre servers
2753 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2754 Details : In case of memory pressure, list_del() can be called twice on
2755 req->rq_history_list, causing a kernel oops.
2759 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2760 Details : fix stack overflow in the distributed lock manager by defering
2761 export eviction after a failed AST to the elt thread instead of
2762 handling it in the dlm interpret routine.
2766 Description: Convert some comments to new format.
2767 Details : Update documenting comments to match doxygen conventions.
2771 Description: Grammar fixes.
2772 Details : A couple of trivial spelling fixes.
2776 Description: OSD_COUNTERS-mandatory
2777 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2778 to simplify the code.
2782 Description: simplify cmm_device freeing logic.
2783 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2787 Description: Add lockdep support to dt_object_operations locking interface.
2788 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2789 parameter indicating lock ordering. Update mdd code to use new
2794 Description: Introduce failloc constants for lockless IO tests.
2795 Details : Add two new failloc constants to test lockless IO. Only one of
2796 them in implemented---another is checked in yet to be landed
2801 Description: Add lockdep support for inode mutex.
2802 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2803 in the situations where Lustre has to lock more than one inode
2808 Description: Add optional invariants checking support.
2809 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2810 configure switch. This macro is to be used for consistency and
2811 sanity checks that are too expensive to be left in `production'
2816 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2817 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2822 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2823 Details : Use spin_lock_nested() in (the only) situation where more than
2824 one ldlm_lock is locked simultaneously. Also, fix possible
2825 dead-lock in ldlm_lock_change_resource() by enforcing particular
2830 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2831 Details : Instead of specifying each ldlm_lock call-back through separate
2832 parameter, wrap them into struct ldlm_callback_suite.
2836 Description: Kill join_lru obd method and its callers.
2837 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2838 in memory---a requirement implemented through ->o_join_lru() obd
2839 method in HEAD. Remove this method and its users.
2843 Description: Add asynchronous ldlm ENQUEUE completion handler.
2844 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2845 case that stock ldlm_completion_ast() cannot handle as it waits
2846 until lock is granted. Introduce new ldlm_completion_ast_async()
2847 for this. Also comment ldlm_completion_ast().
2851 Description: ldlm_error <-> errno conversion.
2852 Details : Add functions to map (rather arbitrary) between LDLM error codes
2853 and standard errno values. CLIO needs this to prevent LDLM specific
2854 constants from escaping ldlm and osc.
2858 Description: Kill unused ldlm_handle2lock_ns() function.
2859 Details : Kill unused ldlm_handle2lock_ns() function.
2863 Description: Add lu_ref support to ldlm_lock
2864 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2865 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2866 and ldlm_resource. LDLM interface has to be changed, because code that
2867 releases a reference on a lock, has to "know" what reference this is.
2868 In the most frequent case
2870 lock = ldlm_handle2lock(handle);
2872 LDLM_LOCK_PUT(lock);
2874 no changes are required. When any other reference (received _not_ from
2875 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2876 instead of LDLM_LOCK_PUT().
2878 Arguably, changes are pervasive, and interface requires some discipline
2879 for proper use. On the other hand, it was very instrumental in finding
2880 a few leaked lock references.
2884 Description: Add ldlm_lock_addref_try().
2885 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2886 attempts to addref a lock that might be being canceled
2891 Description: Add ldlm_weigh_callback().
2892 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2893 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2894 hard-coded `number of pages' logic.
2898 Description: Add lockdep annotations to llog code.
2899 Details : Use appropriately tagged _nested() locking calls in the places
2900 where llog takes more than one ->lgh_lock lock.
2904 Description: Add loi_kms_set().
2905 Details : Wrap kms updates into a helper function.
2909 Description: Constify instances of struct lsm_operations.
2910 Details : Constify instances of struct lsm_operations.
2914 Description: lu_conf support.
2915 Details : On a server, a file system object is uniquely identified
2916 by a fid, which is sufficient to locate and load all object
2917 state (inode). On a client, on the other hand, more data are
2918 necessary instantiate an object. Change lu_object_find() and
2919 friends to take additional `lu_conf' argument describing object.
2920 Typically this includes layout information.
2924 Description: lu_context fixes.
2925 Details : Introduce new lu_context functions that are needed on the client
2926 side, where some system threads (ptlrpcd) are shared by multiple
2927 modules, and so cannot be stopped during module shutdown.
2931 Description: Add start and stop methods to lu_device_type_operations.
2932 Details : Introduce two new methods in lu_device_type_operations, that are
2933 invoked when first instance of a given type is created and last one
2934 is destroyed respectively. This is need by CLIO.
2938 Description: Add lu_ref support to struct lu_device.
2939 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2940 track leaked references.
2944 Description: Introduce lu_kmem_descr.
2945 Details : lu_kmem_descr and its companion interface allow to create
2946 and destroy a number of kmem caches at once.
2950 Description: Fix lu_object finalization race.
2951 Details : Fix a race between lu_object_find() finding an object and its
2952 concurrent finalization. This race is (most likely) not possible
2953 on the server, but might happen on the client.
2957 Description: Introduce lu_ref interface.
2958 Details : lu_ref is a debugging module allowing to track references to
2959 a given object. It is quite cpu expensive, and has to be
2960 explicitly enabled with --enable-lu_ref. See usage description
2965 Description: Factor lu_site procfs stats into a separate function.
2966 Details : Separate lu_site stats printing code into a separate function
2967 that can be reused on a client.
2971 Description: Constify instances of struct {lu,dt,md}_device_operations.
2972 Details : Constify instances of struct {lu,dt,md}_device_operations.
2976 Description: Introduce struct md_site and move meta-data specific parts of
2977 struct lu_site here.
2978 Details : Move md-specific fields out of struct lu_site into special struct
2979 md_site, so that lu_site can be used on a client.
2983 Description: Kill mdd_lov_destroy().
2984 Details : Remove unused mdd code.
2988 Description: Add st_block checking to multistat.c.
2989 Details : Add st_block checking to multistat.c.
2993 Description: Add lu_ref support to struct obd_device.
2994 Details : Add lu_ref tracking to obd_device.
2998 Description: Kill obd_set_fail_loc().
2999 Details : Remove unused code.
3003 Description: Add special type for ptlrpc_request interpret functions.
3004 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
3005 there. Actual usage will be in CLIO.
3009 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
3010 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
3011 doesn't work with lockdep.
3015 Description: Add rwv.c test program.
3016 Details : New testing program exercising readv(2) and writev(2) (Qian).
3020 Description: Add sendfile.c test program.
3021 Details : New testing program exercising sendfile(2) (Jay).
3025 Description: Ratelimit a message that can be very frequent.
3026 Details : Ratelimit a memory allocation failure message that can
3031 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3032 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3036 Description: Liblustre build fixes.
3037 Details : Liblustre build fixes.
3041 Description: libcfs: add cfs_{need,cond}_resched() interface.
3042 Details : libcfs: add cfs_{need,cond}_resched() definition and
3043 implementations for Linux, NT, and liblustre.
3045 Severity : enhancement
3047 Description: More exported tunables for mballoc
3048 Details : Add support for tunable preallocation window and new tunables for
3049 large/small requests
3053 Description: Detect corruption of block bitmap and checking for preallocations
3054 Details : Checks validity of on-disk block bitmap. Also it does better
3055 checking of number of applied preallocations. When corruption is
3056 found, it turns filesystem readonly to prevent further corruptions.
3060 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3061 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3062 window to 0. If it does hit cache miss, and read-pattern is still
3063 stride-io mode, does not reset the stride window, but also does not
3064 increase the stride window length in this case.
3068 Frequency : only for big-endian servers
3069 Description: Check if system is big-endian while mounting fs with extents feature
3070 Details : Mounting a filesystem with extents feature will fail on big-endian
3071 systems since ext3-based ldiskfs is not supported on big-endian
3072 systems. This can be over-riden with "bigendian_extents" mount option.
3074 Severity : enhancement
3076 Description: The root squash functionality
3077 Details : A security feature, which is to prevent users from being able
3078 to mount lustre on their desktop, run as root, and delete
3079 all of the files in the filesystem. The goal is accomplished by
3080 remapping user id (UID) and group id (GID) of the root user to
3081 a UID and GID specified by the system administartor via Lustre
3082 configuration management server (MGS). The functionality also
3083 allows to specify sets of clients for which the remapping does
3088 Description: Excessive recovery window
3089 Details : With AT enabled, the recovery window can be excessively long (6000+
3090 seconds). To address this problem, we no longer use
3091 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3092 timeout no longer depends on the service time, it is set to
3093 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3094 time via pb_service_time.
3098 Description: Watchdog triggered on MDS failover
3099 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3100 the OSTs know that the MDS "UUID" can be reused for the same export
3101 from a different NID, so we do not need to wait for the export to be
3105 Frequency : rare, only if using MMP with Linux RAID
3107 Description: MMP doesn't work with Linux RAID
3108 Details : While using HA for Lustre servers with Linux RAID, it is possible
3109 that MMP will not detect multiple mounts. To make this work we
3110 need to unplug the device queue in RAID when the MMP block is being
3111 written. Also while reading the MMP block, we should read it from
3112 disk and not the cached one.
3114 Severity : enhancement
3116 Description: open file using fid
3117 Details : A file can be opened using just its fid, like
3118 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3121 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3124 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3125 Details : add proper locking for creation of jbd_4k slab cache
3129 Description: MMP check in ext3_remount() fails without displaying any error
3130 Details : When multiple mount protection fails during remount, proper error
3133 Severity : enhancement
3135 Description: Allow stripe size to be up to 4G-64k
3136 Details : Fix math logic to allow large stripe sizes.
3140 Description: add check for >8TB ldiskfs filesystems
3141 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3142 Don't allow >8TB ldiskfs filesystems to be mounted without
3143 force_over_8tb mount option
3148 Descriptoin: OSS complains "too many blocks" and crashes
3149 Details : In filter_commitrw_write():
3150 1. lquota_chkquota();
3151 2. obdo_from_inode();
3152 3. lquota_pending_commit();
3153 "2." has changed uid and gid in this case, then caused
3154 lquota_chkquota() and lquota_pending_commit() to use
3155 different uid and gid. That is the root of the bug.
3159 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3160 Details : Detect and print out the 'stripe count = -1' case properly.
3164 Description: Out or order replies might be lost on replay
3165 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3166 request with tid smaller than one currently being added, add it
3167 to the start, not end of the list.
3169 --------------------------------------------------------------------------------
3171 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3173 * Support for kernels:
3174 2.6.5-7.283 (SLES 9),
3175 2.6.9-55.EL (RHEL 4),
3176 2.6.16.46-0.14 (SLES 10),
3177 2.6.18.8 vanilla (kernel.org)
3178 * Client support for unpatched kernels:
3179 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3180 2.6.16 - 2.6.22 vanilla (kernel.org)
3181 * Due to recently discovered recovery problems, we do not recommend
3182 using patchless RHEL 4 clients with this or any earlier release.
3183 * Recommended e2fsprogs version: 1.39.cfs8
3184 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3185 * Starting with this release, the ldiskfs backing filesystem required
3186 by Lustre is now in its own package, lustre-ldiskfs. This package
3187 should be installed. It is versioned separately from Lustre and
3188 may be released separately in future.
3192 Description: block reactivating mgc import until all deactivates complete
3193 Details : Fix race when failing back MDT/MGS to itself (testing)
3196 Frequency : at statup only
3198 Description: mds_lov_synchronize race leads to various problems
3199 Details : simultaneous MDT->OST connections at startup can cause the
3200 sync to abort, leaving the OSC in a bad state.
3202 Severity : enhancement
3204 Description: add optional extra BUILD_VERSION info
3205 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3206 to override the lustre version.
3209 Frequency : 2.6.18 servers only
3211 Description: ll_kern_mount() doesn't release the module reference
3212 Details : The ldiskfs module reference count never drops down to 0
3213 because ll_kern_mount() doesn't release the module reference.
3218 Description: server LBUG when using old ost_num_threads parameter
3219 Details : Accept the old ost_num_threads parameter but warn that it
3220 is deprecated, and fix an off-by-one error that caused an LBUG.
3225 Description: Transient SCSI error results in persistent IO issue
3226 Details : iobuf->dr_error is not reinitialized to 0 between two
3230 Frequency : sometimes when underlying device returns I/O errors
3232 Description: OSTs not going read-only during write failures
3233 Details : OSTs are not remounted read-only when the journal commit threads
3234 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3235 instead of the ext3 wrappers.
3239 Description: poor connect scaling with increasing client count
3240 Details : Don't run filter_grant_sanity_check for more than 100 exports
3241 to improve scaling for large numbers of clients.
3244 Frequency : SLES10 only
3246 Description: sanity-quota.sh quotacheck failed: rc = -22
3247 Details : Quotas cannot be enabled on SLES10.
3250 Frequency : liblustre clients only
3252 Description: getdirentries does not give error when run on compute nodes
3253 Details : getdirentries does not fail when the size specified as an argument
3254 is too small to contain at least one entry
3256 Severity : enhancement
3258 Description: Add LNET router traceability for debug purposes
3259 Details : If a checksum failure occurs with a router as part of the
3260 IO path, the NID of the last router that forwarded the bulk data
3261 is printed so it can be identified.
3266 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3267 Details : Due to a race condition, liblustre clients were occasionally
3268 evicted incorrectly.
3270 Severity : enhancement
3272 Description: lfs setstripe use optional parameters instead of postional
3275 Severity : enhancement
3277 Description: Nanosecond timestamp support for ldiskfs
3278 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3279 resolution timestamps. There is not yet support for this at
3280 the Lustre filesystem level.
3283 Frequency : during server recovery
3285 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3286 Details : request with rq_no_resend flag not awake l_wait_event if they get
3290 Frequency : nfs export on patchless client
3292 Description: connectathon hang when test nfs export over patchless client
3293 Details : Disconnected dentry cannot be found with lookup, so we do not need
3294 to unhash it or make it invalid
3297 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3298 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3299 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3304 Description: open req refcounting wrong on reconnect
3305 Details : If reconnect happened between getting open reply from server and
3306 call to mdc_set_replay_data in ll_file_open, we will schedule
3307 replay for unreferenced request that we are about to free.
3308 Subsequent close will crash in variety of ways.
3309 Check that request is still eligible for replay in
3310 mdc_set_replay_data().
3315 Description: disable writes to filesystem when reading health_check file
3316 Details : the default for reading the health_check proc file has changed
3317 to NOT do a journal transaction and write to disk, because this
3318 can cause reads of the /proc file to hang and block HA state
3319 checking on a healthy but otherwise heavily loaded system. It
3320 is possible to return to the previous behaviour during configure
3321 with --enable-health-write.
3323 Severity : enhancement
3325 Description: 64-bit inode version
3326 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3327 to the inode. This will be required for version-based recovery.
3332 Description: MDS fails to start if a duplicate client export is detected
3333 Details : in some rare cases it was possible for a client to connect to
3334 an MDS multiple times. Upon recovery the MDS would detect this
3335 and fail during startup. Handle this more gracefully.
3337 Severity : enhancement
3339 Description: Add -o localflock option to simulate old noflock
3341 Details : This will achieve local-only flock/fcntl locks
3347 Description: log_commit_thread vs filter_destroy race leads to crash
3348 Details : Take import reference before releasing llog record semaphore
3353 Description: Wrong request locking in request set processing
3354 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3355 imp_delayed_list, in this place should be used imp_lock.
3358 Frequency : when reconnection
3360 Description: Grant Leak when osc reconnect to OST
3361 Details : When osc reconnect ost, OST(filter) should check whether it
3362 should grant more space to client by comparing fed_grant and
3363 cl_avail_grant, and return the granted space to client instead
3364 of "new granted" space, because client will call osc_init_grant
3365 to update the client grant space info.
3368 Frequency : when client reconnect to OST
3370 Description: Grant Leak when osc do resend and replay bulk write
3371 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3372 bulk write request, because the grant info while be sync between
3373 OSC and OST when reconnect, and we should ignore the grant info
3374 these of resend/replay write req.
3379 Description: Grant space more than available left space sometimes.
3380 Details : When then OST is about to be full, if two bulk writing from
3381 different clients came to OST. Accord the avaliable space of the
3382 OST, the first req should be permitted, and the second one
3383 should be denied by ENOSPC. But if the seconde arrived before
3384 the first one is committed. The OST might wrongly permit second
3385 writing, which will cause grant space > available space.
3388 Frequency : when client is evicted
3390 Description: Grant might be wrongly erased when osc is evicted by OST
3391 Details : when the import is evicted by server, it will fork another
3392 thread ptlrpc_invalidate_import_thread to invalidate the
3393 import, where the grant will be set to 0. While the original
3394 thread will update the grant it got when connecting. So if
3395 the former happened latter, the grant will be wrongly errased
3396 because of this race.
3401 Description: Checking Stale with correct fid
3402 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3403 to check whether it is stale, because sometimes, we want the
3404 enqueue happened anyway, and op_data.fid2 will not be initialized.
3406 Severity : enhancement
3408 Description: update patchless client
3409 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3412 Frequency : only with 2.4 kernel
3414 Description: random memory corruption
3415 Details : size of struct ll_inode_info is to big for union inode.u and this
3416 can be cause of random memory corruption.
3421 Description: Memory leak in recovery
3422 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3423 It should also check obd_fail before fsfilt_start, otherwise if
3424 fsfilt_start return -EROFS,(failover mds during mds recovery).
3425 then the req will return with repmsg->transno = 0 and rc = EROFS.
3426 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3427 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3428 be freed no matter whether fsfilt_commit success or not.
3431 Frequency : only with huge count clients
3433 Description: Prevents from taking the superblock lock in llap_from_page for
3435 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3436 for llap_from_page call in ll_removepage prevents from taking the
3437 superblock lock for a soon died page.
3442 Description: Not check open intent error before release open handle
3443 Details : in some rare cases, the open intent error is not checked before
3444 release open handle, which may cause
3445 ASSERTION(open_req->rq_transno != 0), because it tries to release
3446 the failed open handle.
3451 Description: Set cat log bitmap only after create log success.
3452 Details : in some rare cases, the cat log bitmap is set too early. and it
3453 should be set only after create log success.
3457 Description: Accessing a block bevice can re-enable I/O when Lustre is
3458 tearing down a device.
3459 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3464 Description: service threads may hog cpus when there are a lot of requests
3465 Details : Insert cond_resched to give other threads a chance to use some CPU
3470 Description: the cat log was not initialized in recovery
3471 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3472 unlink log on mds will not be initialized until mds post
3473 recovery. And also in mds post recovery, the unlink log will
3474 initialization will be done asynchronausly, so there will be race
3475 between add unlink log and unlink log initialization.
3479 Description: brw_stats were being printed incorrectly
3480 Details : brw_stats were being printed as log2 but all of them were not
3481 recorded as log2. Also remove some code duplication arising from
3482 filter_tally_{read,write}.
3486 Frequency : rare, only in recovery.
3487 Description: ASSERTION(req->rq_type != LI_POISON) failed
3488 Details : imp_lock should be held while iterating over imp_sending_list for
3489 prevent destroy request after get timeout in ptlrpc_queue_wait.
3493 Description: replay-single.sh test 52 fails
3494 Details : A lock's skiplist need to be cleanup when it being unlinked
3495 from its resource list.
3499 Description: Short directio read returns full requested size rather than
3501 Details : Direct I/O operations should return actual amount of bytes
3502 transferred rather than requested size.
3504 Severity : enhancement
3506 Description: metadata RPC reduction (e.g. for rm performance)
3507 Details : decrease the amount of synchronous RPC between clients and servers
3508 by canceling conflicing lock before the operation on the client
3509 and packing thier handles into the main operation RPC to server.
3511 Severity : enhancement
3513 Description: Async OSC create to avoid the blocking unnecessarily.
3514 Details : If an OST has no remaining object, system will block on the
3515 creation when it needs to create a new object on this OST. Now,
3516 ways use pre-created objects when available, instead of blocking on
3517 an empty osc while others are not empty. If we must block, we
3518 block for the shortest possible period of time.
3522 Description: improve handling recoverable errors
3523 Details : if request processig with error which can be recoverable on server
3524 request should be resend, otherwise page released from cache and
3527 Severity : enhancement
3529 Description: refine locking for avoid write wrong info into lov_objid file
3530 Details : fix possible races with add new target and write/update data in
3533 --------------------------------------------------------------------------------
3535 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3540 Frequency : on some architectures
3542 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3543 Details : Uninitialized flags sometimes cause configuration commands to
3547 Frequency : patchless clients only
3549 Description: missing __iget() symbol export
3550 Details : The __iget() symbol export is missing. To avoid the need for
3551 this on patchless clients the deathrow inode reaper is turned
3552 off, and we depend on the VM to clean up old inodes. This
3553 dependency was during via the fix for bug 12181.
3555 --------------------------------------------------------------------------------
3557 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3559 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3560 INTEROPERATE with older versions automatically. In many cases a
3561 special upgrade step is needed. Please read the
3562 user documentation before upgrading any part of a live system.
3563 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3564 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3565 * WARNING: Lustre configuration and startup changes are required with
3566 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3570 Severity : enhancement
3572 Description: MountConf
3573 Details : Lustre configuration is now managed via mkfs and mount
3574 commands instead of lmc and lconf. New obd types (MGS, MGC)
3575 are added for dynamic configuration management. See
3576 https://mail.clusterfs.com/wikis/lustre/MountConf for
3579 Severity : enhancement
3581 Description: dynamic OST addition
3582 Details : OSTs can now be added to a live filesystem
3584 Severity : enhancement
3586 Description: startup order invariance
3587 Details : MDTs and OSTs can be started in any order. Clients only
3588 require the MDT to complete startup.
3590 Severity : enhancement
3592 Description: parallel, asynchronous orphan cleanup
3593 Details : orphan cleanup is now performed in separate threads for each
3594 OST, allowing parallel non-blocking operation.
3596 Severity : enhancement
3598 Description: optimized stripe assignment
3599 Details : stripe assignments are now made based on ost space available,
3600 ost previous usage, and OSS previous usage, in order to try
3601 to optimize storage space and networking resources.
3603 Severity : enhancement
3605 Description: Permanently set tunables
3606 Details : All writable /proc/fs/lustre tunables can now be permanently
3607 set on a per-server basis, at mkfs time or on a live system.
3609 Severity : enhancement
3611 Description: Lustre message v2
3612 Details : Add lustre message format v2.
3614 Severity : enhancement
3616 Description: client OST exclusion list
3617 Details : Clients can be started with a list of OSTs that should be
3618 declared "inactive" for known non-responsive OSTs.
3622 Description: SPEC SFS validation failure on NFS v2 over lustre.
3623 Details : Changes the blocksize for regular files to be 2x RPC size,
3624 and not depend on stripe size.
3626 Severity : enhancement
3628 Description: Multiple MD RPCs in flight.
3629 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3630 To avoid overly-overloading MDS, introduce a limit on number of
3631 MDS RPCs in flight for a single client and add /proc controls
3632 to adjust this limit.
3634 Severity : enhancement
3636 Description: client read/write statistics
3637 Details : Add client read/write call usage stats for performance
3638 analysis of user processes.
3639 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3640 file access. extents_stats shows chunk size distribution.
3641 extents_stats_per_process show chunk size distribution per
3644 Severity : enhancement
3646 Description: mds statistics
3647 Details : Add detailed mds operations statistics in
3648 /proc/fs/lustre/mds/*/stats.
3652 Description: Failure of copying files with lustre special EAs.
3653 Details : Client side always return success for setxattr call for lustre
3654 special xattr (currently only "trusted.lov").
3659 Description: Refcount LNET uuids
3660 Details : The global LNET uuid list grew linearly with every startup;
3661 refcount repeated list entries instead of always adding to
3664 Severity : enhancement
3666 Description: Dynamic service threads
3667 Details : Within a small range, start extra service threads
3668 automatically when the request queue builds up.
3671 Frequency : mixed-endian client/server environments
3673 Description: mixed-endian crashes
3674 Details : The new msg_v2 system had some failures in mixed-endian
3677 Severity : enhancement
3679 Description: Easy OST removal
3680 Details : OSTs can be permanently deactivated with e.g. 'lctl
3681 conf_param lustre-OST0001.osc.active=0'
3683 Severity : enhancement
3685 Description: MGS proc entries
3686 Details : Added basic proc entries for the MGS showing what filesystems
3689 Severity : enhancement
3691 Description: provide MGS failover
3692 Details : Added config lock reacquisition after MGS server failover.
3694 Severity : enhancement
3696 Description: add Linux 2.4 support
3697 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3701 Description: a large application tries to do I/O to the same resource and dies
3702 in the middle of it.
3703 Details : Check the req->rq_arrival time after the call to
3704 ost_brw_lock_get(), but before we do anything about
3705 processing it & sending the BULK transfer request. This
3706 should help move old stale pending locks off the queue as
3707 quickly as obd_timeout.
3710 Frequency : when an incorrect nid is specified during startup
3712 Description: ptlrpc connect to non-existant node causes kernel crash
3713 Details : LNET can't be re-entered from an event callback, which
3714 happened when we expire a message after the export has been
3715 cleaned up. Instead, hand the zombie cleanup off to another
3718 Severity : enhancement
3720 Description: plain/inodebits lock performance improvement
3721 Details : Grouping plain/inodebits in granted list by their request modes
3722 and bits policy, thus improving the performance of search through
3726 Frequency : only if OST filesystem is corrupted
3728 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3729 Details : for a short time RPCs with bulk IO are in the replay list,
3730 but replay of bulk IOs is unimplemented. If the OST filesystem
3731 is corrupted due to disk cache incoherency and then replay is
3732 started it is possible to trip an assertion. Avoid putting
3733 committed RPCs into the replay list at all to avoid this issue.
3736 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3738 Description: System hang on startup
3739 Details : This bug allowed the liblustre (e.g. catamount) client to
3740 return to the app before handling all startup RPCs. This
3741 could leave the node unresponsive to lustre network traffic
3742 and manifested as a server ptllnd timeout.
3744 Severity : enhancement
3746 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3747 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3748 boolean to control whether to print peer debug info when a
3749 client's RPC times out.
3752 Frequency : only for kernels with patches from Lustre below 1.4.3
3754 Description: Remove old rdonly API
3755 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3758 Frequency : only for devices with external journals
3760 Description: Set external device read-only also
3761 Details : During a commanded failover stop, we set the disk device
3762 read-only while the server shuts down. We now also set any
3763 external journal device read-only at the same time.
3766 Frequency : when upgrading from 1.4 while trying to change parameters
3768 Description: The wrong (new) MDC name was used when setting parameters for
3769 upgraded MDT's. Also allows changing of OSC (and MDC)
3770 parameters if --writeconf is specified at tunefs upgrade time.
3773 Frequency : when setting specific ost indicies
3775 Description: QOS code breaks on skipped indicies
3776 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3777 created with --index need not be sequential.
3779 Severity : enhancement
3781 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3782 Details : The uninit_groups feature works in conjunction with the kernel
3783 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3784 up the pass1 processing of e2fsck. This is a read-only feature
3785 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3786 mount filesystems that have had this feature enabled.
3788 Severity : enhancement
3790 Description: Improve multi-block allocation algorithm to avoid fragmentation
3791 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3792 allocation locality and avoid filesystem fragmentation.
3794 ------------------------------------------------------------------------------
3796 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3798 * Support for kernels:
3799 2.4.21-47.0.1.EL (RHEL 3)
3800 2.6.5-7.283 (SLES 9)
3801 2.6.9-42.0.10.EL (RHEL 4)
3802 2.6.12.6 vanilla (kernel.org)
3803 2.6.16.27-0.9 (SLES 10)
3804 * Recommended e2fsprogs version: 1.39.cfs5
3806 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3810 Frequency : occasional, depends on client load and configuration
3811 Bugzilla : 12181, 12203
3812 Description: data loss for recently-modified files
3814 Details : In some cases it is possible that recently written or created
3815 files may not be written to disk in a timely manner (this should
3816 normally be within 30s unless client IO load is very high).
3817 The problem appears as zero-length files or files that are a
3818 multiple of 1MB in size after a client crash or client eviction
3819 that are missing data at the end of the file.
3821 This problem is more likely to be hit on clients where files are
3822 repeatedly created and unlinked in the same directory, clients
3823 have a large amount of RAM, have many CPUs, the filesystem has
3824 many OSTs, the clients are rebooted frequently, and/or the files
3825 are not accessed by other nodes after being written.
3827 The presence of the problem can be detected by looking at
3828 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3829 smaller than the second (nr_unused) then dirty files will not
3830 be flushed automatically to disk. "sync; sleep 10" should be
3831 run several times on the node before unmounting it to update
3832 Lustre (this is also safe to run on nodes without this problem).
3834 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3835 that can cause this same problem, so customers using that kernel
3836 also need to update the kernel in addition to Lustre. In order
3837 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3839 It is normal that files written just before a client crash (less
3840 than 30s) may not yet have been flushed to disk, even for local
3844 Frequency : frequent on thin XT3 nodes
3846 Description: UUID collision on thin XT3 Linux nodes
3847 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3848 randomly, since we relied on an insufficiently-seeded PRNG.
3853 Description: OSS hangs after "All ost request buffers busy"
3854 Details : A deadlock between quota and journal operations caused OSS
3855 hangs after printing "All ost request buffers busy."
3858 Frequency : always on liblustre builds
3860 Description: Cleanup compiler warnings on liblustre
3863 Frequency : always on liblustre builds on XT3
3865 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3869 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3870 > 32000 subdirectories
3871 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3872 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3873 subdirectory count crosses 32000. This will aid e2fsck to
3874 correctly handle more than 32000 subdirectories.
3877 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3879 Description: System hang on startup
3880 Details : This bug allowed the liblustre (e.g. catamount) client to
3881 return to the app before handling all startup RPCs. This
3882 could leave the node unresponsive to lustre network traffic
3883 and manifested as a server ptllnd timeout.
3885 Severity : enhancement
3887 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3888 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3889 boolean to control whether to print peer debug info when a
3890 client's RPC times out.
3895 Description: make O_SYNC working on 2.6 kernels
3896 Details : 2.6 kernels use different method for mark pages for write,
3897 so need add a code to lustre for O_SYNC work.
3902 Description: Failure to close file and release space on NFS
3903 Details : Put inode details into lock acquired in ll_intent_file_open.
3904 Use mdc_intent_lock in ll_intent_open to properly
3905 detect all kind of errors unhandled by mdc_enqueue.
3910 Description: proc file read during shutdown sometimes raced obd removal,
3912 Details : Add lock to prevent obd access after proc file removal.
3915 Frequency : Only for files larger than 4GB on 32-bit clients.
3917 Description: improperly doing page alignment of locks
3918 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3919 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3924 Description: RPCs being resent when they shouldn't be
3925 Details : Some RPCs that should not be resent are being resent. This
3926 can cause inconsistencies in the RPC state machine. Do not
3927 resend such requests.
3930 Frequency : rare, only with NFS export
3932 Description: Crash on NFS re-export node
3933 Details : under very unusual load conditions an assertion is hit in
3934 ll_intent_file_open()
3937 Frequency : only if OST filesystem is corrupted
3939 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3940 Details : for a short time RPCs with bulk IO are in the replay list,
3941 but replay of bulk IOs is unimplemented. If the OST filesystem
3942 is corrupted due to disk cache incoherency and then replay is
3943 started it is possible to trip an assertion. Avoid putting
3944 committed RPCs into the replay list at all to avoid this issue.
3949 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3950 Details : Large single O_DIRECT read and write calls can fail to allocate
3951 a sufficiently large buffer to process the request. In case of
3952 allocation failure the allocation is retried with a smaller
3953 buffer and broken into smaller requests.
3955 Severity : enhancement
3957 Description: Add -o localflock option to simulate old noflock behaviour.
3958 Details : This will achieve local-only flock/fcntl locks coherentness.
3963 Description: versioning check is incomplete
3964 Details : Checking the version difference of client vs. server, report
3965 error if the gap is too big.
3970 Description: add support PG_writeback bit
3971 Details : add support for PG_writeback bit for Lustre, for more carefull
3972 work with page cache in 2.6 kernel. This also fix some deadlocks
3973 and remove hack for work O_SYNC with 2.6 kernel.
3975 Severity : enhancement
3977 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3978 Details : The uninit_groups feature works in conjunction with the kernel
3979 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3980 up the pass1 processing of e2fsck. This is a read-only feature
3981 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3982 mount filesystems that have had this feature enabled.
3984 Severity : enhancement
3986 Description: Improve multi-block allocation algorithm to avoid fragmentation
3987 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3988 allocation locality and avoid filesystem fragmentation.
3990 ------------------------------------------------------------------------------
3992 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3994 * Support for kernels:
3995 2.6.9-42.0.3.EL (RHEL 4)
3996 2.6.5-7.276 (SLES 9)
3997 2.4.21-47.0.1.EL (RHEL 3)
3998 2.6.12.6 vanilla (kernel.org)
3999 2.6.16.21-0.8 (SLES10)
4000 * Recommended e2fsprogs version: 1.39.cfs2-0
4002 * The backwards-compatible /proc/sys/portals symlink has been removed
4003 in this release. Before upgrading, please ensure that you change
4004 any configuration scripts or /etc/sysctl.conf files that access
4005 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4006 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4007 in advance of the upgrade on any system running Lustre 1.4.6 or
4008 newer, since /proc/sys/lnet was added in that version.
4009 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4013 Frequency : only when quota is used
4015 Description: avoid scanning export list for quota master
4016 Details : Change the algorithms to avoid scanning export list in order
4017 to improve the efficiency.
4020 Frequency : MDS failover only, very rarely
4022 Description: "went back in time" messages on mds failover
4023 Details : The greatest transno may be lost when the current operation
4024 finishes with an error (transno==0) and the client's last_rcvd
4025 record is over-written. Save the greatest transno in the
4026 mds_last_transno for this case.
4029 Frequency : always for specific kernels and striping counts
4031 Description: client may get "Matching packet too big" without ACL support
4032 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4033 when trying to access files in certain configurations. The
4034 clients should in fact be denied when mounting because they do
4035 not understand ACLs.
4038 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4040 Description: many clients connecting with IO in progress causes connect timeouts
4041 Details : Avoid synchronous journal commits to avoid delays caused by many
4042 clients connecting/disconnecting when bulk IO is in progress.
4043 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4044 OST_IO_PORTAL to avoid delays behind potentially many pending
4048 Frequency : occasionally with multiple writers to a single file
4050 Description: shared writes to file may result in wrong size reported by stat()
4051 Details : Allow growing of kms when extent lock is cancelled
4054 Frequency : always with random mmap IO to multi-striped file
4056 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4057 Details : Only if the hole is at the end of OST object so that kms is too
4058 small. Fix is to increase kms accordingly in ll_nopage.
4061 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4063 Description: writes to a missing object would leak memory on the OST
4064 Details : If there is an inconsistency between the MDS and OST filesystems,
4065 such that the MDS references an object that doesn't exist, writes
4066 to that object will leak memory due to incorrect cleanup in the
4067 error handling path, eventually running out of memory on the OST.
4072 Description: Creating too long symlink causes lustre errors
4073 Details : Check symlink and name lengths before sending requests to MDS.
4076 Frequency : only if flock is enabled (not on by default)
4078 Description: posix locks not released on fd closure on 2.6.9+
4079 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4080 kernels, this caused such locks not to be released on fd close and
4081 then assertions on fs unmount about still used locks.
4084 Frequency : MDS failover only, very rarely
4086 Description: clients may get ASSERTION(granted_lock != NULL)
4087 Details : When request was taking a long time, and a client was resending
4088 a getattr by name lock request. The were multiple lock requests
4089 with the same client lock handle and
4090 mds_getattr_name->fixup_handle_for_resent_request found one of the
4091 lock handles but later failed with ASSERTION(granted_lock != NULL).
4096 Description: handle->h_buffer_credits > 0, assertion failure
4097 Details : h_buffer_credits is zero after truncate, causing assertion
4098 failure. This patch extends the transaction or creates a new
4102 Frequency : NFS re-export or patchless client
4103 Bugzilla : 11179, 10796
4104 Description: Crash on NFS re-export node (__d_move)
4105 Details : We do not want to hash the dentry if we don't have a lock.
4106 But if this dentry is later used in d_move, we'd hit uninitialised
4107 list head d_hash, so we just do this to init d_hash field but
4108 leave dentry unhashed.
4111 Frequency : NFS re-export or patchless client
4113 Description: NFS exports has problem with symbolic link
4114 Details : lustre client didn't properly install dentry when re-exported
4115 to NFS or running patchless client.
4118 Frequency : NFS re-export or patchless client
4120 Description: Various nfs/patchless fixes.
4121 Details : fixes reuse disconected alias for lookup process - this fixes
4122 warning "find_exported_dentry: npd != pd",
4123 fix permission error with open files at nfs.
4124 fix apply umask when do revalidate.
4127 Frequency : occasional
4129 Description: Crash on NFS re-export node
4130 Details : calling clear_page() on the wrong pointer triggered oops in
4131 generic_mapping_read().
4134 Frequency : rarely, using O_DIRECT IO
4136 Description: unaligned directio crashes client with LASSERT
4137 Details : check for unaligned buffers before trying any requests.
4140 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4142 Description: stale data returned from RAID cache
4143 Details : If only a small amount of IO is done to the RAID device before
4144 reading it again it is possible to get stale data from the RAID
4145 cache instead of reading it from disk.
4148 Frequency : always for sles10 kernel
4150 Description: sles10 support
4151 Details : ll_follow_link: compile fixes and using of nd_set_link
4152 under newer kernels.
4155 Frequency : depends on arch, kernel and compiler version, always on sles10
4158 Description: recursive or deep enough symlinks cause stack overflow
4159 Details : getting rid of large stack-allocated variable in
4163 Frequency : depends on hardware
4165 Description: lustre write performance loss in the SLES10 kernel
4166 Details : the performance loss is caused by using of write barriers in the
4167 ext3 code. The SLES10 kernel turns barrier support on by
4168 default. The fix is to undo that change for ldiskfs.
4170 ------------------------------------------------------------------------------
4172 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4174 * Support for kernels:
4175 2.6.9-42.0.3EL (RHEL 4)
4176 2.6.5-7.276 (SLES 9)
4177 2.4.21-47.0.1.EL (RHEL 3)
4178 2.6.12.6 vanilla (kernel.org)
4182 Frequency : quota enabled and large files being deleted
4184 Description: releasing more than 4GB of quota at once hangs OST
4185 Details : If a user deletes more than 4GB of files on a single OST it
4186 will cause the OST to spin in an infinite loop. Release
4187 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4192 Description: statfs data retrieved from /proc may be stale or zero
4193 Details : When reading per-device statfs data from /proc, in the
4194 {kbytes,files}_{total,free,avail} files, it may appear
4195 as zero or be out of date.
4198 Frequency : systems with MD RAID1 external journal devices
4200 Description: lconf's call to blkid is confused by RAID1 journal devices
4201 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4202 of returning all block devices that match the journal UUID.
4205 Frequency : always, for aggregate stripe size over 4GB
4207 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4208 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4209 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4212 Frequency : always if "lfs find" used on a local file/directory
4214 Description: "lfs find" segfaults if used on a local file/directory
4215 Details : The case where a directory component was not specified wasn't
4216 handled correctly. Handle this properly.
4219 Frequency : always on ppc64
4221 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4222 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4223 perform endianness conversion.
4226 Frequency : rarely (truncate to non-zero file size after write under load)
4227 Bugzilla : 10730, 10687
4228 Description: Files padded with zeros to next 4K multiple
4229 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4230 it is possible that files that are truncated to a non-zero size
4231 immediately after being written are filled with zero bytes beyond
4232 the truncated size. No file data is lost.
4234 Severity : enhancement
4236 Description: Allow recovery/failover for liblustre clients.
4237 Details : liblustre clients were unaware of failover configurations until
4240 Severity : enhancement
4242 Description: user file locks should fail when not mounting with flock option
4243 Details : Set up an error-returning stub in ll_file_operations.lock field
4244 to prevent incorrect behaviour when client is mounted without
4245 flock option. Also, set up properly f_op->flock field for
4249 Frequency : always on ia64
4251 Description: "lfs df" loops on printing out MDS statfs information
4252 Details : The obd_ioctl_data was not initialized and in some systems
4253 this caused a failure during the ioctl that did not return
4254 an error. Initialize the struct and return an error on failure.
4257 Frequency : SLES 9 only
4259 Description: Error of copying files with lustre special EAs as root
4260 Details : Client side always return success for setxattr call for lustre
4261 special xattr (currently only "trusted.lov").
4264 Frequency : rarely on clusters with both ia64+i386 clients
4266 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4267 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4268 clients are doing concurrent writes to the same file it is
4269 possible that the ia64 clients may LASSERT because the OST
4270 extent locks are not PAGE_SIZE aligned. Ensure that grown
4271 locks are always aligned on the request boundary.
4274 Frequency : specific use, occasional
4276 Description: Overwriting in use executable truncates on-disk binary image
4277 Details : If one node attempts to overwrite an executable in use by
4278 another node, we now correctly return ETXTBSY instead of
4279 truncating the file.
4281 Severity : enhancement
4283 Description: Async OSC create to avoid the blocking unnecessarily.
4284 Details : If an OST has no remaining object, system will block on the creating
4285 when need to create a new object on this OST. Now, ways use
4286 pre-created objects when available, instead of blocking on an
4287 empty osc while others are not empty. If we must block, we block
4288 for the shortest possible period of time.
4293 Description: chmod on Lustre root is propagated to other clients
4294 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4295 invalid by the follow_mount time.
4298 Frequency : liblustre clients only
4300 Description: Race in 'instant cancel' lock handling could lead to such locks
4301 never to be granted in case of SMP MDS
4302 Details : Do not destroy not yet granted but cbpending locks in
4306 Frequency : replay/resend of open
4308 Description: non null lock assetion failure in mds_intent_policy
4309 Details : Trying to replay/resend lockless open requests resulted in
4310 mds_open() returning 0 with no lock. Now it sets a flag if
4311 a lock is going to be returned.
4313 Severity : enhancement
4315 Description: Checksum enhancements
4316 Details : New checksum enhancements allow for resending RPCs that failed
4319 Severity : enhancement
4321 Description: Tunables on number of dirty pages in cacche
4322 Details : Allow to set limit on number of dirty pages cached.
4327 Description: client crash on unmount - lock still has references
4328 Details : In some error handling cases it was possible to leak a lock
4329 reference on a client while accessing a file. This was not
4330 harmful to the client during operation, but would cause the
4331 client to crash when the filesystem is unmounted.
4334 Frequency : specific case, rare
4336 Description: ETXTBSY on mds though file not in use by client
4337 Details : ETXTBSY is no longer incorrectly returned when attempting to
4338 chmod or chown a directory that the user previously tried to
4339 execute or a currently-executing binary.
4342 Frequency : extremely rare except on liblustre-based clients
4344 Description: Lustre space not freed when files are deleted
4345 Details : Clean up open-unlinked files after client eviction. Previously
4346 the unlink was skipped and the files remained as orphans.
4351 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4352 Details : In some cases it was possible to send a blocking callback to
4353 a client doing a glimpse, even though that client didn't get
4354 a lock granted. When the glimpse lock is cancelled on the OST
4355 the freed lock is left on the waiting list and corrupted the list.
4358 Frequency : all core dumps
4360 Description: Broke core dumps to lustre
4361 Details : Negative dentry may be unhashed if parent does not have UPDATE
4362 lock, but some callers, e.g. do_coredump, expect dentry to be
4363 hashed after successful create, hash it in ll_create_it.
4365 ------------------------------------------------------------------------------
4367 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4369 * Support for kernels:
4370 2.6.9-42.0.2.EL (RHEL 4)
4371 2.6.5-7.276 (SLES 9)
4372 2.4.21-40.EL (RHEL 3)
4373 2.6.12.6 vanilla (kernel.org)
4377 Frequency : always on RHEL 3
4379 Description: Number of open files grows over time
4380 Details : The number of open files grows over time, whether or not
4381 Lustre is started. This was due to a filp leak introduced
4382 by one of our kernel patches.
4384 ------------------------------------------------------------------------------
4386 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4388 * Support for kernels:
4389 2.6.9-42.EL (RHEL 4)
4390 2.6.5-7.276 (SLES 9)
4391 2.4.21-40.EL (RHEL 3)
4392 2.6.12.6 vanilla (kernel.org)
4397 Bugzilla : 5719, 9635, 9792, 9684
4398 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4399 Details : If a server is under heavy load and cannot reply to new
4400 connection requests before the client resends the (re)connect,
4401 the connection handling code can behave badly if two service
4402 threads are concurrently handing separate (re)connections from
4403 the same client. Add better locking to the connection handling
4404 code, and ensure that only a single connection will be processed
4405 for a given client UUID, even if the lock is dropped.
4407 Severity : enhancement
4409 Description: add TCP zero-copy support to kernel
4410 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4411 sends if the hardware supports scatter-gather and checksumming.
4412 This allows socklnd to do client-write and server-read more
4413 efficiently and reduce CPU utilization from skbuf copying.
4416 Frequency : only if NFS exporting from client
4418 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4419 Details : If setting attributes on a file created under NFS that had
4420 never been opened it would be possible to oops the client
4421 if the file had no objects.
4424 Frequency : always for liblustre
4426 Description: liblustre client does MDS+OSTs setattr RPC for each write
4427 Details : When doing a write from a liblustre client, the client
4428 incorrectly issued an RPC to the MDS and each OST the file was
4429 striped over in order to update the timestamps. When writing
4430 with small chunks and many clients this could overwhelm the MDS
4431 with RPCs. In all cases it would slow down the write because
4432 these RPCs are unnecessary.
4434 Severity : enhancement
4436 Description: allow number of MDS service threads to be changed at module load
4437 Details : It is now possible to change the number of MDS service threads
4438 running. Adding "options mds mds_num_threads={N}" to the MDS's
4439 /etc/modprobe.conf will set the number of threads for the next
4440 time Lustre is restarted (assuming the "mds" module is also
4441 reloaded at that time). The default number of threads will
4442 stay the same, 32 for most systems.
4447 Description: OST crash if filesystem is unformatted or corrupt
4448 Details : If an OST is started on a device that has never been formatted
4449 or if the filesystem is corrupt and cannot even mount then the
4450 error handling cleanup routines would dereference a NULL pointer.
4455 Description: NULL pointer deref in llap_from_page.
4456 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4457 incorrect) mapping if the page was truncated/reclaimed while it was
4458 searched for. Check for this condition and skip such pages when
4459 doing readahead. Introduce extra check to llap_from_page() to
4460 verify page->mapping->host is non-NULL (so page is not anonymous).
4463 Frequency : Sometimes when using sys_sendfile
4465 Description: "page not covered by a lock" warnings from ll_readpage
4466 Details : sendfile called ll_readpage without right page locks present.
4467 Now we introduced ll_file_sendfile that does necessary locking
4468 around call to generic_file_sendfile() much like we do in
4472 Frequency : with certain MDS communication failures at client mount time
4474 Description: NULL pointer deref after failed client mount
4475 Details : a client connection request may delayed by the network layer
4476 and not be sent until after the PTLRPC layer has timed out the
4477 request. If the client fails the mount immediately it will try
4478 to clean up before the network times out the request. Add a
4479 reference from the request import to the obd device and delay
4480 the cleanup until the network drops the request.
4483 Frequency : occasionally during client (re)connect
4485 Description: assertion failure during client (re)connect
4486 Details : processing a client connection request may be delayed by the
4487 client or server longer than the client connect timeout. This
4488 causes the client to resend the connection request. If the
4489 original connection request is replied in this interval, the
4490 client may trip an assertion failure in ptlrpc_connect_interpret()
4491 which thought it would be the only running connect process.
4494 Frequency : only with obd_echo servers and clients that are rebooted
4496 Description: kernel BUG accessing uninitialized data structure
4497 Details : When running an obd_echo server it did not start the ping_evictor
4498 thread, and when a client was evicted an uninitialized data
4499 structure was accessed. Start the ping_evictor in the RPC
4500 service startup instead of the OBD startup.
4502 Severity : enhancement
4503 Bugzilla : 10193 (patchless)
4504 Description: Remove dependency on various unexported kernel interfaces.
4505 Details : No longer need reparent_to_init, exit_mm, exit_files,
4506 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4509 Frequency : rare (only users of deprecated and unsupported LDAP config)
4511 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4512 Details : LDAP apparently contains 'lustreName' attributes instead of
4513 'name'. A simple remapping of the name is sufficient.
4516 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4518 Description: waiting_locks_callback trips kernel BUG if client is evicted
4519 Details : Running with the dump_on_timeout debug flag turned on makes
4520 it possible that the waiting_locks_callback() can try to dump
4521 the Lustre kernel debug logs from an interrupt handler. Defer
4522 this log dumping to the expired_lock_main() thread.
4524 Severity : enhancement
4526 Description: Support NFS exporting on 2.6 kernels.
4527 Details : Implement non-rawops metadata methods for NFS server to use without
4528 changing NFS server code.
4531 Frequency : very rare (synthetic metadata workload only)
4533 Description: two racing renames might cause an MDS thread to deadlock
4534 Details : Running the "racer" program may cause one MDS thread to rename
4535 a file from being the source of a rename to being the target of
4536 a rename at exactly the same time that another thread is doing
4537 so, and the second thread has already enqueued these locks after
4538 doing a lookup of the target and is trying to relock them in
4539 order. Ensure that we don't try to re-lock the same resource.
4542 Frequency : only very large systems with liblustre clients
4544 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4545 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4546 all OSTs in parallel. This allows the eviction of stale liblustre
4547 clients to proceed much faster than if they were done in series,
4548 and also offers similar improvements for other set_info RPCs.
4553 Description: excessive CPU usage during initial read phase on client
4554 Details : During the initial read phase on a client, it would agressively
4555 retry readahead on the file, consuming too much CPU and impacting
4556 performance (since 1.4.5.8). Improve the readahead algorithm
4557 to avoid this, and also improve some other common cases (read
4558 of small files in particular, where "small" is files smaller than
4559 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4564 Description: MDS crash when receiving packet with unknown intent.
4565 Details : Do not LBUG in unknown intent case, just return -EFAULT
4567 Severity : enhancement
4568 Bugzilla : 9293, 9385
4569 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4570 Details : Do not serialize getattr (non-intent version) and statfs.
4573 Frequency : occasional, when OST network is overloaded/intermittent
4575 Description: client evicted by OST after bulk IO timeout
4576 Details : If a client sends a bulk IO request (read or write) the OST
4577 may evict the client if it is unresposive to its data GET/PUT
4578 request. This is incorrect if the network is overloaded (takes
4579 too long to transfer the RPC data) or dropped the OST GET/PUT
4580 request. There is no need to evict the client at all, since
4581 the pinger and/or lock callbacks will handle this, and the
4582 client can restart the bulk request.
4585 Frequency : Always when mmapping file with no objects
4587 Description: client crashes when mmapping file with no objects
4588 Details : Check that we actually have objects in a file before doing any
4589 operations on objects in ll_vm_open, ll_vm_close and
4595 Description: Request leak when working with deleted CWD
4596 Details : Introduce advanced request refcount tracking for requests
4597 referenced from lustre intent.
4599 Severity : Enhancement
4601 Description: Cache open file handles on client.
4602 Details : MDS now will return special lock along with openhandle, if
4603 requested and client is allowed to hold openhandle, even if unused,
4604 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4605 closing files for every read/write openration.
4607 Severity : Enhancement
4609 Description: Cache open negative dentries on client when possible.
4610 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4611 negative dentries on lock revocation.
4616 Description: Remounting a client read-only wasn't possible with a zconf mount
4617 Details : It wasn't possible to remount a client read-only with llmount.
4619 Severity : enhancement
4620 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4621 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4622 MPICH is included to add Lustre-specific ADIO interfaces.
4623 This is based closely on the UFS ADIO layer and only differs
4624 in file creation, in order to allow the OST striping to be set.
4625 This is user-contributed code and not supported by CFS.
4630 Description: extended inode attributes (immutable, append-only) work improperly
4631 when 2.4 and 2.6 kernels are used on client/server or vice versa
4632 Details : Introduce kernel-independent values for these flags.
4634 Severity : enhancement
4637 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4638 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4639 granularity. Now, Fractional MB granularity is be supported,
4640 this is very useful for low memory system.
4642 Severity : enhancement
4644 Description: Getattr by fid
4645 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4646 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4647 have locks and avoids getattr rpc requests for it on every stat.
4650 Frequency : Always, for filesystems larger than 2TB
4652 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4653 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4654 filesystems larger than 2TB to be mounted with mballoc enabled.
4657 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4659 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4660 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4661 filesystems larger than 2TB is mounted then the kernel will
4662 silently corrupt the start of the filesystem. CONFIG_LBD is
4663 enabled for all CFS-supported kernels, but the possibility of
4664 this happening with a modified kernel config exists.
4666 Severity : enhancement
4668 Description: add client O_DIRECT support for 2.6 kernels
4669 Details : It is now possible to do O_DIRECT reads and writes to files
4670 in the Lustre client mountpoint on 2.6 kernel clients.
4672 Severity : enhancement
4674 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4675 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4676 parallel, not waiting for response from every OST before sending
4677 a rpc to the next OST.
4682 Description: setattr vs write race when updating file timestamps
4683 Details : Client processes that update a file timestamp into the past
4684 right after writing to the file (e.g. tar) it is possible that
4685 the updated file modification time can be reset to the current
4686 time due to a race between processing the setattr and write RPC.
4688 Severity : enhancement
4690 Description: Bring 'lfs find' closer in line with regular Linux find.
4691 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4692 -print0 options and obtains all the needed info through the lustre
4695 Severity : enhancement
4697 Description: support up to 1024 configured devices on one node
4698 Details : change obd_dev array from statically allocated to dynamically
4699 allocated structs as they are first used to reduce memory usage
4704 Description: Flush dirty partially truncated pages during truncate
4705 Details : Immediatelly flush partially truncated pages in filter_setattr,
4706 this way we completely avoid having any pages in page cache on OST
4707 and can retire ugly workarounds during writes to flush such pages.
4712 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4713 Details : protect inode from truncation within vfs_unlink() context
4714 just take a reference before calling vfs_unlink() and release it
4715 when parent's i_sem is free.
4720 Description: last_id value checked outside lock on OST caused LASSERT failure
4721 Details : If there were multiple MDS->OST object precreate requests in
4722 flight, it was possible that the OST's last object id was checked
4723 outside a lock and incorrectly tripped an assertion. Move checks
4724 inside locks, and discard old precreate requests.
4727 Frequency : always, if extents are used on OSTs
4729 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4730 Details : OSTs using the extents format would not zero the high 16 bits of
4731 the index physical block number. This is not a problem for any
4732 OST filesystems smaller than 16TB, and no kernels support ext3
4733 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4734 new/modified files) and can be fixed for existing filesystems
4735 with e2fsprogs-1.39-cfs1.
4740 Description: import connection selection may be incorrect if timer wraps
4741 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4742 import connections to be ignored if the 32-bit jiffies counter
4743 wraps. Use a 64-bit jiffies counter.
4746 Frequency : very large clusters immediately after boot
4748 Description: LNET request buffers exhausted under heavy short-term load
4749 Details : If a large number of client requests are generated on a service
4750 that has previously never seen so many requests it is possible
4751 that the request buffer growth cannot keep up with the spike in
4752 demand. Instead of dropping incoming requests, they are held in
4753 the LND until the RPC service can accept more requests.
4756 Frequency : Sometimes during replay
4758 Description: Assertion failure in ll_local_open after replay.
4759 Details : If replay happened on an open request reply before we were able
4760 to set replay handler, reply will become not swabbed tripping the
4761 assertion in ll_local_open. Now we set the handler right after
4762 recognising of open request
4765 Frequency : very rare
4767 Description: kernel reports "badness in vsnprintf"
4768 Details : Reading from the "recovery_status" /proc file in small chunks
4769 may cause a negative length in lprocfs_obd_rd_recovery_status()
4770 call to vsnprintf() (which is otherwise harmless). Exit early
4771 if there is no more space in the output buffer.
4773 Severity : enhancement
4775 Description: clear OBD RPC statistics by writing to them
4776 Details : It is now possible to clear the OBD RPC statistics by writing
4777 to the "stats" file.
4782 Description: Client mtime is not the same on different clients after utimes
4783 Details : In some cases, the client was using the utimes() syscall on
4784 a file cached on another node. The clients now validate the
4785 ctime from the MDS + OSTs to determine which one is right.
4790 Description: Inability to activate failout mode
4791 Details : lconf script incorrectly assumed that in pythong string's numeric
4792 value is used in comparisons.
4795 Frequency : always with multiple stripes per file
4797 Description: Inefficient object allocation for mutli-stripe files
4798 Details : When selecting which OSTs to stripe files over, for files with
4799 a stripe count that divides evenly into the number of OSTs,
4800 the MDS is always picking the same starting OST for each file.
4801 Return the OST selection heuristic to the original design.
4806 Description: mount failures may take full timeout to return an error
4807 Details : Under some heavy load conditions it is possible that a
4808 failed mount can wait for the full obd_timeout interval,
4809 possibly several minutes, before reporting an error.
4810 Instead return an error as soon as the status is known.
4812 Frequency : quota enabled and large files being deleted
4814 Description: releasing more than 4GB of quota at once hangs OST
4815 Details : If a user deletes more than 4GB of files on a single OST it
4816 will cause the OST to spin in an infinite loop. Release
4817 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4822 Description: statfs data retrieved from /proc may be stale or zero
4823 Details : When reading per-device statfs data from /proc, in the
4824 {kbytes,files}_{total,free,avail} files, it may appear
4825 as zero or be out of date.
4828 Frequency : systems with MD RAID1 external journal devices
4830 Description: lconf's call to blkid is confused by RAID1 journal devices
4831 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4832 of returning all block devices that match the journal UUID.
4835 Frequency : always, for aggregate stripe size over 4GB
4837 Description: assertion fails when trying to use 4GB stripe size
4838 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4839 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4842 Frequency : always on ppc64
4844 Description: the first write on an ext3 filesystem with mballoc got stuck
4845 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4846 perform endianness conversion.
4848 ------------------------------------------------------------------------------
4850 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4852 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4853 INTEROPERATE with older versions automatically. Please read the
4854 user documentation before upgrading any part of a live system.
4855 * WARNING: Lustre networking configuration changes are required with
4856 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4859 * Support for kernels:
4860 2.6.9-22.0.2.EL (RHEL 4)
4861 2.6.5-7.244 (SLES 9)
4862 2.6.12.6 vanilla (kernel.org)
4865 Severity : enhancement
4866 Bugzilla : 7981/8208
4867 Description: Introduced Lustre Networking (LNET)
4868 Details : LNET is new networking infrastructure for Lustre, it includes
4869 a reorganized network configuration mode (see the user
4870 documentation for full details) as well as support for routing
4871 between different network fabrics. Lustre Networking Devices
4872 (LNDs) for the supported network fabrics have also been
4873 created for this new infrastructure.
4875 Severity : enhancement
4876 Description: Introduced Access control lists
4877 Details : clients can set ACLs on files and directories in order to have
4878 more fine-grained permissions than the standard Unix UGO+RWX.
4879 The MDS must be started with the "-o acl" mount option.
4881 Severity : enhancement
4882 Description: Introduced filesystem quotas
4883 Details : Administrators may now establish per-user quotas on the
4886 Severity : enhancement
4888 Description: Configuration change for the XT3
4889 The PTLLND is now used to run Lustre over Portals on the XT3
4890 The configure option(s) --with-cray-portals are no longer used.
4891 Rather --with-portals=<path-to-portals-includes> is used to
4892 enable building on the XT3. In addition to enable XT3 specific
4893 features the option --enable-cray-xt3 must be used.
4898 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4899 Details : A race between allocating a new llcd and re-getting the llcd_lock
4900 allowed another thread to grab newly-allocated llcd.
4902 Severity : enhancement
4904 Description: 2.6 OST async journal commit and locking fix to improve performance
4905 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4906 2.6 kernels are now async as they already were in 2.4 kernels so
4907 that they can commit concurrently with the network bulk transfer.
4908 For block-allocated files the filter allocation semaphore is held
4909 to avoid filesystem fragmentation during allocation. BKL lock
4910 removed for 2.6 xattr operations where it is no longer needed.
4915 Description: lconf incorrectly determined whether two IP networks could talk
4916 Details : In some more complicated routing and multiple-network
4917 configurations, lconf will avoid trying to make a network
4918 connection to a disjoint part of the IP space. It was doing the
4919 math incorrectly for one set of cases.
4924 Description: Fix for potential infinite loop processing records in an llog.
4925 Details : If an llog record is corrupted/zeroed, it is possible to loop
4926 forever in llog_process(). Validate the llog record length
4927 and skip the remainder of the block on error.
4930 Frequency : occasional (liblustre only)
4932 Description: liblustre could not open files whose last component is a symlink
4933 Details : sysio_path_walk() would incorrectly pass the open intent to
4934 intermediate path components.
4937 Frequency : rare (liblustre only with non-standard tuning)
4938 Bugzilla : 7201 (7350)
4939 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4940 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4941 after it has been released, only do this for referenced locks
4943 Severity : enhancement
4945 Description: specifying an (invalid) directory default stripe_size of -1
4946 would reset the directory default striping
4947 Details : stripe_size -1 was used internally to signal directory stripe
4948 removal, now use "all default" to signal dir stripe removal
4949 as a directory striping of "all default" is not useful
4952 Frequency : common for large clusters running liblustre clients
4954 Description: doing an ls when liblustre clients are running is slow
4955 Details : sending a glimpse AST to a liblustre client waits for every AST
4956 to time out, as liblustre clients will not respond. Since they
4957 cannot cache data we refresh the OST lock LVB from disk instead.
4959 Severity : enhancement
4961 Description: doing an ls at the same time as file IO can be slow
4962 Details : enqueue and other "small" requests can be blocked behind many
4963 large IO requests. Create a new OST IO portal for non-IO
4964 requests so they can be processed faster.
4967 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4969 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4970 created file with mode 000
4971 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4972 allow an ftruncate() as a non-root user to a file with mode 000.
4973 Lustre now respects this flag to disable mode checks when
4974 truncating a file owned by the user
4977 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4980 Description: Revoking locks from clients that went dead or catatonic might take
4982 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4983 cancellation of such locks instant on servers without waiting for
4984 any reply from clients. Clients drops these locks when cancel
4985 notification from server is received without replying.
4988 Frequency : liblustre-only, when liblustre client dies or becomes busy
4990 Description: Doing ls on Linux clients can take a long time with active
4992 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4993 granting such locks to it in the first place if possible. Locks
4994 are taken by proxy on the OST during the read or write and
4995 dropped immediately afterward. Add connect flags handling, do
4996 not grant locks to liblustre clients for glimpse ASTs.
4998 Severity : enhancement
5000 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
5001 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5002 sequential reads and behaves badly with applications doing random
5003 reads. Improve it by reading ahead at least read region, and
5004 avoiding excessive large RPC for small reads.
5006 Severity : enhancement
5008 Description: Creating more than 1000 files for a single job may cause a load
5009 imbalance on the OSTs if there are also a large number of OSTs.
5010 Details : qos_prep_create() uses an OST index reseed value that is an
5011 even multiple of the number of available OSTs so that if the
5012 reseed happens in the middle of the object allocation it will
5013 still utilize the OSTs as uniformly as possible.
5018 Description: OST or MDS may oops in ping_evictor_main()
5019 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5020 but doesn't restart at beginning of obd_exports_timed list
5023 Severity : enhancement
5025 Description: improve by-nid export eviction on the MDS and OST
5026 Details : allow multiple exports with the same NID to be evicted at one
5027 time without re-searching the exports list.
5030 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5032 Description: MDS may oops in groups_free()
5033 Details : in rare race conditions a newly allocated group_info struct is
5034 freed again, and this can be NULL. The 2.4 compatibility code
5035 for groups_free() checked for a NULL pointer, but 2.6 did not.
5038 Frequency : common for liblustre clients doing little filesystem IO
5039 Bugzilla : 9352, 7313
5040 Description: server may evict liblustre clients accessing contended locks
5041 Details : if a client is granted a lock or receives a completion AST
5042 with a blocking AST already set it would not reply to the AST
5043 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5046 Frequency : lfs setstripe, only systems with more than 160 OSTs
5048 Description: unable to set striping with a starting offset beyond OST 160
5049 Details : llapi_create_file() incorrectly limited the starting stripe
5050 index to the maximum single-file stripe count.
5053 Frequency : LDAP users only
5055 Description: lconf did not handle in-kernel recovery with LDAP properly
5056 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5058 Severity : enhancement
5060 Description: bind OST threads to NUMA nodes to improve performance
5061 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5062 node and do their allocations there to localize memory access
5064 Severity : enhancement
5066 Description: llmount can determine client NID directly from Myrinet (GM)
5067 Details : the client NID code from gmnalnid was moved directly into
5068 llmount, removing the need to use this or specifying the
5069 client NID explicitly when mounting GM clients with zeroconf
5072 Frequency : if client is started with down MDS
5074 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5075 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5076 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5077 ptlrpc_import_delayed_req() because we didn't check if the
5078 request was interrupted, and we also didn't break out of the
5079 event loop if there was no timeout
5084 Description: data loss during non-page-aligned writes to a single file from
5085 both multiple nodes and multiple threads on one node at same time
5086 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5087 inconsistency led to false short-reads, that were cached and later
5088 used by ->prepare_write() to fill in partially written page,
5089 leading to data loss.
5092 Frequency : always, if lconf --abort_recovery used
5094 Description: lconf --abort_recovery fails with 'Operation not supported'
5095 Details : lconf was attempting to abort recovery on the MDT device and not
5098 Severity : enhancement
5100 Description: remove cleanup logs
5101 Details : replace lconf-generated cleanup logs with lustre internal
5102 cleanup routines. Eliminates the need for client-cleanup and
5105 Severity : enhancement
5107 Description: add support for EAs (user and system) on lustre filesystems
5108 Details : it is now possible to store extended attributes in the Lustre
5109 client filesystem, and with the user_xattr mount option it
5110 is possible to allow users to store EAs on their files also
5112 Severity : enhancement
5114 Description: Add possibility (config option) to show minimal available OST free
5116 Details : When compiled with --enable-mindf configure option, statfs(2)
5117 (and so, df) will return least minimal free space available from
5118 all OSTs as amount of free space on FS, instead of summary of
5119 free spaces of all OSTs.
5121 Severity : enhancement
5123 Description: do not expand extent locks acquired on OST-side
5124 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5125 by server: they are not cached anyway.
5128 Frequency : when mmap is used/binaries executed from Lustre
5130 Description: Unmmap pages before throwing them away from read cache.
5131 Details : llap_shrink cache now attempts to unmap pages before discarding
5132 them (if unmapping failed - do not discard). SLES9 kernel has
5133 extra checks that trigger if this unmapping is not done first.
5138 Description: lconf didn't resolve symlinks before checking to see whether a
5139 given mountpoint was already in use
5142 Frequency : when migrating failover services
5143 Bugzilla : 6395, 9514
5144 Description: When migrating a subset of services from a node (e.g. failback
5145 from a failover service node) the remaining services would
5146 time out and evict clients.
5147 Details : lconf --force (implied by --failover) sets the global obd_timeout
5148 to 5 seconds in order to quickly disconnect, but this caused
5149 other RPCs to time out too quickly. Do not change the global
5150 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5152 Severity : enhancement
5153 Frequency : if MDS is started with down OST
5154 Bugzilla : 9439,5706
5155 Description: Allow startup/shutdown of an MDS without depending on the
5156 availability of the OSTs.
5157 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5158 Add appropriate locking and lov-osc refcounts for safe
5159 cleaning. Add osc abort_inflight calls in case the
5160 synchronize never started.
5163 Frequency : occasional (Cray XT3 only)
5165 Description: root not authorized to access files in CRAY_PORTALS environment
5166 Details : The client process capabilities were not honoured on the MDS in
5167 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5168 been accessed by an authorized user then root was able to access
5169 the file on the local client also. The root user capabilities
5170 are now allowed on the MDS, as this environment has secure UID.
5173 Frequency : occasional
5175 Description: ldiskfs "too long searching" message happens too often
5176 Details : A debugging message (otherwise harmless) prints too often on
5177 the OST console. This has been reduced to only happen when
5178 there are fragmentation problems on the filesystem.
5183 Description: Division by zero in statfs when all OSCs are inactive
5184 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5185 lov_statfs divides by value returned by lov_get_stripecnt().
5189 Bugzilla : 9489, 3273
5190 Description: First write from each client to each OST was only 4kB in size,
5191 to initialize client writeback cache, which caused sub-optimal
5192 RPCs and poor layout on disk for the first written file.
5193 Details : Clients now request an initial cache grant at (re)connect time
5194 and so that they can start streaming writes to the cache right
5195 away and always do full-sized RPCs if there is enough data.
5196 If the OST is rebooted the client also re-establishes its grant
5197 so that client cached writes will be honoured under the grant.
5202 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5203 Details : Now I/O RPCs go to different portal number and (presumably) fast
5204 lock requests (and glimses) and other RPCs get their own service
5205 threads pool that should be able to service those RPCs
5208 Severity : enhancement
5210 Description: Ability to exchange lustre version between client and servers and
5211 issue warnings at client side if client is too old. Also for
5212 liblustre clients there is ability to refuse connection of too old
5214 Details : New 'version' field is added to connect data structure that is
5215 filled with version info. That info is later checked by server and
5219 Frequency : rare, liblustre only.
5220 Bugzilla : 9296, 9581
5221 Description: Two simultaneous writes from liblustre at offset within same page
5222 might proceed at the same time overwriting eachother with stale
5224 Details : I/O lock withing llu_file_prwv was released too early, before data
5225 actually was hitting the wire. Extended lock-holding time until
5226 server acknowledges receiving data.
5229 Frequency : extremely rare. Never observed in practice.
5231 Description: avoid generating lustre_handle cookie of 0.
5232 Details : class_handle_hash() generates handle cookies by incrementing
5233 global counter, and can hit 0 occasionaly (this is unlikely, but
5234 not impossible, because initial value of cookie counter is
5235 selected randonly). Value of 0 is used as a sentinel meaning
5236 "unassigned handle" --- avoid it. Also coalesce two critical
5237 sections in this function into one.
5239 Severity : enhancement
5241 Description: allow liblustre clients to delegate truncate locking to OST
5242 Details : To avoid overhead of locking, liblustre client instructs OST to
5243 take extent lock in ost_punch() on client's behalf. New connection
5244 flag is added to handle backward compatibility.
5246 Severity : enhancement
5247 Bugzilla : 4928, 7341, 9758
5248 Description: allow number of OST service threads to be specified
5249 Details : a module parameter allows the number of OST service threads
5250 to be specified via "options ost ost_num_threads={N}" in the
5251 OSS's /etc/modules.conf or /etc/modprobe.conf.
5255 Bugzilla : 6146, 9635, 9895
5256 Description: servers crash with bad pointer in target_handle_connect()
5257 Details : In rare cases when a client is reconnecting it was possible that
5258 the connection request was the last reference for that export.
5259 We would temporarily drop the export reference and get a new
5260 one, but this may have been the last reference and the export
5261 was just destroyed. Get new reference before dropping old one.
5263 Severity : enhancement
5264 Frequency : if client is started with failover MDS
5266 Description: Allow multiple MDS hostnames in the mount command
5267 Details : Try to read the configuration from all specified MDS
5268 hostnames during a client mount in case the "primary"
5271 Severity : enhancement
5273 Description: Stop sending data to evicted clients as soon as possible.
5274 Details : Check if the client we are about to send or are sending data to
5275 was evicted already. (Check is done every second of waiting,
5276 for which l_wait_event interface was extended to allow checking
5277 of exit condition at specified intervals).
5280 Frequency : rare, normally only when NFS exporting is done from client
5282 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5284 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5287 Frequency : always, if srand() is called before liblustre initialization
5289 Description: Liblustre uses system PRNG disturbing its usage by user application
5290 Details : Introduce internal to lustre fast and high-quality PRNG for
5291 lustre usage and make liblustre and some other places in generic
5292 lustre code to use it.
5294 Severity : enhancement
5295 Bugzilla : 9477, 9557, 9870
5296 Description: Verify that the MDS configuration logs are updated when xml is
5297 Details : Check if the .xml configuration logs are newer than the config
5298 logs stored on the MDS and report an error if this is the case.
5299 Request --write-conf, or allow starting with --old_conf.
5301 Severity : enhancement
5303 Description: Handle symlinks in the path when checking if Lustre is mounted.
5304 Details : Resolve intermediate symlinks when checking if a client has
5305 mounted a filesystem to avoid duplicate client mounts.
5310 Description: lconf can hit an error exception but still return success.
5311 Details : The lconf command catches the Command error exception at the top
5312 level script context and will exit with the associated exit
5313 status, but doesn't ensure that this exit status is non-zero.
5318 Description: failure of ptlrpc thread startup can cause oops
5319 Details : Starting a ptlrpc service thread can fail if there are a large
5320 number of threads or the server memory is very fragmented.
5321 Handle this without oopsing.
5324 Frequency : always, only if liblustre and non-default acceptor port was used
5326 Description: liblustre cannot connect to servers with non-default acceptor port
5327 Details : tcpnal_set_default_params() was not called and was therefore
5328 ignoring the environment varaible TCPNAL_PORT, as well as other
5329 TCPNAL_ environment variables
5334 Description: two objects could be created on the same OST for a single file
5335 Details : If an OST is down, in some cases it was possible to create two
5336 objects on a single OST for a single file. No problems other
5337 than potential performance impact and spurious error messages.
5341 Bugzilla : 5681, 9562
5342 Description: Client may oops in ll_unhash_aliases
5343 Details : Client dcache may become inconsistent in race condition.
5344 In some cases "getcwd" can fail if the current directory is
5350 Description: Inode refcounting problems in NFS export code
5351 Details : link_raw functions used to call d_instantiate without obtaining
5352 extra inode reference first.
5356 Bugzilla : 9942, 9903
5357 Description: Referencing freed requests leading to crash, memleaks with NFS.
5358 Details : We used to require that call to ll_revalidate_it was always
5359 followed by ll_lookup_it. Also with revalidate_special() it is
5360 possible to call ll_revalidate_it() twice for the same dentry
5361 even if first occurence returned success. This fix changes semantic
5362 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5363 reference on a request referred from the intent.
5364 ll_intent_release() then releases such a request.
5367 Frequency : rare, normally benchmark loads only
5369 Description: unlinked inodes were kept in memory on the client
5370 Details : If a client is repeatedly creating and unlinking files it
5371 can accumulate a lot of stale inodes in the inode slab cache.
5372 If there is no other client load running this can cause the
5373 client node to run out of memory. Instead flush old inodes
5374 from client cache that have the same inode number as a new inode.
5377 Frequency : SLES9 2.6.5 kernel and long filenames only
5378 Bugzilla : 9969, 10379
5379 Description: utime reports stale NFS file handle
5380 Details : SLES9 uses out-of-dentry names in some cases, which confused
5381 the lustre dentry revalidation. Change it to always use the
5385 Frequency : rare, unless heavy write-truncate concurrency is continuous
5386 Bugzilla : 4180, 6984, 7171, 9963, 9331
5387 Description: OST becomes very slow and/or deadlocked during object unlink
5388 Details : filter_destroy() was holding onto the parent directory lock
5389 while truncating+unlinking objects. For very large objects this
5390 may block other threads for a long time and slow overall OST
5391 responsiveness. It may also be possible to get a lock ordering
5392 deadlock in this case, or run out of journal credits because of
5393 the combined truncate+unlink. Solution is to do object truncate
5394 first in one transaction without parent lock, and then do the
5395 final unlink in a new transaction with the parent lock. This
5396 reduces the lock hold time dramatically.
5399 Frequency : rare, 2.4 kernels only
5401 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5402 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5403 before unmounting their filesystem in order to avoid deadlock.
5404 The kernel_locked() function in 2.4 kernels only checks whether
5405 the kernel lock is held, not whether it is this process that is
5406 holding it as 2.6 kernels do.
5411 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5412 Details : The client ptlrpc code may be trying to reconnect to a down
5413 server before a previous connection attempt has timed out.
5414 Increase the reconnect interval to be longer than the connection
5415 timeout interval to avoid sending duplicate connections to
5419 Frequency : echo_client brw_test command
5421 Description: fix echo_client to work with OST preallocated code
5422 Details : OST preallocation code (5137) didn't take echo_client IO path
5423 into account: echo_client calls filter methods outside of any
5424 OST thread and, hence, there is no per-thread preallocated
5425 pages and buffers to use. Solution: hijack pga pages for IO. As
5426 a byproduct, this avoids unnecessary data copying.
5430 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5431 Description: Client can oops in mdc_commit_close() after open replay
5432 Details : It was possible for the MDS to return an open request with no
5433 transaction number in mds_finish_transno() if the client was
5434 evicted, but without actually returning an error. Clients
5435 would later try to replay that open and may trip an assertion
5436 Simplify the client close codepath, and always return an error
5437 from the MDS in case the open is not successful.
5440 Frequency : rare, 2.6 OSTs only
5442 Description: OST may deadlock under high load on fragmented files
5443 Details : If there was a heavy load and highly-fragmented OST filesystems
5444 it was possible to have all the OST threads deadlock waiting on
5445 allocation of biovecs, because the biovecs were not released
5446 until the entire RPC IO was completed. Instead, release biovecs
5447 as soon as they are complete to ensure forward IO progress.
5449 Severity : enhancement
5451 Description: Support for specifying external journal device at mount
5452 Details : If an OST or MDS device is formatted with an external journal
5453 device, this device major/minor is stored in the ext3 superblock
5454 and may not be valid for failover. Allow detecting and
5455 specifying the external journal at mount time.
5460 Description: Mounting an MDS with pending unlinked files may cause oops
5461 Details : target_finish_recovery() calls mds_postrecov() which returned
5462 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5463 considers this an error and immediately begins cleaning up the
5464 lov, just after starting the mds_lov process
5466 Severity : enhancement
5468 Description: Implement 'lfs df' to report actual free space on per-OST basis
5469 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5470 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5471 usage of objects; '-h' to report in human readable format.
5473 ------------------------------------------------------------------------------
5475 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5482 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5483 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5484 at startup during orphan unlinking. Instead, skip buddy update
5485 before setup, as it will be regenerated after recovery is complete.
5490 Description: If an OST is inactive, its locks might reference stale inodes.
5491 Details : lov_change_cbdata() must iterate over all namespaces, even if
5492 they are inactive to clear inode references from the lock.
5494 Severity : enhancement
5495 Frequency : occasional, if non-standard max_dirty_mb used
5497 Description: Client will block write RPCs if not enough grant
5498 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5499 then the client will block writes while waiting for another RPC
5500 to complete instead of consuming its dirty limit. With change
5501 we get improved performance when max_dirty_mb is small.
5503 Severity : enhancement
5504 Bugzilla : 3389, 6253
5505 Description: Add support for supplementary groups on the MDS.
5506 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5507 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5508 lookups for user supplementary groups into a cache.
5512 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5513 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5514 for liblustre O_EXCL is enforced by MDS.
5518 Description: readdir never returns NULL in liblustre.
5519 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5520 dentry in struct dirent.
5524 Frequency : liblustre only, depends on application IO pattern
5525 Description: liblustre clients evicted if not contacting servers
5526 Details : Don't put liblustre clients into the ping_evictor list, so
5527 they will not be evicted by the pinger ever.
5529 Severity : enhancement
5531 Description: Add ability to evict clients by NID from MDS.
5532 Details : By echoing "nid:$NID" string into
5533 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5534 $NID would be instantly evicted from this MDS and from all active
5535 OSTs connected to it.
5539 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5540 Details : lookup_it_finish() used to query file size from OSTs that was not
5545 Description: service threads change working directory to that of init
5546 Details : Starting lustre service threads may pin the working directory
5547 of the parent thread, making that filesystem busy. Threads
5548 now change to the working directory of init to avoid this.
5552 Frequency : during shutdown only
5553 Description: shutdown with a failed MDS or OST can cause unmount to hang
5554 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5559 Frequency : 2.6 only
5560 Description: chmod/chown may include an extra supplementary group
5561 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5562 supplementary group and if none is specified this is used.
5565 Bugzilla : 5479 (6816)
5567 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5568 Details : If lookup is in progress on a file that is unlinked we might try
5569 to revalidate the inode and fail in revalidate after lookup is
5570 complete and ll_file_open() enqueues the open again but
5571 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5574 Frequency : always, if lconf --abort_recovery used
5576 Description: lconf --abort_recovery fails with 'Operation not supported'
5577 Details : lconf was attempting to abort recovery on the MDT device and not
5580 ------------------------------------------------------------------------------
5582 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5587 Frequency : rare (only unsupported configurations with a node running as an
5589 Bugzilla : 6514, 5137
5590 Description: Mounting a Lustre file system on a node running as an OST could
5592 Details : OSTs now preallocates memory needed to write out data at
5593 startup, instead of when needed, to avoid having to
5594 allocate memory in possibly low memory situations.
5595 Specifically, if the file system is mounted on on OST,
5596 memory pressure could force it to try to write out data,
5597 which it needed to allocate memory to do. Due to the low
5598 memory, it would be unable to do so and the node would
5599 become unresponsive.
5601 Severity : enhancement
5603 Description: Addition of lconf --service command line option
5604 Details : lconf now accepts a '--service <arg>' option, which is
5605 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5607 Severity : enhancement
5609 Description: Failover mode is now the default for OSTs.
5610 Details : By default, OSTs will now run in failover mode. To return to
5611 the old behaviour, add '--failout' to the lmc line for OSTs.
5613 Severity : enhancement
5615 Description: Health checks are now provided for MDS and OSTs
5616 Details : Additional detailed health check information on MSD and OSTs
5617 is now provided through the procfs health_check value.
5620 Frequency : occasional, depends on IO load
5622 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5623 after numerous create/delete cycles
5624 Details : The ext3 inode allocation policy would not allocate new inodes
5625 very well on the OSTs because there are no new directories
5626 being created. Instead we look for groups with free space if
5627 the parent directories are nearly full.
5632 Description: Network or server problems during mount may cause partially
5633 mounted clients instead of returning an error.
5634 Details : The config llog parsing code may overwrite the error return
5635 code during mount error handling, returning success instead
5641 Description: MDS can fail to allocate large reply buffers
5642 Details : After long uptimes the MDS can fail to allocate large reply
5643 buffers (e.g. zconf client mount config records) due to memory
5644 fragmentation or consumption by the buffer cache. Preallocate
5645 some large reply buffers so that these replies can be sent even
5646 under memory pressure.
5650 Frequency : rare (liblustre)
5651 Description: fsx running with liblustre complained that using truncate() to
5652 extend the file doesn't work. This patch corrects that issue.
5653 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5654 ATTR_SIZE and lsm use in llu_setattr_raw.
5658 Frequency : rare, only 2.6 kernels
5659 Description: Unusual file access patterns on the MDS may result in inode
5660 data being lost in very rare circumstances.
5661 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5662 code in the 2.6 kernel caused the inode and/or EA data not to
5663 be read from disk, causing single-file corruption.
5667 Frequency : rare, only 2.6 filesystems using extents
5668 Description: Heavy concurrent write and delete load may cause data corruption.
5669 Details : It was possible under high-load situations to have an extent
5670 metadata block in the block device cache from a just-unlinked
5671 file overwrite a newly-allocated data block. We now unmap any
5672 metadata buffers that alias just-allocated data blocks.
5676 Frequency : filesystems with default stripe_count larger than 77
5677 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5678 Details : lconf specifies an inode size of 4096 bytes when the default
5679 stripe_count is larger than 77. This conflicts with the default
5680 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5681 this case to avoid pinning too much memory for large EAs.
5683 ------------------------------------------------------------------------------
5685 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5690 Frequency : rare (extremely heavy IO load with hundreds of clients)
5692 Description: Client is evicted, gets IO error writing to file
5693 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5694 caused clients to be evicted for AST timeouts. The fixes in
5695 bug 5192 mean we no longer need to have such short AST timeouts
5696 so ldlm_timeout has been increased.
5699 Frequency : occasional during --force or --failover shutdown under load
5700 Bugzilla : 5949, 4834
5701 Description: Server oops/LBUG if stopped with --force or --failover under load
5702 Details : a collection of import/export refcount and cleanup ordering
5703 issues fixed for safer force cleanup
5706 Frequency : only filesystems larger than 120 OSTs
5707 Bugzilla : 5990, 6223
5708 Description: lfs getstripe would oops on a very large filesystem
5709 Details : lov_getconfig used kfree on vmalloc'd memory
5712 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5713 Bugzilla : 6242, 6243
5714 Description: reading from files that had been truncated to a non-zero size
5715 but never opened returned no data
5716 Details : ll_file_read() reads zeros from no-object files to EOF
5721 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5722 Details : There's little harm in aborting MDS/OSS recovery and letting it
5723 try again, so I removed the LASSERT and return an error instead.
5725 Severity : enhancement
5727 Description: New debugging infrastructure for tracking down data corruption
5728 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5729 (b) cover more of the client-side code path, and (c) try to narrow
5730 down where problems occurred
5734 Bugzilla : 3819, 4364, 4397, 6313
5735 Description: Racing close and eviction MDS could cause assertion in mds_close
5736 Details : It was possible to get multiple mfd references during close and
5737 client eviction, leading to one thread referencing a freed mfd.
5739 Severity: : enhancement
5740 Bugzilla : 3262, 6359
5741 Description: Attempts to reconnect to servers are now more aggressive.
5742 Details : This builds on the enhanced upcall-less recovery that was added
5743 in 1.4.2. When trying to reconnect to servers, clients will
5744 now try each server in the failover group every 10 seconds. By
5745 default, clients would previously try one server every 25 seconds.
5750 Description: After recovery, certain operations trigger a failed
5751 assertion on a client.
5752 Details : Failing over an mds, using lconf -d --failover, while a
5753 client was doing a readdir() call would cause the client to
5754 LBUG after recovery completed and the readdir() was resent.
5756 Severity : enhancement
5758 Description: Default groups are now added by lconf
5759 Details : You can now run lconf --group <servicename> without having to
5760 manually add groups with lmc.
5763 Frequency : occasional
5765 Description: Nodes with an elan id of 0 trigger a failed assertion
5768 Frequency : always when accessing e.g. tty/console device nodes
5770 Description: tty and some other devices nodes cannot be used on lustre
5771 Details : file's private_data field is used by device data and lustre
5772 values in there got lost. New field was added to struct file to
5773 store fs-specific private data.
5776 Frequency : when exporting Lustre via NFS
5778 Description: NFSD failed occasionally when looking up a path component
5779 Details : NFSD is looking up ".." which was broken in ext3 directories
5780 that had grown large enough to become hashed.
5783 Frequency : Clusters with multiple interfaces not on the same subnet
5785 Description: Nodes will repeatedly try to reconnect to an interface which it
5786 cannot reach and report an error to the log.
5787 Details : Extra peer list entries will be created by lconf with some peers
5788 unreachable. lconf now validates the peer before adding it.
5791 Frequency : Only if a default stripe is set on the filesystem root.
5793 Description: Setting a default stripe on the filesystem root prevented the
5794 filesystem from being remounted.
5795 Details : The client was sending extra request flags in the root getattr
5796 request and did not allocate a reply buffer for the dir EA.
5799 Frequency : occasional, higher if lots of files are accessed by one client
5800 Bugzilla : 6159, 6097
5801 Description: Client trips assertion regarding lsm mismatch/magic
5802 Details : While revalidating inodes the VFS looks up inodes with ifind()
5803 and in rare cases can find an inode that is being freed.
5804 The ll_test_inode() code will free the lsm during ifind()
5805 when it finds an existing inode and then the VFS later attaches
5806 this free lsm to a new inode.
5810 Bugzilla : 6422, 7030
5811 Description: MDS deadlock between mkdir and client eviction
5812 Details : Creating a new file via mkdir or mknod (starting a transaction
5813 and getting the ns lock) can deadlock with client eviction
5814 (gets ns lock and trying to finish a synchronous transaction).
5817 Frequency : occasional
5818 Description: While starting a server, the fsfilt_ext3 module could not be
5820 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5821 kernels. Previously, lconf would still use the ext3 name
5822 when trying to load modules. Now, it will correctly use
5823 ext3 on 2.4 and ldiskfs on 2.6.
5825 Severity : enhancement
5826 Description: The default stripe count has been changed to 1
5827 Details : The interpretation of the default stripe count (0, to lfs
5828 or lmc) has been changed to mean striping across a single
5829 OST, rather than all available. For general usage we have
5830 found a stripe count of 1 or 2 works best.
5832 Severity : enhancement
5833 Description: Add support for compiling against Cray portals.
5834 Details : Conditional compiling for some areas that are different
5838 Frequency : occasional
5839 Bugzilla : 6409, 6834
5840 Description: Creating files with an explicit stripe count may lead to
5841 a failed assertion on the MDS
5842 Details : If some OSTs are full or unavailable, creating files may
5843 trigger a failed assertion on the MDS. Now, Lustre will
5844 try to use other servers or return an error to the
5848 Frequency : occasional
5850 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5851 caused lock timeout during readahead (since 1.4.2).
5852 Details : Processes doing readahead might match a lock that hasn't been
5853 granted yet if there are overlapping and conflicting lock
5854 requests. The readahead process waits on ungranted lock
5855 (original lock is CBPENDING), while OST waits for that process
5856 to cancel CBPENDING read lock and eventually evicts client.
5858 Severity : enhancement
5860 Description: Initial enabling of flock support for clients
5861 Details : Implements fcntl advisory locking and file status functions.
5862 This feature is provided as an optional mount flag (default
5863 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5864 locking are implemented yet, and those that are are not guaranteed
5865 to be completely correct in production environments.
5866 mount -t lustre -o [flock|noflock] ...
5869 Frequency : occasional
5871 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5872 assertion in the ext3 JBD (journaling) layer.
5873 Details : The b_committed_data struct is protected by the big kernel lock
5874 in 2.4 kernels, serializing journal_commit_transaction() and
5875 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5876 access to this struct. In 2.6 kernels there is finer grained
5877 locking to improve SMP performance of the JBD layer.
5881 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5886 Description: merge and cleanup kernel patches.
5889 -----------------------------------------------------------------------------
5891 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5893 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5894 versions of Lustre. Please update all servers and clients to
5895 version 1.4.2 or later at the same time. You must also run
5896 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5897 to update the configuration logs.
5899 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5900 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5901 - swab llog records of type '0' so we get proper header size/idx (5861)
5902 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5903 - fix rename of one directory over another leaking an inode (5953)
5904 - avoid SetPageDirty on 2.6 (5981)
5905 - don't re-add just-being-destroyed locks to the waiting list (5653)
5906 - when creating new directories, inherit the parent's custom
5907 striping settings if present parent (3048)
5908 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5909 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5910 - don't hold client locks on temporary worklist from l_lru (5666)
5911 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5912 - automatically evict dead clients (5921)
5913 - Update file size properly in create+truncate+fstat case (6196)
5914 - Do not unhash mountpoint dentries, do not allow removal of
5916 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5917 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5918 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5919 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5920 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5921 - Correctly handle reads of files with no objects (6243)
5922 - lctl recover will also mark a device active if deactivate used (5933)
5924 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5925 - allow --write-conf on an MDS with different nettype than client (5619)
5926 - don't write config llogs to MDS for mounts not from that MDS (5617)
5927 - lconf should create multiple TCP connections from a client (5201)
5928 - init scripts are now turned off by default; run chkconfig --on
5929 lustre and chkconfig --on lustrefs to use them
5930 - upcalls are no longer needed for clients to recover to failover
5932 - add --abort-recovery option to lconf to abort recovery on device
5934 - add support for an arbitrary number of OSTs (3026)
5935 - Quota support protocol changes.
5936 - forward compatibility changes to wire structs (6007)
5937 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5938 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5939 - improved "lustre status" script
5940 - initialize blocksize for non-regular files (6062)
5941 - added --disable-server and --disable-client configure options (5782)
5942 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5943 - Vanilla 2.4.29 support
5944 - increase maximum number of obd devices to 520 (6242)
5945 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5946 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5947 SLES 9 2.6.5 kernels
5948 - limit stripes per file to 160 (the maximum EA size) (6093)
5950 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5953 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5954 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5955 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5956 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5957 - small ext3 extents cleanups and fixes (5733)
5958 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5959 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5960 - avoid needless client->OST connect, fix handle mismatch (5317)
5961 - fix DLM error path that led to out-of-sync client, long delays (5779)
5962 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5963 - fix several locking issues related to i_size (5492,5624,5654,5672)
5964 - don't move pending lock onto export if it is already evicted (5683)
5965 - fix kernel oops when creating .foo in unlinked directory (5548)
5966 - fix deadlock in obdfilter statistics vs. object create (5811)
5967 - use time_{before,after} to avoid timer jiffies wrap (5882)
5968 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5969 - Do not leak request if server was not able to process it (5154)
5970 - If mds_open unable to find parent dir, make that negative lookup(5154)
5971 - don't create new directories with extent-mapping (5909, 5936)
5973 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5974 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5975 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5976 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5977 - support for RHEL4 kernel on i686 (5773)
5978 - provide error messages when incompatible logs are encountered (5898)
5980 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5981 * version 1.4.0.10 (1.4.1 release candidate 1)
5983 - don't keep a lock reference when lock is not granted (4238)
5984 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5985 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5986 - drop import inflight refcount on signal_completed_replay error (5255)
5987 - unlock page after async write error during send (3677)
5988 - handle missing objects in filter_preprw_read properly (5265)
5989 - no transno return for symlink open, don't save no-trasno open (3440)
5990 - don't try to complete elan receive that already failed (4012)
5991 - free RPC server reply state on error (5406)
5992 - clean up thread from ptlrpc_start_thread() on error (5160)
5993 - readahead could read extra page into cache that wasn't ejected (5388)
5994 - prevent races in class_attach/setup/cleanup/detach (5260)
5995 - don't dereference de->d_inode after l_dput of de (5458)
5996 - use "int" for stripe value returned from lock_to_stripe (5544)
5997 - mballoc allocation and error-checking fixes in 2.6 (5504)
5998 - block device patches to fix I/O request sizes in 2.6 (5482)
5999 - look up hostnames for IB nals (5602)
6000 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
6001 - don't start multiple acceptors for the same port (5277)
6002 - fix incorrect LASSERT in mds_getattr_name (5635)
6003 - export a proc file for general "ping" checking (5628)
6004 - fix "lfs check" to not block when the MDS is down (5628)
6006 - service request history (4965)
6007 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6008 - create an "evict_client" /proc entry on OSTs, like the MDS has
6009 - fix mount usage message, return errors per mount(8) (5168)
6010 - change grep [] to grep "[]" in tests so they work in more UMLs
6011 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6012 - remove ancient LOV_MAGIC_V0 EA support (5047)
6013 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6014 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6015 - allow readahead allocations to fail when low on memory (5383)
6016 - mmap locking landed again, after considerable improvement (2828)
6017 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6019 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6022 - send OST transaction number in read/write reply to free req (4966)
6023 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6024 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6025 - the watchdog thread now runs as interruptible (5246)
6026 - flock/lockf fixes (but it's still disabled, pending 5135)
6027 - don't use EXT3 constants in llite code (5094)
6028 - memory shortage at startup could cause assertion (5176)
6030 - reorganization of lov code
6031 - single portals codebase
6033 - add extents/mballoc support (5025)
6034 - direct I/O reads in the obdfilter (4048)
6035 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6037 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6040 - send OST transaction number in read/write reply to free req (4966)
6041 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6042 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6043 - the watchdog thread now runs as interruptible (5246)
6044 - handle missing objects in filter_preprw_read properly (5265)
6045 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6046 - drop import inflight refcount on signal_completed_replay error (5255)
6047 - unlock page after async write error during send (3677)
6048 - return original error code on reconstructed replies (3761)
6049 - no transno return for symlink open, don't save no-trasno open (3440)
6051 - add pid to ldlm debugging output (4922)
6052 - bump the watchdog timeouts -- we can't handle 30sec yet
6053 - extra debugging for orphan dentry/inode bug (5259)
6055 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6058 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6059 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6060 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6061 - fix NULL dereference in /proc/sys/portals/routes (4827)
6062 - allow failed mdc_close() operations to be interrupted (4561)
6063 - stop precreate on OST before MDS would time out on it (4778)
6064 - don't send partial-page writes before EOF from client (4410)
6065 - discard client grant for sub-page writes on large-page clients (4520)
6066 - don't free dentries not owned by NFS code, check generation (4806)
6067 - fix lsm leak if mds_create_objects() fails (4801)
6068 - limit debug_daemon file size, always print CERROR messages (4789)
6069 - use transno after validating reply (3892)
6070 - process timed out requests if import state changes (3754)
6071 - update mtime on OST during writes, return in glimpse (4829)
6072 - add mkfsoptions to LDAP (4679)
6073 - use ->max_readahead method instead of zapping global ra (5039)
6074 - don't interrupt __l_wait_event() during strace
6076 - add software watchdogs to catch hung threads quickly (4941)
6077 - make lustrefs init script start after nfs is mounted
6078 - fix CWARN/ERROR duplication (4930)
6079 - return async write errors to application if possible (2248)
6080 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6081 - print NAL number in %x format (4645)
6082 - update barely-supported suse-2.4.21-171 series (4842)
6083 - support for sles 9 %post scripts
6084 - support for building 2.6 kernel-source packages
6085 - support for sles km_* packages
6087 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6090 - ignore -ENOENT errors in osc_destroy (3639)
6091 - notify osc create thread that OSC is being cleaned up (4600)
6092 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6093 - reconstruct ost_handle() like mds_handle() (4657)
6094 - create a new thread to do import eviction to avoid deadlock (3969)
6095 - let lconf resolve symlinked-to devices (4629)
6096 - don't unlink "objects" from directory with default EA (4554)
6097 - hold socknal file ref over connect in case target is down (4394)
6098 - allow more than 32000 subdirectories in a single directory (3244)
6099 - fix blocks count for O_DIRECT writes (3751)
6100 - OST returns ENOSPC from object create when no space left (4539)
6101 - don't send truncate RPC if file size isn't changing (4410)
6102 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6103 - bind to privileged port in socknal and tcpnal (3689)
6105 - rate limit CERROR/CWARN console message to avoid overload (4519)
6106 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6107 - basic mmap support (3918)
6108 - kernel patch series update from b1_4 (4711)
6110 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6113 - avoid crash during MDS cleanup with OST shut down (2775)
6114 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6115 - don't use bad inodes on the MDS (3744)
6116 - dynamic object preallocation to improve recovery speed (4236)
6117 - don't hold spinlock over lock dumping or change debug flags (4401)
6118 - don't zero obd_dev when it is force cleaned (3651)
6119 - print grants to console if they go negative (4431)
6120 - "lctl deactivate" will stop automatic recovery attempts (3406)
6121 - look for existing locks in ldlm_handle_enqueue() (3764)
6122 - don't resolve lock handle twice in recovery avoiding race (4401)
6123 - revalidate should check working dir is a directory (4134)
6125 - don't always mark "slow" obdfilter messages as errors (4418)
6127 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6130 - don't close LustreDB during write_conf until it is done (3860)
6131 - fix typo in lconf for_each_profile (3821)
6132 - allow dumping logs from multiple threads at one time (3820)
6133 - don't allow multiple threads in OSC recovery (3812)
6134 - fix debug_size parameters (3864)
6135 - fix mds_postrecov to initialize import for llog ctxt (3121)
6136 - replace config semaphore with spinlock (3306)
6137 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6138 - don't allow enqueue to complete on a destroyed export (3822)
6139 - down write_lock before checking llog header bitmap (3825)
6140 - recover from lock replay timeout (3764)
6141 - up llog sem before sending rpc (3652)
6142 - reduce ns lock hold times when setting kms (3267)
6143 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6144 - fix NULL deref and obd_dev leak on setup error (3312)
6145 - replace some LBUG about llog ops with error handling (3841)
6146 - don't match INVALID dentries from d_lookup and spin (3784)
6147 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6148 - fix invalid assertion in ptlrpc_set_wait (3880)
6150 - add libwrap support for the TCP acceptor (3996)
6151 - add /proc/sys/portals/routes for non-root route listing (3994)
6152 - allow setting MDS UUID in .xml (2580)
6153 - print the stack of a process that LBUGs (4228)
6155 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6158 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6159 - make sure to unset replay flag from failed open requests (3440)
6160 - if default stripe count is 0, use OST count for inode size (3636)
6161 - update parent mtime/ctime on client for create/unlink (2611)
6162 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6163 - free recovery state on server during a forced cleanup (3571)
6164 - unregister_reply for resent reqs (3063)
6165 - loop back devices mounting and status check on 2.6 (3563)
6166 - fix resource-creation race that can provoke i_size == 0 (3513)
6167 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6168 - more debugging for page-accounting assertion (3746)
6169 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6170 - avoid deadlock after precreation failure (3758)
6171 - fix race and lock order deadlock in orphan handling (3450, 3750)
6172 - add validity checks when grabbing inodes from l_ast_data (3599)
6174 - add /proc/.../recovery_status to obdfilter (3428)
6175 - lightweight CDEBUG infrastructure, debug daemon (3668)
6176 - change default OSC RPC parameters to be better on small clusters
6177 - turn off OST read cache for files smaller than 32MB
6178 - install man pages and include them in rpms (3100)
6179 - add new init script for (un)mounting lustre filesystems (2593)
6180 - run chkconfig in %post for init scripts (3701)
6181 - drop scimac NAL (unmaintained)
6183 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6186 - clean kiobufs before and after use (3485)
6187 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6188 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6189 - change init script to fail more clearly if not run as root (1528)
6190 - allow clients to reconnect during replay (1742)
6191 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6192 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6193 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6194 - fix cond_resched (3554)
6195 - only evict unfinished clients after recovery (3515)
6196 - allow bulk resend, prevent data loss (3570)
6197 - dynamic ptlrpc request buffer allocation (2102)
6198 - don't allow unlinking open directory if it isn't empty (2904)
6199 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6200 - remove extraneous obd dereference causing LASSERT failure (3334)
6201 - don't use get_cycles() when creating temp. files on the mds (3156)
6202 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6203 - handle EEXIST for set-stripe, set proper directory name (3336)
6205 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6206 - fix ksocknal_fmb_callback() error messages (2918)
6208 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6211 - don't copy lvb into (possibly NULL) reply on error (2983)
6212 - don't deref dentry after dput, don't free lvb on error (2922)
6213 - use the kms to determine writeback rpc length (2947)
6214 - increment oti_logcookies when osc is inactive (2948)
6215 - update client's i_blocks count via lvb messages (2543)
6216 - handle intent open/close of special files properly (1557)
6217 - mount MDS with errors=remount-ro, like obdfilter (2009)
6218 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6219 - don't use cancelling-locks' kms values (2947)
6220 - use highest lock extent for kms, not last one (2925)
6221 - don't dereference ERR_PTR() dentry in error handling path (3107)
6222 - fix thread race in portals_debug_dumplog() (3122)
6223 - create lprocfs device entries at setup instead of at attach (1519)
6224 - common AST error handler, don't evict client on completion race (3145)
6225 - zero nameidata in detach_mnt in 2.6 (3118)
6226 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6227 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6228 - initialize RPC timeout timer earlier for 2.6 (3219)
6229 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6230 - print nal/nid for unknown nid (3258)
6231 - additional checks for oscc recovery before doing precreate (3284)
6232 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6233 - don't crash in mdc_close for bad permissions on open (3285)
6234 - zero i_rdev for non-device files (3147)
6235 - clear page->private before handing to FS, better assertion (3119)
6236 - tune the read pipeline (3236)
6237 - fix incorrect decref of invalidated dentry (2350)
6238 - provide read-ahead stats and refine rpc in flight stats (3328)
6239 - don't hold journal transaction open across create RPC (3313)
6240 - update atime on MDS at close time (3265)
6241 - close LDAP connection when recovering to avoid server load (3315)
6242 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6243 - don't leak open file on MDS after open resend (3325)
6244 - serialize filter_precreate and filter_destroy_precreated (3329)
6245 - loop device shouldn't call sync_dev() for nul device (3092)
6246 - clear page cache after eviction (2766)
6247 - resynchronize MDS->OST in background (2824)
6248 - refuse to mount the same filesystem twice on same mountpoint (3394)
6249 - allow llmount to create routes for mounting behind routers (3320)
6250 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6251 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6252 - fix rare problem with rename on htree directories (3417)
6254 - allow default OST striping configuration per directory (1414)
6255 - fix compilation for qswnal for 2.6 kernels (3125)
6256 - increase maximum number of MDS request buffers for large systems
6257 - change liblustreapi to be useful for external progs like lfsck (3098)
6258 - increase local configuration timeout for slow disks (3353)
6259 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6261 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6264 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6265 - don't overwrite extent policy data in reply if lock was blocked (2901)
6266 - drop filter export grants atomically with removal from device (2663)
6267 - del obd_self_export from work_list in class_disconnect_exports (2908)
6268 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6269 - swab reply message in mdc_close, other PPC fixes (2464)
6270 - fix destroying of named logs (2325)
6271 - overwrite old logs when running lconf --write_conf (2264)
6272 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6273 - fix race in target_handle_connect (2898)
6274 - mds_reint_create() should take same inode create lock (2926)
6275 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6276 - don't close files for self_export to avoid uninitialized obd (2936)
6277 - allow MDS with the same name as client node (2939)
6278 - hold dentry reference for closed log files for unlink (2325)
6279 - reserve space for all logs during transactions (2059)
6280 - don't evict page beyond end of stripe extent (2925)
6281 - don't oops on a deleted current working directory (2399)
6282 - handle hard links to targets without a parent properly (2517)
6283 - don't dereference NULL lock when racing during eviction (2867)
6284 - don't grow lock extents when lots of conflicting locks (2919)
6286 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6289 - account for cache space usage on clients to avoid data loss (974)
6290 - lfsck support in lustre kernel code (2349)
6291 - reduce journal credits needed for BRW writes (2370)
6292 - orphan handling to avoid losing space on client/server crashes
6293 - ptlrpcd can be blocked, stopping ALL progress (2477)
6294 - use lock value blocks to assist in proper KMS, faster stat (1021)
6295 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6296 - recovery for initial connections (2355)
6297 - fixes for mds_cleanup_orphans (1934)
6298 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6299 - block all file creations until orphan recovery completes (1901)
6300 - client remove rq_connection from request struct (2423)
6301 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6302 - recovery timer race (2670)
6303 - mdc_close recovey bug (2532)
6304 - ptlrpc cleanup bug (2710)
6305 - mds timeout on local locks (2588)
6306 - namespace lock held during RPCs (2431)
6307 - handle interrupted sync write properly (2503)
6308 - don't try to handle a message that hasn't been replied to (2699)
6309 - client assert failure during cleanup after abort recovery (2701)
6310 - leak mdc device after failed mount (2712)
6311 - ptlrpc_check_set allows timedout requests to complete (2714)
6312 - wait for inflight reqs when ptlrpcd finishes (2710)
6313 - make sure unregistered services are removed from the srv_list
6314 - reset bulk XID's when resending them (caught by 1138 test)
6315 - unregister_bulk after timeout
6316 - fix lconf error (2694)
6317 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6318 - readahead locks pages, leaves pending causing memory pressure (2673)
6319 - increase OST request buffers to 4096 on large machines (2729)
6320 - fix up permission of existing directories in simple_mkdir (2661)
6321 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6322 - don't assign transno to errored transactions (2742)
6323 - don't delete objects on OST if given a bogus objid from MDS (2751)
6324 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6325 - if rq_no_resend, then timeout request after recovery (2432)
6326 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6327 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6328 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6329 - MDS recovery shouldn't delete objects that it has given out (2730)
6330 - if enqueue arrives after completion, don't clobber LVB (2819)
6331 - don't unlock pages twice when trigger_group_io returns error (2814)
6332 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6333 - don't write pages to disk if there was an error (1450)
6334 - don't ping imports that have recovery disabled (2676)
6335 - take buffered bytes into account when balancing socknal conn (2817)
6336 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6337 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6338 - remove little-endian swabbing of llog records (1987)
6339 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6340 - retry reposting request buffers if they fail (1191)
6341 - grow extent at grant time to avoid granting a revoked lock (2809)
6342 - lock revoke doesn't evict page if covered by a second lock (2765)
6343 - disable VM readahead to avoid reading outside lock extents (2805)
6345 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6346 - updated kernel patches for hp-2.4.20 kernel (2681)
6348 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6351 - fix truncated write corruption (2366)
6352 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6354 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6355 - ensure len(uuid) < 37 in lmc (1171)
6356 - fix ia64 OOPS in llog_test (2255)
6357 - zero end of page at obdfilter for partial page writes (2648)
6358 - don't leave stale dentries around after renames (bug 2428)
6359 - fix timeouts when evicting a client with a single lock held (2642)
6360 - set deadline for the initial HELLO message to drain (2634)
6361 - print out dotted-quad IP addresses in the socknal (2302)
6363 - additional debugging for MDS client eviction problem (2443)
6364 - fix mkfsoptions support for osts (2603, 2604)
6366 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6369 - add series for the vanilla 2.6.0 kernel
6370 - add series for the vanilla 2.4.24 kernel
6371 - add series for a cray x86/64 UL kernel drop
6372 - fix xattr patches for the vanilla 2.4.19 series
6374 - generate true UUIDs in lmc (1171)
6375 - have portals stack dumping break in UML (2466)
6376 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6377 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6378 - wake callback waiting threads on client eviction (2460)
6379 - Add --ptldebug and --subsystem to lmc (1719)
6380 - update assertion to allow safe interrupt allocation
6381 - set rq_no_resend for cancel requests (2432)
6382 - recalculate ptlrpcd timeout after resend (2494)
6383 - call vfs_rmdir when removing pending directories (2368)
6384 - fix renaming a file to itself (2429)
6385 - lmc creates a default one-stripe lov (2454)
6386 - expand procfs space to handle large clusters (2326)
6387 - increase UML stack to avoid overflow
6388 - update lconf's list of debug and subsystem masks
6389 - fix lfs find --obd (2510)
6390 - /proc tunable for disabling filter read caching (2591)
6391 - stop rpm packages from altering slapd.conf (2301)
6392 - disable nagle in the socknal under 0conf (2578)
6393 - choose mds inode size based on stripe count (2572)
6394 - fix kernel-source rpm problems (2516)
6396 - add --disable-doc to avoid pdf generation (2421)
6397 - update documentation, tests, type-os, comments
6398 - avoid format warnings on ia64
6399 - remove the TOE NAL
6400 - tiny code cleanups by removing unused fields
6402 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6405 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6406 - remove the most blatant lies from BUILDING file (2371)
6407 - change default debug level to reasonable production setting
6408 - reduce client side cache size to reduce cache flush time
6409 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6410 - make TCP zerocopy and pinger support enabled by default (2476)
6411 - sync writes completed after process exits caused crashes (2319)
6412 - maintain correct mount count on the MDS (2356)
6413 - backout 1557, because 2316 wasn't really fixed
6414 - better file I/O statistics gathering in /proc
6415 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6416 - another kernel patch to fix zero-copy TCP function export
6417 - don't take duplicate lock when processing re-sent getattr (2420)
6418 - lctl uses obd_self_export instead of creating new conn (2353)
6419 - MDS/OST recovery case which requires object creation asserted (2425)
6420 - move lfs from /usr/sbin to /usr/bin in packages
6421 - fix race between mds_client_add and mds_client_free (2417)
6422 - use kmalloc instead of slabs in portals (2430)
6423 - don't create duplicate records when a failover MDS is present (2442)
6424 - remove unnecessary mount age check (2332)
6425 - don't remove directory inodes from locks prematurely (2451)
6426 - don't break if MDS service name is the same as hostname (2103)
6427 - fix races in client write RPC generation when cache full (2482)
6429 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6432 - remove now-unused request->rq_obd (278)
6433 - if an allocation fails, print out how much memory we've used (1933)
6434 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6435 - add the "configurable stack size" patch to most series files (1256)
6436 - ability to write large log records, for 100+ OST configs (2306)
6437 - fix NULL deref when filter_prep fails (2314)
6438 - fix operator precedence error in filter_sync
6439 - dynamic allocation of socknal TX descriptors (2315)
6440 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6441 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6442 - ensure that CWARN messages in Portals always get to the syslog
6443 - __init/__exit are not for prototype decls (ldlm_init/exit)
6444 - x86-64 compile warning fixes
6445 - fix gateway LMC keyword conflict (2318)
6446 - fix MDS lock inversions in getattr/reint paths (1844)
6447 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6448 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6449 - DLM race condition prevented some lock evictions (2328)
6450 - ENOMEM detection and retry on socknal sends (2230)
6451 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6452 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6453 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6454 - better error messages when a client is rejected during recovery (1505)
6455 - avoid cancelling locks which were never granted, after failure (2330)
6456 - fix i_sem/journal inversion in mds_client_add (2333)
6457 - fix truncate/getattr lock cycle deadlock (2334)
6458 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6459 - fix two rare exit paths which could leak an l_lock() ref (2321)
6460 - fencepost error in MDS/OST orphan recovery (2226)
6461 - make log record alignment 8 bytes (1988)
6462 - lstripe now fails when requested offset > ost_count (2237)
6463 - ensure that all kernel series have a complete list.h (1607)
6464 - fix crashes in special-file operations (2316)
6465 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6467 - allow configurable automake binary, for testing new versions
6468 - small update to the lfs documentation
6470 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6472 * fix negative export reference count in fsfilt_sync (2312)
6474 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6475 * release candidate 0.9.1
6477 - orphans are moved into the PENDING directory for possible recovery
6478 - replayed opens now open by fid for orphan/rename safety (1042)
6479 - last close of an orphan inode generates a transno (683)
6480 - chdir() and mount() now pin the directory entry (1020)
6481 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6482 - discard very old requests without processing them (1502)
6483 - remove some common, well-understood CERRORs (1505)
6484 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6485 - clear "grant" flags in OST replies until OST grant code lands (1644)
6486 - fix read performance by not clobbering i_blksize on client (1598)
6487 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6488 - make LRU size a /proc tunable, clears locks when reduced (707)
6489 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6490 - server locks take references on exports now (1558)
6491 - build fixes for 2.4.20-rh trees (1663)
6492 - return an error from lov_create if all OSCs are inactive (1751)
6493 - fix import levels when a reconnect happens without a timeout (1597)
6494 - exit early from mds_open if we get a lookup error (1749)
6495 - partial page read at EOF wouldn't wait for disk before sending (1642)
6496 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6497 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6498 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6499 - write/truncate lock inversion (1639)
6500 - Don't auto-load obdclass, portals modules during cleanup (1495)
6501 - fix timestamps from jumping to "now" (1763)
6502 - extra journal assertions (1648)
6503 - add an extra multiunlink test (1771)
6504 - fix read_record/write_record API (1776)
6505 - fix leak of offset_extent, possible incorrect i_size later (1772)
6506 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6507 - Debugging for the kqswnal_get_idle_tx problems (1820)
6508 - Allow recovery to be attempted multiple times (1536)
6509 - Write out MDS last_rcvd file after it is first created (1600)
6510 - Fix tx_descriptor leak in failed transmit situations (1827)
6511 - ext3 journaling fixes for assertion failure after IO error (1871)
6512 - class_export_put() on freed export after completion AST error (1896)
6513 - Fix revalidate looping in VFS (1322)
6514 - Don't access a freed export during MDS_REINT timeout (1521)
6515 - Add open-unlink recovery support on the MDS (1673,1764)
6516 - Return an error if no MDS data was read from last_rcvd (1946)
6517 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6518 - Don't setup a disk device that doesn't match exported UUID (317)
6519 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6520 - avoid committing NULL handle in force close
6521 - local.sh is now a one-stripe LOV configuration
6522 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6523 - fix ext3 htree duplicate directory entry corruption (1516)
6524 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6525 - update to new LOV EA format (2097)
6526 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6528 2003-06-15 Phil Schwan <phil@clusterfs.com>
6531 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6532 - per-import/export recovery handling (958, 931, 959)
6533 - multiple last-rcvd slots, for serving multiple FSes (949)
6534 - connections are again shared between multiple imp/exports (963, 964)
6535 - "umount -f" would hang if any requests needed to be sent (393, 978)
6536 - avoid pinning large req buffer by copying for queued messages (989)
6537 - add "uuid" to "lctl device" command to help upcalls (991)
6538 - "open" RPCs with transnos would confuse recovery counters (1037)
6539 - do proper endian conversion of all wire messages (288, 340, 891)
6540 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6541 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6542 - fix ptlrpc_connection leak in target_handle_connect (1174)
6543 - fix import refcounting bug in OST and MDS cleanup (1134)
6544 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6545 - fix very unlikely obd_types race condition (501)
6546 - remove osc_open hack for echo_client (1187)
6547 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6548 - a failure in read_inode2 leads to deadlock (1139)
6549 - cancel ack-locks as soon as transaction is committed (1072)
6550 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6551 - make sure to commitrw after any preprw to avoid deadlock (1162)
6552 - failing to execute a file in a lustre FS would lock inode (1203)
6553 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6554 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6555 - obd_uuid2tgt would walk off the end of the list (1255)
6556 - on IA64 the osc would give portals incorrect bulk size (1258)
6557 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6558 - fix lock inversion caused by new llite matching code (1282)
6559 - limit the number of dirty pages on a client to 10MB (1286)
6560 - timed out locks were not being corrected cancelled (1289)
6561 - fix O_DIRECT above 4GB on IA-32 (1292)
6562 * major user-visible changes
6563 - fail out/fail over policy now controlled by the upcall (993)
6565 - add OBD_PING to check server availability and failure (954)
6566 - lustre messages are now sent in sending host order (288, 340, 891)
6567 - add eadatalen to MDS getattr reply (340)
6568 - OST read replies may contain second buffer, with per-page status (593)
6570 2003-03-11 Phil Schwan <phil@clusterfs.com>
6573 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6574 - failed open()s could cause deadlock; fixed (867, 869)
6575 - stop cancelling OST locks when files are closed (481)
6576 - overlapping XID spaces caused network corruption (851, 853)
6577 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6578 - setattr_raw added, to do single-RPC, server-side setattrs
6579 - lmc/lconf syntax change for OST UUIDs
6580 - fix crashy race condition between ptlrpc_free_req and osc_close
6581 - don't use request in mdc_enqueue if we hit a timeout (889)
6582 - don't set the inode i_size for regular files from the MDS (896)
6583 - handle out of order completion AST (842)
6584 - don't LBUG if a lock request times out after receiving AST (913)
6585 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6586 - if a bad lock AST arrives, send an error instead of dropping entirely
6587 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6588 - fix leak in bulk IO when only partially completed (899, 900, 926)
6589 - fix O_DIRECT for ia64 (55)
6590 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6591 - C-z after timeout could hang a process forever; fixed (977)
6593 - client-side I/O cache (678, 924, 929, 941, 970)
6595 - READPAGE and SETATTRs which don't take server-side locks get
6598 2003-02-11 Phil Schwan <phil@clusterfs.com>
6601 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6602 - Fix fsfilt_extN_readpage() to read a full page of directory
6603 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6604 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6605 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6606 - Most intent operations were removed, in favour of a new RPC mode
6607 that does a single RPC to the server and bypasses most of the VFS
6608 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6609 - Aggressively cancel local locks on DLM servers
6610 - mds_reint_unlink sends EA to the client if it's the last nlink.
6611 client uses that EA to unlink OST objects.
6612 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6613 - recursive symlinks were fixed (439)
6614 - fixed NULL deref in DEBUG_REQ
6615 - filter_update_lastobjid no longer calls sync, which annoyed extN
6616 - fixed multi-client small-writes to a single file problem (445)
6617 - fixed mtime updates during file writes (607)
6618 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6619 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6620 - fixed Linux OST size reporting problem (444, 656)
6621 - OST now updates object mtime with writes or setattr (607, 619)
6622 - client verifies file size before zeroing page past EOF (445)
6623 - OST now writes last allocated objid to disk with allocation (108)
6624 - LOV on echo now works (409)
6626 - mds_reint_unlink sends a new buffer, with the EA included. this
6627 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6628 set if a regular file was being unlinked, and it was the last link
6629 - use PtlGet from the target for bulk writes (315)
6630 - OST now updates object mtime with writes or setattr (607, 619)
6631 - LDLM now has a grant-time callback to revalidate locked items, if
6633 - Many MDS operations were reorganized to combat race conditions
6635 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6636 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6638 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6641 - Fully reactivate OST imports after reconnection (512, others)
6642 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6643 - More graceful error handling for truncating on dead OST (515)
6644 - Don't error out unless we're actually accessing dead stripes (474)
6645 - Fix garbage sizes when stripes are missing (410)
6646 - LRU counters were broken, causing constant lock purge (433, 432)
6647 - garbage on read from stripes with failed OSTs (441)
6648 - mark OSCs as active before reconnecting during recovery (438)
6649 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6650 - lfind did not preserve OST order in output (443)
6651 - symlinks cause hung clients, incorrect data (439)
6652 - stop dereferencing request after dropping refcount (457)
6653 - don't LASSERT(spin_is_locked) on non-SMP (455)
6654 - fixes for many rename() bugs
6655 - fstat didn't correctly synchronize attributes (399)
6656 - server must handle lock cancellation during blocking AST prep (487)
6657 - bulk descriptors were free()d too soon (511)
6658 - fix paths in lconf, which would load incorrect modules (451, 507)
6659 - fix confusing lconf 'host not found' error message (386)
6660 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6661 - fix race condition in mdc_blocking_ast() for inode access (526)
6662 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6663 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6664 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6665 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6666 - fix lstripe in several ways (kernel side) (527)
6667 - fix request leak in ldlm_cli_enqueue (262)
6668 - incorrect OSC was marked inactive after OST failure
6669 - call mds_fs_cleanup before unmounting filesystem (524)
6670 - fix races between taking ns_lock and ldlm_lock_change_resource
6671 - fix races updating LOV export open file list
6672 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6673 - fix recovery NULL deref in ldlm_cli_cancel_unused
6674 - fix some DLM races by using new hash table for lock handles (419)
6675 - permit the client to specify desired inodes, at replay
6676 - duplicate requests when we queue them for replay reintegration
6677 - fix last_rcvd offset calculation
6678 - sync after each recovered transaction, so we always make progress
6679 - never, not always, ERESTART requests without transnos
6680 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6682 - skip replay if the MDS says that the client is already connected
6683 - don't check for a recovery-enabled export to match lctl's UUID
6684 - don't INC_USE_COUNT for phantom exports
6685 - don't crash when cleaning up phantom exports (567)
6686 - don't double-finish or set replay data for errored mdc_open requests
6687 - abort requests when they time out, so we don't get old replies
6688 - send/receive replies for AST messages again
6689 - if the client says that it doesn't have the lock, cancel it on the
6691 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6692 mark it as destroyed, it will all work out when decref is called
6693 - fix module use counts (22, 581)
6695 - ASTs now expect a reply (server cancels lock on error reply)
6697 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6700 - fix many simultaneous client startup (392)
6701 - fix dentry->d_it clobbering
6702 - credentials weren't being shipped for readdir/getattr operations
6703 - remove invalid assertions triggered during some concurrent MD
6705 - proper Lustre versions added (336, 389)
6706 - fix memory leak for create error case (398)
6707 - fix LOV locking bug that would get cli/srv out of sync
6708 - fix echo client over LOV (409)
6709 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6710 - fix double-O_EXCL intent crash (424)
6711 - avoid sending multiple lock CANCELs (352)
6713 - MDS can do multi-client recovery (modulo bugs in new code)
6715 - many updates, edits, cleanups
6717 2002-11-18 Phil Schwan <phil@clusterfs.com>
6720 - fix null d_it dereference (346)
6721 - fix full OST/dbench hang (333)
6722 - fix permission problem with file removal (286)
6723 - fix removal of OSCs from LOV when they fail
6724 - fix NULL deref during bulk timeout (214)
6725 - fix problems related to multiple filesystems on one MDS (241)
6726 - fixed serious subtle metadata locking bugs
6727 - free locks on clients when inodes are removed due to memory
6729 - fix inode pointer in lock data (285)
6730 - partial support for multiple MDS on a single host (241)
6731 - data locks weren't cancelled at clear_inode time (290, 311)
6732 - intent locks could lead to unbounded lock growth (205)
6733 - added a maximum lock count, an LRU list, and a flusher
6734 - fix multiple rename (365)
6735 - properly abstracted the echo client
6736 - OSC locked 1 byte too many; fixed
6737 - rewrote brw callback code:
6738 - fixed recovery bugs related to LOVs (306)
6739 - fixed too-many-pages-in-one-write crash (191)
6740 - fixed (again) crash in sync_io_timeout (214)
6741 - probably fixed callback-related race (385)
6743 - Add capability to MDS protocol
6744 - LDLM cancellations and callbacks on different portals
6746 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6749 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6750 - defer open object destruction to close time (601981, 138)
6751 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6752 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6753 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6754 - don't call obd_setattr in ll_file_release for destroyed objects
6755 * protocol change to lustre_msg: move |version| and add |flags|
6756 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6757 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6758 * added replay of create, unlink, link and rename operations during
6759 MDS failover; recovery should be much more robust now
6760 * remove failed OSCs from LOVs (only lov_create uses this so far)
6761 * the lustre-HOWTO was brought (more) up to date (582544)
6763 2002-10-23 Phil Schwan <phil@clusterfs.com>
6766 - in-use dentries weren't being reused properly (617851)
6767 - prevent multiple LDLM setup (599178)
6768 - fix LOV size calculations for truncate (617853)
6769 - fix client handling of MDS intent errors (POSIX)
6770 - fix permission bug in lovstripe.c test (624321)
6771 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6772 - truncate past end of file could corrupt data
6773 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6774 - a race in recovery could return ETIMEDOUT to apps (623947)
6775 - building outside the source directory was fixed
6776 * the lustre-HOWTO was brought (more) up to date (582544)
6777 * major progress was made on recovery functionality
6779 2002-10-10 Phil Schwan <phil@clusterfs.com>
6782 - recovery deadlock fix
6783 - rm -rf causes LBUG fix (617817)
6784 - file open by multiple tasks fix (618962)
6785 - directory permissions bugs (602707 and 620007)
6786 - journal_stop fixed with locking (611313)
6787 - O_APPEND failures resolved (618273, perhaps 614459)
6788 - lconf PATH fix (619770)
6789 - IA64 build fix (621450)
6790 - RPC buffer sizes scale with amount of memory
6792 2002-10-01 Phil Schwan <phil@clusterfs.com>
6795 - locks would be cancelled without throwing away data pages,
6796 resulting in inconsistent data (605627)
6797 - inode attributes were not always being refreshed (605627, 612449)
6798 - lconf now continues to cleanup after lctl reports an error
6799 - MDS now enforces user permissions (602707)
6800 - lprocfs cleanup fixed, but not yet enabled (614157)
6801 - fixed infinite server hang, should a client not respond to an AST
6802 - avoid going into recovery if user calls readlink() with a buffer
6803 that's too small (613941)
6804 - AST RPCs no longer require replies (614867) -- this may be changed
6805 - don't crash server if client sends an IOV that's too big (611336)
6806 - fixed lock conversion deadlock (611892)
6807 - fixed the following of symlinks (614622)
6808 * recovery: the server can remove locks from a client that dies, other
6809 clients can make progress
6810 * more extN patch fixes
6811 * compile-time configurable ptlrpc buffer allocations
6813 - collaborative read cache document
6814 - Lustre Lite Performance CDR document-in-progress
6816 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6819 - fix typo in patch-2.4.18
6821 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6824 - clear ptlrpc request each time in handle_incoming_request()
6825 - unlink of files now destroys the object on the OST
6827 2002-09-19 Peter Braam <braam@clusterfs.com>
6829 * add hard link support
6830 * change obdfile creation method
6831 * kernel patch changed
6833 2002-09-19 Peter Braam <braam@clusterfs.com>
6836 - stack overflow bug in extN fixed
6838 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6840 * documentation updates
6841 - add man pages for config tools
6842 - update tests/README to describe testing with new config tools
6843 - finish metadata API descriptions
6844 * bug fixes and cleanups
6845 - statfs workaround for 16TB limit
6846 - LOV stripe allocation improved, can stripe on subset of OSTs
6847 - LOV file size/IO offset was wrong for files > 4GB in size
6848 - object EA data was being dropped, caused files to be unreadable
6849 - memory overflow with non-LOV OST caused memory corruption
6850 - fixed regression tests to work with new config tools, obdfilter
6851 - fixed bug when directory size became larger than 1 block
6852 - fixed bug (for single client case) when PWD was deleted
6853 - invalidate local directory pages when doing intent-based ops
6854 - avoid LDLM oops when lock callback contained bad data
6856 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6858 * documentation updates
6859 * bug fixes and cleanups
6860 - configuration tools
6863 - 64-bit compile warnings
6864 - 64-bit internal statfs data
6866 * test_brw on persistent OST devices
6868 * lprocfs (disabled)
6870 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6872 * documentation updates
6873 * bug fixes and cleanups
6874 * configuration tools
6876 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6879 * many small fixes to 0_5_4
6880 * io/network handling
6881 * thinkos in MDS operations
6883 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6886 * crucial basic fixes to 0.5.3
6887 * IOR, Iozone work over Elan
6890 2002-08-07 Phil Schwan <phil@clusterfs.com>
6891 * version 0_5_3, our first alpha
6892 * we use the new Portals iovs
6893 * documentation updates
6894 * bug fixes and cleanups
6895 * small changes in the DLM wire protocol
6897 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6898 * version 0_5_1 with some initial stability,
6899 * locking on MD and file I/O.
6900 * documentation updates
6901 * several bug fixes since 0.5.0
6902 * small changes in wire protocol
6904 2002-07-18 Phil Schwan <phil@clusterfs.com>
6906 * delivered as Lustre Light Alpha
6907 * fixed a crash after handling invalid MDS requests
6908 * fixed directory pages for architectures with non-4k pages sizes
6910 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6911 * release version v0_4_4
6912 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6914 * Fixes a number of interruption problems with OST operations.
6915 * Update documentation for portals header changes
6916 * Move all wire protocol structs/defines to lustre_idl.h
6917 * Fixes symlink length bug.
6918 * Add tcpdump to repository.
6920 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6921 * release version v0_4_3
6922 * Fixes statfs for inodes on extN.
6923 * Fixes bug in runtests which would delete /etc/hosts.
6924 * Use 64-bit object IDs wherever possible (not into VFS though)
6925 Remove ost_get_info, which is unused by lustre, and out of date.
6927 2002-07-03 Peter Braam <braam@clusterfs.com>
6928 * release version v0_4_2 Fixes a lookup error (type not passed)
6929 * move forward to head of Portals
6930 * move forward to latest Lustre kernel
6932 2002-06-25 Peter Braam <braam@clusterfs.com>
6933 * release version v0_4_1. Hopefully stable on single node use.