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 3.10.0-1127.8.2.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 3.10.0-1127.8.2.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 4.18.0-193.6.3.el8 (RHEL8.2)
37 4.4.120-92.70 (SLES12 SP2)
38 4.4.0-142 (Ubuntu 16.04)
39 vanilla linux 4.15.0 (ZFS only)
40 * Recommended e2fsprogs version: 1.45.6.wc1 or newer
41 * Recommended ZFS / SPL version: 0.7.13
42 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
43 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
44 stack overflow. For more information, please refer to bugzilla 17630.
45 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
46 the re-exporting nodes to be mounted with "32bitapi" mount option
47 * Note for Ubuntu 14.04.5:
48 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
49 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
53 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
54 for currently supported client and server kernel versions.
55 * Server primary kernels built and tested during release cycle:
56 3.10.0-957.27.2.el7 (RHEL7.6)
57 3.10.0-1062.18.1.el7 (RHEL7.7)
58 3.10.0-1127.8.2.el7 (RHEL7.8)
59 * Server known to build on the following kernels (others may also work):
60 2.6.32-696.18.7.el6 (RHEL6.9)
61 3.10.0-862.14.4.el7 (RHEL7.5)
62 4.4.120-92.70 (SLES12 SP2)
63 4.4.180-94.97 (SLES12 SP3)
64 4.4.0-142 (Ubuntu 16.04)
65 4.15.0-45 (Ubuntu 18.04)
66 vanilla linux 4.6.7 (ZFS only)
67 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
68 * Client primary unpatched kernels built and tested during release cycle:
69 3.10.0-957.27.2.el7 (RHEL7.6)
70 3.10.0-1062.18.1.el7 (RHEL7.7)
71 3.10.0-1127.8.2.el7 (RHEL7.8)
72 4.18.0-80.11.2.el8 (RHEL8.0)
73 4.4.180-94.97 (SLES12 SP3)
74 4.12.14-95.45 (SLES12 SP4)
75 4.12.14-122.20 (SLES12 SP5)
76 4.15.0-45 (Ubuntu 18.04)
77 * Client known to build on unpatched kernels (others may also work):
78 2.6.32-696.18.7.el6 (RHEL6.9)
79 2.6.32-754.6.3.el6 (RHEL6.10)
80 3.10.0-862.14.4.el7 (RHEL7.5)
81 4.14.0-49.13.1.el7a (RHEL7.5)
82 4.14.0-115.2.2.el7a (RHEL7.6)
83 4.18.0-147.8.1.el8 (RHEL8.1)
84 4.18.0-193.1.2.el8 (RHEL8.2)
85 4.4.120-92.70 (SLES12 SP2)
86 4.4.0-142 (Ubuntu 16.04)
87 vanilla linux 4.15.0 (ZFS only)
88 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
89 * Recommended ZFS / SPL version: 0.7.13
90 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
91 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
92 stack overflow. For more information, please refer to bugzilla 17630.
93 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
94 the re-exporting nodes to be mounted with "32bitapi" mount option
95 * Note for Ubuntu 14.04.5:
96 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
97 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
101 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
102 for currently supported client and server kernel versions.
103 * Server primary kernels built and tested during release cycle:
104 3.10.0-957.27.2.el7 (RHEL7.6)
105 3.10.0-1062.9.1.el7 (RHEL7.7)
106 * Server known to build on the following kernels (others may also work):
107 2.6.32-696.18.7.el6 (RHEL6.9)
108 3.10.0-862.14.4.el7 (RHEL7.5)
109 4.4.120-92.70 (SLES12 SP2)
110 4.4.180-94.97 (SLES12 SP3)
111 4.4.0-142 (Ubuntu 16.04)
112 4.15.0-45 (Ubuntu 18.04)
113 vanilla linux 4.6.7 (ZFS only)
114 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
115 * Client primary unpatched kernels built and tested during release cycle:
116 3.10.0-957.27.2.el7 (RHEL7.6)
117 3.10.0-1062.9.1.el7 (RHEL7.7)
118 4.18.0-80.11.2.el8 (RHEL8.0)
119 4.4.180-94.97 (SLES12 SP3)
120 4.12.14-95.29 (SLES12 SP4)
121 4.15.0-45 (Ubuntu 18.04)
122 * Client known to build on unpatched kernels (others may also work):
123 2.6.32-696.18.7.el6 (RHEL6.9)
124 2.6.32-754.6.3.el6 (RHEL6.10)
125 3.10.0-862.14.4.el7 (RHEL7.5)
126 4.14.0-49.13.1.el7a (RHEL7.5)
127 4.14.0-115.2.2.el7a (RHEL7.6)
128 4.18.0-147.3.1.el8 (RHEL8.1)
129 4.4.120-92.70 (SLES12 SP2)
130 4.4.0-142 (Ubuntu 16.04)
131 vanilla linux 4.15.0 (ZFS only)
132 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
133 * Recommended ZFS / SPL version: 0.7.13
134 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
135 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
136 stack overflow. For more information, please refer to bugzilla 17630.
137 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
138 the re-exporting nodes to be mounted with "32bitapi" mount option
139 * Note for Ubuntu 14.04.5:
140 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
141 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
145 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
146 for currently supported client and server kernel versions.
147 * Server primary kernels built and tested during release cycle:
148 3.10.0-957.27.2.el7 (RHEL7.6)
149 3.10.0-1062.1.1.el7 (RHEL7.7)
150 * Server known to build on the following kernels (others may also work):
151 2.6.32-696.18.7.el6 (RHEL6.9)
152 3.10.0-862.14.4.el7 (RHEL7.5)
153 4.4.120-92.70 (SLES12 SP2)
154 4.4.180-94.97 (SLES12 SP3)
155 4.4.0-142 (Ubuntu 16.04)
156 4.15.0-45 (Ubuntu 18.04)
157 vanilla linux 4.6.7 (ZFS only)
158 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
159 * Client primary unpatched kernels built and tested during release cycle:
160 3.10.0-957.27.2.el7 (RHEL7.6)
161 3.10.0-1062.1.1.el7 (RHEL7.7)
162 4.18.0-80.7.1.el8 (RHEL8.0)
163 4.4.180-94.97 (SLES12 SP3)
164 4.12.14-95.29 (SLES12 SP4)
165 4.15.0-45 (Ubuntu 18.04)
166 * Client known to build on unpatched kernels (others may also work):
167 2.6.32-696.18.7.el6 (RHEL6.9)
168 2.6.32-754.6.3.el6 (RHEL6.10)
169 3.10.0-862.14.4.el7 (RHEL7.5)
170 4.14.0-49.13.1.el7a (RHEL7.5)
171 4.14.0-115.2.2.el7a (RHEL7.6)
172 4.4.120-92.70 (SLES12 SP2)
173 4.4.0-142 (Ubuntu 16.04)
174 vanilla linux 4.15.0 (ZFS only)
175 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
176 * Recommended ZFS / SPL version: 0.7.13
177 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
178 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
179 stack overflow. For more information, please refer to bugzilla 17630.
180 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
181 the re-exporting nodes to be mounted with "32bitapi" mount option
182 * Note for Ubuntu 14.04.5:
183 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
184 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
188 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
189 for currently supported client and server kernel versions.
190 * Server known to build on the following kernels (others may also work):
191 2.6.32-696.18.7.el6 (RHEL6.9)
192 3.10.0-862.14.4.el7 (RHEL7.5)
193 3.10.0-957.10.1.el7 (RHEL7.6)
194 4.4.120-92.70 (SLES12 SP2)
195 4.4.162-94.69 (SLES12 SP3)
196 4.4.0-142 (Ubuntu 16.04)
197 4.15.0-45 (Ubuntu 18.04)
198 vanilla linux 4.6.7 (ZFS only)
199 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
200 * Client known to build on unpatched kernels (others may also work):
201 2.6.32-696.18.7.el6 (RHEL6.9)
202 2.6.32-754.6.3.el6 (RHEL6.10)
203 3.10.0-862.14.4.el7 (RHEL7.5)
204 3.10.0-957.10.1.el7 (RHEL7.6)
205 4.14.0-49.13.1.el7a (RHEL7.5)
206 4.14.0-115.2.2.el7a (RHEL7.6)
207 4.4.120-92.70 (SLES12 SP2)
208 4.4.162-94.69 (SLES12 SP3)
209 4.12.14-95.13 (SLES12 SP4)
210 4.4.0-142 (Ubuntu 16.04)
211 4.15.0-45 (Ubuntu 18.04)
212 vanilla linux 4.15.0 (ZFS only)
213 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
214 * Recommended ZFS / SPL version: 0.7.13
215 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
216 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
217 stack overflow. For more information, please refer to bugzilla 17630.
218 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
219 the re-exporting nodes to be mounted with "32bitapi" mount option
220 * Note for Ubuntu 14.04.5:
221 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
222 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
227 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
228 for currently supported client and server kernel versions.
229 * Server known to build on the following kernels (others may also work):
230 2.6.32-696.18.7.el6 (RHEL6.9)
231 3.10.0-862.14.4.el7 (RHEL7.5)
232 3.10.0-957.10.1.el7 (RHEL7.6)
233 4.4.120-92.70 (SLES12 SP2)
234 4.4.162-94.69 (SLES12 SP3)
235 4.4.0-142 (Ubuntu 16.04)
236 4.15.0-45 (Ubuntu 18.04)
237 vanilla linux 4.6.7 (ZFS only)
238 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
239 * Client known to build on unpatched kernels (others may also work):
240 2.6.32-696.18.7.el6 (RHEL6.9)
241 2.6.32-754.6.3.el6 (RHEL6.10)
242 3.10.0-862.14.4.el7 (RHEL7.5)
243 3.10.0-957.10.1.el7 (RHEL7.6)
244 4.14.0-49.13.1.el7a (RHEL7.5)
245 4.14.0-115.2.2.el7a (RHEL7.6)
246 4.4.120-92.70 (SLES12 SP2)
247 4.4.162-94.69 (SLES12 SP3)
248 4.12.14-95.13 (SLES12 SP4)
249 4.4.0-142 (Ubuntu 16.04)
250 4.15.0-45 (Ubuntu 18.04)
251 vanilla linux 4.15.0 (ZFS only)
252 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
253 * Recommended ZFS / SPL version: 0.7.13
254 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
255 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
256 stack overflow. For more information, please refer to bugzilla 17630.
257 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
258 the re-exporting nodes to be mounted with "32bitapi" mount option
259 * Note for Ubuntu 14.04.5:
260 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
261 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
265 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
266 for currently supported client and server kernel versions.
267 * Server known to build on the following kernels (others may also work):
268 2.6.32-696.18.7.el6 (RHEL6.9)
269 3.10.0-862.14.4.el7 (RHEL7.5)
270 3.10.0-957.el7 (RHEL7.6)
271 4.4.120-92.70 (SLES12 SP2)
272 4.4.162-94.69 (SLES12 SP3)
273 4.4.0-131 (Ubuntu 16.04)
274 4.15.0-32 (Ubuntu 18.04)
275 vanilla linux 4.6.7 (ZFS only)
276 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
277 * Client known to build on unpatched kernels (others may also work):
278 2.6.32-696.18.7.el6 (RHEL6.9)
279 2.6.32-754.6.3.el6 (RHEL6.10)
280 3.10.0-862.14.4.el7 (RHEL7.5)
281 3.10.0-957.el7 (RHEL7.6)
282 4.14.0-49.13.1.el7a (RHEL7.5)
283 4.14.0-115.2.2.el7a (RHEL7.6)
284 4.4.120-92.70 (SLES12 SP2)
285 4.4.162-94.69 (SLES12 SP3)
286 4.4.0-131 (Ubuntu 16.04)
287 4.15.0-32 (Ubuntu 18.04)
288 vanilla linux 4.15.0 (ZFS only)
289 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
290 * Recommended ZFS / SPL version: 0.7.9
291 * Tested with ZFS / SPL version: 0.7.9
292 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
293 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
294 stack overflow. For more information, please refer to bugzilla 17630.
295 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
296 the re-exporting nodes to be mounted with "32bitapi" mount option
297 * Note for Ubuntu 14.04.5:
298 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
299 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
301 --------------------------------------------------------------------------------
304 2018-04-03 Intel Corporation
306 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
307 for currently supported client and server kernel versions.
308 * Server known to build on patched kernels:
309 2.6.32-431.29.2.el6 (RHEL6.5)
310 2.6.32-504.30.3.el6 (RHEL6.6)
311 2.6.32-573.26.1.el6 (RHEL6.7)
312 2.6.32-642.15.1.el6 (RHEL6.8)
313 2.6.32-696.18.7.el6 (RHEL6.9)
314 3.10.0-693.21.1.el7 (RHEL7.4)
315 3.0.101-0.47.71 (SLES11 SP3)
316 3.0.101-107 (SLES11 SP4)
317 3.12.74-60.64.40 (SLES12 SP1)
318 4.4.114-92.64 (SLES12 SP2)
319 4.4.114-94.11 (SLES12 SP3)
320 3.13.0-101 (Ubuntu 14.04, ZFS only)
321 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
322 4.4.0-116 (Ubuntu 16.04)
323 vanilla linux 4.6.7 (ZFS only)
324 * Client known to build on unpatched kernels:
325 2.6.32-431.29.2.el6 (RHEL6.5)
326 2.6.32-504.30.3.el6 (RHEL6.6)
327 2.6.32-573.26.1.el6 (RHEL6.7)
328 2.6.32-642.15.1.el6 (RHEL6.8)
329 2.6.32-696.18.7.el6 (RHEL6.9)
330 3.10.0-693.21.1.el7 (RHEL7.4)
331 3.0.101-0.47.71 (SLES11 SP3)
332 3.0.101-107 (SLES11 SP4)
333 3.12.74-60.64.40 (SLES12 SP1)
334 4.4.114-92.64 (SLES12 SP2)
335 4.4.114-94.11 (SLES12 SP3)
336 3.13.0-101 (Ubuntu 14.04)
337 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
338 4.4.0-116 (Ubuntu 16.04)
340 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
341 * Recommended ZFS / SPL version: 0.7.6
342 * Tested with ZFS / SPL version: 0.7.6
343 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
344 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
345 stack overflow. For more information, please refer to bugzilla 17630.
346 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
347 the re-exporting nodes to be mounted with "32bitapi" mount option
348 * Note for Ubuntu 14.04.5:
349 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
350 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
352 --------------------------------------------------------------------------------
354 2017-07-05 Intel Corporation
356 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
357 for currently supported client and server kernel versions.
358 * Server known to build on patched kernels:
359 2.6.32-431.29.2.el6 (RHEL6.5)
360 2.6.32-504.30.3.el6 (RHEL6.6)
361 2.6.32-573.26.1.el6 (RHEL6.7)
362 2.6.32-642.15.1.el6 (RHEL6.8)
363 2.6.32-696.3.1.el6 (RHEL6.9)
364 3.10.0-514.21.1.el7 (RHEL7.3)
365 3.0.101-0.47.71 (SLES11 SP3)
366 3.0.101-100 (SLES11 SP4)
367 3.12.74-60.64.40 (SLES12 SP1)
368 4.4.59-92.17 (SLES12 SP2)
369 vanilla linux 4.6.7 (ZFS only)
370 * Client known to build on unpatched kernels:
371 2.6.32-431.29.2.el6 (RHEL6.5)
372 2.6.32-504.30.3.el6 (RHEL6.6)
373 2.6.32-573.26.1.el6 (RHEL6.7)
374 2.6.32-642.15.1.el6 (RHEL6.8)
375 2.6.32-696.3.1.el6 (RHEL6.9)
376 3.10.0-514.21.1.el7 (RHEL7.3)
377 3.0.101-0.47.71 (SLES11 SP3)
378 3.0.101-100 (SLES11 SP4)
379 3.12.74-60.64.40 (SLES12 SP1)
380 4.4.59-92.17 (SLES12 SP2)
382 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
383 * Recommended ZFS / SPL version: 0.6.5.9
384 * Tested with ZFS / SPL version: 0.6.5.9
385 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
386 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
387 stack overflow. For more information, please refer to bugzilla 17630.
388 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
389 the re-exporting nodes to be mounted with "32bitapi" mount option
391 --------------------------------------------------------------------------------
393 12-07-2016 Intel Corporation
395 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
396 for currently supported client and server kernel versions.
397 * Server known to build on patched kernels:
398 2.6.32-431.29.2.el6 (RHEL6.5)
399 2.6.32-504.30.3.el6 (RHEL6.6)
400 2.6.32-573.26.1.el6 (RHEL6.7)
401 2.6.32-642.11.1.el6 (RHEL6.8)
402 3.10.0-514.el7 (RHEL7.3)
403 3.0.101-0.47.71 (SLES11 SP3)
404 3.0.101-80 (SLES11 SP4)
405 3.12.62-60.64.8 (SLES12 SP1)
406 vanilla linux 4.5.7 (ZFS only)
407 * Client known to build on unpatched kernels:
408 2.6.32-431.29.2.el6 (RHEL6.5)
409 2.6.32-504.30.3.el6 (RHEL6.6)
410 2.6.32-573.26.1.el6 (RHEL6.7)
411 2.6.32-642.11.1.el6 (RHEL6.8)
412 3.10.0-514.el7 (RHEL7.3)
413 3.0.101-0.47.71 (SLES11 SP3)
414 3.0.101-80 (SLES11 SP4)
415 3.12.62-60.64.8 (SLES12 SP1)
417 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
418 * Recommended ZFS / SPL version: 0.6.5.8
419 * Tested with ZFS / SPL version: 0.6.5.8
420 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
421 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
422 stack overflow. For more information, please refer to bugzilla 17630.
423 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
424 the re-exporting nodes to be mounted with "32bitapi" mount option
426 --------------------------------------------------------------------------------
428 02-29-2016 Intel Corporation
430 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
431 for currently supported client and server kernel versions.
432 * Server known to build on patched kernels:
433 2.6.32-431.29.2.el6 (RHEL6.5)
434 2.6.32-504.30.3.el6 (RHEL6.6)
435 2.6.32-573.12.1.el6 (RHEL6.7)
436 3.10.0-327.3.1.el7 (RHEL7.2)
437 3.0.101-0.47.71 (SLES11 SP3)
438 3.0.101-68 (SLES11 SP4)
439 vanilla linux 4.2.1 (ZFS only)
440 * Client known to build on unpatched kernels:
441 2.6.32-431.29.2.el6 (RHEL6.5)
442 2.6.32-504.30.3.el6 (RHEL6.6)
443 2.6.32-573.12.1.el6 (RHEL6.7)
444 3.10.0-327.3.1.el7 (RHEL7.2)
445 3.0.101-0.47.71 (SLES11 SP3)
446 3.0.101-68 (SLES11 SP4)
449 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
450 * Recommended ZFS / SPL version: 0.6.4.2
451 * Tested with ZFS / SPL version: 0.6.4.2
452 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
453 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
454 stack overflow. For more information, please refer to bugzilla 17630.
455 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
456 the re-exporting nodes to be mounted with "32bitapi" mount option
458 --------------------------------------------------------------------------------
460 03-10-2015 Intel Corporation
462 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
463 for currently supported client and server kernel versions.
464 * Server known to build on patched kernels:
465 2.6.32-431.29.2.el6 (RHEL6.5)
466 2.6.32-504.8.1.el6 (RHEL6.6)
467 3.0.101-0.46 (SLES11 SP3)
468 * Client known to build on unpatched kernels:
469 2.6.32-431.29.2.el6 (RHEL6.5)
470 2.6.32-504.8.1.el6 (RHEL6.6)
471 3.10.0-123.20.1.el7 (RHEL7)
472 3.0.101-0.46 (SLES11 SP3)
473 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
474 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
475 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
476 stack overflow. For more information, please refer to bugzilla 17630.
477 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
478 the re-exporting nodes to be mounted with "32bitapi" mount option
480 Severity : enhancement
482 Description: control OST-index in IDIF via ROCOMPAT flag.
483 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
484 last_rcvd file. For new formatted OST device, it will be auto set;
485 for the case of upgrading from old OST device, you can enable it
486 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
487 enabled, for new created OST-object, its IDIF-in-LMA will contain
488 the OST-index; for the existing OST-object, the OSD will convert
489 old format IDIF as new format IDIF with OST-index stored in the
490 LMA EA when accessing such OST-object or via OI scrub. Once such
491 flag is enabled, it cannot be reverted back, so the system cannot
492 be downgraded to the orignal incompatible version.
494 --------------------------------------------------------------------------------
496 07-30-2014 Intel Corporation
498 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
499 for currently supported client and server kernel versions.
500 * Server known to build on patched kernels:
501 2.6.32-431.20.3.el6 (RHEL6.5)
502 3.0.101-0.31 (SLES11 SP3)
503 * Client known to build on unpatched kernels:
504 2.6.32-431.20.3.el6 (RHEL6.5)
505 3.0.101-0.31 (SLES11 SP3)
508 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
509 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
510 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
511 stack overflow. For more information, please refer to bugzilla 17630.
512 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
513 the re-exporting nodes to be mounted with "32bitapi" mount option
515 --------------------------------------------------------------------------------
517 10-31-2013 Intel Corporation
519 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
520 * Server builds for kernels:
521 2.6.32-358.18.1.el6 (RHEL6.4)
522 * Client builds for unpatched kernels:
523 2.6.32-358.18.1.el6 (RHEL6.4)
524 3.0.92-0.8 (SLES11 SP3)
526 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
527 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
528 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
529 stack overflow. For more information, please refer to bugzilla 17630.
530 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
531 the re-exporting nodes to be mounted with "32bitapi" mount option
533 --------------------------------------------------------------------------------
535 05-30-2013 Intel Corporation
537 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
538 * Server builds for kernels:
539 2.6.32-358.6.2.el6 (RHEL6.4)
540 * Client builds for unpatched kernels:
541 2.6.32-358.6.2.el6 (RHEL6.4)
542 3.0.74-0.6.6 (SLES11 SP2)
544 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
545 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
546 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
547 stack overflow. For more information, please refer to bugzilla 17630.
548 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
549 the re-exporting nodes to be mounted with "32bitapi" mount option
551 --------------------------------------------------------------------------------
553 9-30-2012 Whamcloud, Inc.
555 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
556 * Server support for kernels:
557 2.6.32-279.5.1.el6 (RHEL6)
558 * Client support for unpatched kernels:
559 2.6.32-279.5.1.el6 (RHEL6)
561 2.6.18-238.19.1.el5 (RHEL5)
562 2.6.16.60-0.69.1 (SLES10)
563 * Recommended e2fsprogs version: 1.41.90.wc4
564 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
565 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
566 stack overflow. For more information, please refer to bugzilla 17630.
567 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
568 the re-exporting nodes to be mounted with "32bitapi" mount option
570 Severity : enhancement
572 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
574 Severity : enhancement
576 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
578 Severity : enhancement
580 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
582 Severity : enhancement
584 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
586 Severity : enhancement
588 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
590 --------------------------------------------------------------------------------
592 04-30-2012 Whamcloud, Inc.
594 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
595 * Server support for kernels:
596 2.6.32-220.4.2.el6 (RHEL6)
597 * Client support for unpatched kernels:
598 2.6.32-220.4.2.el6 (RHEL6)
600 2.6.18-238.19.1.el5 (RHEL5)
601 2.6.16.60-0.69.1 (SLES10)
602 * Recommended e2fsprogs version: 1.41.90.wc4
603 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
604 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
605 stack overflow. For more information, please refer to bugzilla 17630.
606 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
607 the re-exporting nodes to be mounted with "32bitapi" mount option
609 Severity : enhancement
611 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
613 Severity : enhancement
615 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
617 Severity : enhancement
619 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
621 Severity : enhancement
623 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
625 --------------------------------------------------------------------------------
627 09-30-2011 Whamcloud, Inc.
629 * Server support for kernels:
630 2.6.18-238.19.1.el5 (RHEL5)
631 2.6.32-131.6.1.el6 (RHEL6)
632 * Client support for unpatched kernels:
633 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
634 2.6.18-238.19.1.el5 (RHEL5)
635 2.6.32-131.6.1.el6 (RHEL6)
636 2.6.29.4-167.fc11 (FC11)
637 2.6.16.60-0.69.1 (SLES10)
639 * Recommended e2fsprogs version: 1.41.90.wc3
640 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
641 Lustre filesystem with 4K stack may cause a stack overflow. For more
642 information, please refer to bugzilla 17630.
643 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
644 mounted with "32bitapi" mount option
646 Severity : enhancement
647 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
649 Severity : enhancement
650 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
652 Severity : enhancement
653 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
656 Frequency : only with SLES10
658 Description: Use OFED "KMP" provided by Novell
659 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
660 Lustre is now built against this package. That means you need to
661 install the ofed-kmp package from Novell for the patchless client.
662 Note that the ofed-kmp that Novell ships may not exactly match the
663 kernel version but should still be compatible.
665 Severity : enhancement
667 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
668 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
670 --------------------------------------------------------------------------------
672 2010-07-15 Oracle, Inc.
674 * Server support for kernels:
675 2.6.18-164.11.1.el5 (RHEL 5)
676 2.6.18-164.11.1.0.1.el5 (OEL 5)
677 * Client support for unpatched kernels:
678 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
679 2.6.18-164.11.1.el5 (RHEL 5),
680 2.6.18-164.11.1.0.1.el5 (OEL 5)
681 2.6.16.60-0.42.8 (SLES 10),
683 2.6.29.4-167.fc11 (FC11)
684 * Recommended e2fsprogs version: 1.41.10-sun2
686 Severity : enhancement
688 Description: Distribute lfs_migrate script with lustre
689 Details : The lfs_migrate mentioned in the manual is now included in the
690 lustre tree in lustre/scripts and will distributed in lustre RPMs.
692 Severity : enhancement
694 Description: Failover nid list added to import proc file.
696 Severity : enhancement
698 Description: "som_preview" mount option required to use SOM
699 Details : The clients will need to use som_preview mount option
703 Description: Lustre does not do 1MB IOs to HW RAID
704 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
706 Severity : enhancement
708 Description: lfs quota output is not very convenient for awk/sed-parsing
709 Details : Some positions in lfs quota output table could be empty or
710 non-empty which made it hard to parse it with scripts, now
711 a dash is put instead of space where there is not supposed
716 Description: Update to OFED 1.5.1
717 Details : OFED 1.5.1 will be used in environments that don't use
718 vendor supplied OFED.
722 Description: Remove lnet. prefix from lctl params display
723 Details : Also renamed lnet.memused to lnet_memused to
728 Description: Use conf_param -d to remove permanant settings.
732 Description: Add list_param -R
733 Details : lctl list_param now supports recursive mode.
736 Bugzilla : 21829, 22312
737 Description: Fix llobdstat.
738 Details : 1. Now updating the write_bytes counter
739 2. Port script fixes from 1.8.
743 Description: decrease the usage of memory on clients.
744 Details : 1. On clients, recycle dentries and inodes unused.
745 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
750 Description: llobdstat fix and enhancement
751 Details : add a counter to set a limit to how many samples will be returned
752 fix a wildcard in the path to limit to obdfilter stats only
756 Description: lfs check only allowed for root
757 Details : Code cleanups and added a regresion test of non-root use of
762 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
763 Details : Fix hr->hr_index overload.
767 Description: Fix FIEMAP functionality.
771 Description: FC11 patchless client support added
775 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
776 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
780 Description: New acc-sm script and various test-framework enhancments.
784 Description: Previously landed patch was reverted.
788 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
792 Description: Add ldiskfs maxdirsize mount option.
796 Description: Fix mkfs.lustre support for 16TB LUNs.
800 Description: Print mmp_check_interval
801 Details : print mmp_check_interval and make it possible to abort
802 mount operation in case it takes too long.
806 Description: Add cascading_rw to lustre mpi tests.
810 Description: Store SOM data in EA.
811 Details : Also several SOM related cleanups.
815 Description: Add refcount to md_open_data
819 Description: Use refcount on lov_request_set to prevent memory corruption.
823 Description: Statahead: use parent dir's i_mutex for synchronization.
827 Description: loadgen improvements.
831 Description: Enable peer health detection in ksocklnd and ko2iblnd.
835 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
839 Description: Fix race on obd_devcie:md_stats
843 Description: Fix SLES 11 client builds.
844 Details : Resolved uninitalized pointer gcc warnings.
848 Description: Allow non-root access for "lfs check".
849 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
853 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
855 Severity : enhancement
857 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
858 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
862 Description: Sanity test_27p fails.
863 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
864 Sanity tests 27 were cleaned up and improved.
868 Description: Unify req format on client/servers
869 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
870 buffers and swabbers approach..
875 Description: drop join file stuff.
877 Severity : enhancement
879 Description: Add 2.6.27 and SLES11 (patchless client) support
883 Description: Fix in ptlrpc_expire_one_request() to print the signed time
886 Severity : enhancement
888 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
891 Severity : enhancement
893 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
895 Severity : enhancement
897 Description: correctly handle big reply message.
898 Details : send LNet event if reply is bigger than buffer and adjust this buffer
903 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
905 Severity : enhancement
907 Description: Add support for OEL5.
911 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
913 Severity : enhancement
915 Description: "weak-modules" support
916 Details : Implement "weak-modules" support which enables kernel modules
917 to be used with any kernel that implements the same kABI. In
918 order to achieve this modules are now installed in
919 /lib/modules/$(uname -r)/updates/kernel on all distributions.
923 Description: client could not reconnect to OST because of an active request.
924 Details : The client now retries to reconnect to the same server, if a
925 connect request failed with EBUSY or -EAGAIN. Test case 35b
926 conf-sanity.sh has been created.
930 Description: Don't allow make backward step on assignin osc next id.
931 Details : race between allocation next id and ll_sync thread can be cause
932 of set wrong osc next id and can be kill valid ost objects.
936 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
940 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
944 Description: Error handling in osc_statfs_interpret() has been improved.
945 Details : Check in osc_statfs_interpret() for EBADR.
949 Description: Do not send statfs() requests to OSTs disabled by administrator.
950 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
952 Severity : enhancement
954 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
956 Severity : enhancement
958 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
960 Severity : enhancement
962 Description: Enhancement for lfs(1) command to use numeric uid/gid.
966 Description: Improved cleanup procedure for conf-sanity test_46a.
970 Description: Do not start bulk transfer, if client tried to reconnect.
971 Details : Do not start bulk transfer, if client tried to reconnect,
972 but failed due to active rpcs. Check for exp_abort_active_req
973 has been implemented.
977 Description: Prevent inconsistences between linux and lustre mount structures.
978 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
979 Make the sleep interruptible.
981 Severity : enhancement
983 Description: provide server to client comms path
984 Details : server to client communications path via new LDLM_SET_INFO rpc
988 Description: truncate starts GFP_FS allocation under transaction causing deadlock
989 Details : ldiskfs_truncate calls grab_cache_page which may start page
990 allocation under an open transaction. This may lead to
991 calling prune_icache with consequent lustre reentrance.
995 Description: send correctly lsm on open replay
996 Details : MDS is trust to LSM size on replay open, but client can set wrong size
1001 Description: enable client interface failover
1002 Details : When a child reconnects from another NID, properly update export
1003 nid hash position and ldlm reverse import.
1005 Severity : enhancement
1007 Description: Communicate OST degraded/readonly state via statfs to MDS
1008 Details : Flags in the statfs returned from OSTs indicate whether the
1009 OST is in a degraded RAID state, or if the filesystem has
1010 turned read-only after a filesystem error is detected.
1015 Description: don't panic if EPROTO was hit when reading symlink
1016 Details : correctly handling request reference in error cases.
1018 Severity : enhancement
1020 Description: Add LustreNetLink, a kernel-userspace communcation path.
1022 Severity : enhancement
1024 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1029 Description: access to llog context before init.
1030 Details : move handling CATALOGS file at osc layer and forbid access to llog
1031 context before init.
1034 Frequency : rare, only when client is writing to OST on local node
1036 Description: Avoid deadlock for local client writes
1037 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1038 memory freeing context. This allows OST threads to set the
1039 PF_MEMALLOC flag on task structures in order to allocate memory
1040 from reserved pools and complete IO.
1041 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1042 so that the OST threads generate memory pressure and allow
1043 inactive pages to be reclaimed.
1045 Severity : enhancement
1047 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1052 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1053 Details : move ldlm namespace creation in setup phase to avoid grab
1054 _lprocfs_lock with cli_sem held.
1058 Description: Temporarily disable grant shrink.
1059 Details : Disable the feature for debugging.
1062 Frequency : only when formatting test filesystems
1064 Description: Unable to run several mkfs.lustre on loop devices at the same time
1065 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1066 formatting. The solution is to proper handle the error.
1068 Severity : enhancement
1070 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1072 Severity : enhancement
1074 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1079 Description: resolve race between obd_disconnect and class_disconnect_exports
1080 Details : if obd_disconnect will be called to already disconnected export he
1081 forget release one reference and osc module can't unloaded.
1083 Severity : enhancement
1085 Description: Allow tuning service thread via /proc
1086 Details : For each service a new
1087 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1088 created that can be used to set min/max thread counts, and get the
1089 current number of running threads.
1094 Descriptoin: don't return error if have partially created objects for file.
1095 Details : lov_update_create_set uses set->set_success as index for created
1096 objects, so if some requests will be failed, they will have hole
1097 at end of array and qos_shrink_lsm for allocate correct lsm.
1099 Severity : enhancement
1101 Description: Update OFED support to 1.4.1
1106 Description: Reduce small size read RPC
1107 Details : Set read-ahead limit for every file and only do read-ahead when
1108 available read-ahead pages are bigger than 1M to avoid small size
1111 Severity : enhancement
1113 Description: move AT tunable parameters for more consistent usage
1114 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1116 Severity : enhancement
1118 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1119 Details : allow skip disconnected ost for send statfs request and hide error
1125 Description: fix racy locking of mballoc block bitmaps causing BUG
1126 Details : The locking of the mballoc buddy bitmap and the in-memory
1127 block bitmap was using two different spin locks in some
1128 cases. This made it possible to incorrectly access the
1129 mballoc bitmap while another process was modifying it,
1130 causing a sanity assertion to fail. While no on-disk corruption
1131 was reported, there was some risk of this happening.
1134 Frequency : rare, on llog test 6
1136 Descriptoin: don't allow connect to already connected import
1137 Details : allowing connect to already connected import is hide connecting problem.
1142 Descriptoin: allow kill process which wait statahead result
1143 Details : for some reasons 'ls' can stick in waiting result from statahead,
1144 in this case need way for kill this process.
1146 Severity : enhancement
1148 Description: Add state history info file, enhance import info file
1149 Details : Track import connection state changes in a new osc/mdc proc file;
1150 add overview-type data to the osc/mdc import proc file.
1152 Severity : enhancement
1154 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1156 Severity : enhancement
1158 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1163 Descriptoin: OST grant too much space to client even there are not enough space.
1164 Details : Client will shrink its grant cache to OST if there are no write
1165 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1166 this grant cache if there are already not enough avaible space
1167 (left_space < total_clients * 32M).
1170 Frequency : start MDS on uncleanly shutdowned MDS device
1172 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1173 Details : stay in waiting mds<>ost recovery finished produce random bugs
1174 due race between two ll_sync thread for one lov target. send
1175 ACTIVATE event only if connect realy finished and import have
1179 Frequency : rare, connect and disconnect target at same time
1181 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1182 Details : don't call obd_disconnect under lov_lock. this long time
1183 operation and can block ptlrpcd which answer to connect request.
1188 Descriptoin: don't lose wakeup for imp_recovery_waitq
1189 Details : recover_import_no_retry or invalidate_import and import_close can
1190 both sleep on imp_recovery_waitq, but we was send only one wakeup
1194 Frequency : always with long access acl
1196 Descriptoin: mds can't pack reply with long acl.
1197 Details : mds don't control size of acl but they limited by reint/getattr
1200 Severity : enhancement
1202 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1204 Severity : enhancement
1206 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1209 Frequency : start MDS on uncleanly shutdowned MDS device
1211 Descriptoin: aborting recovery hang on MDS
1212 Details : don't throttle destroy RPCs for the MDT.
1215 Frequency : on remount
1217 Description: external journal device not working after the remount
1218 Details : clear dev_rdonly flag for external journal devices in
1224 Description: shutdown vs evict race
1225 Details : client_disconnect_export vs connect request race.
1226 if client will evicted at this time - we start invalidate
1227 thread without referece to import and import can be freed
1231 Frequency : rare, need acl's on inode.
1233 Description: client can't handle ost additional correctly
1234 Details : if ost was added after client connected to mds client can have
1235 hit lnet_try_match_md ... to big messages to wide striped files.
1236 in this case need teach client to handle config events about add
1237 lov target and update client max ea size at that event.
1239 Severity : enhancement
1241 Description: Changelogs
1242 Details : Changelogs are a lightweight mechanism to track filesystem
1243 metadata and namespace changes. The changelog is recorded
1244 permanently on the MDTs, and is periodically "consumed" / purged
1245 when records are no longer needed.
1247 Severity : enhancement
1249 Description: compact fld format with extents
1250 Details : Store range of seq rather than every seq in FLD. Seq
1251 controller update FLD rather than clients. In Case of CMD, mdt0
1252 has FLD, all other metadata server act as non persistent proxy
1253 for FLD queries and cache fld entries in fld cache.
1258 Description: don't skip ost target if they assigned to file
1259 Details : Drop slow OSCs if we can, but not for requested start idx.
1260 This means "if OSC is slow and it is not the requested
1261 start OST, then it can be skipped, otherwise skip it only
1262 if it is inactive/recovering/out-of-space.
1266 Description: more cleanup in mds_lov
1267 Details : not send LOV EA under replay, we can't know about they size at this
1268 time. Don't allow client connect to mds before any ost connected,
1269 for avoid problems with LOV EA size and returning EIO to client.
1271 Severity : enhancement
1273 Description: Interoperability at server side (Disk interoperability)
1275 Severity : enhancement
1277 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1279 Severity : enhancement
1281 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1283 Severity : enhancement
1285 Description: New client IO stack (CLIO).
1287 Severity : enhancement
1289 Description: Commit on sharing. Eliminate inter-client dependencies between
1290 uncommitted transactions by doing transaction commits.
1291 Thereby clients may recovery independently.
1294 Frequency : Create a symlink file with a very long name
1296 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1297 Details : If there is no extra space in the request for early cancels,
1298 ldlm_req_handles_avail() returns 0 instead of a negative value.
1300 Severity : enhancement
1302 Description: Add /proc entry for import status
1303 Details : The mdc, osc, and mgc import directories now have
1304 an import directory that contains useful import data for debugging
1305 connection problems.
1307 Severity : enhancement
1309 Description: Re-disable certain /proc logging
1310 Details : Enable and disable client's offset_stats, extents_stats and
1311 extents_stats_per_process stats logging on the fly.
1314 Frequency : Only on FC kernels 2.6.22+
1316 Description: oops in statahead
1317 Details : Do not drop reference count for the dentry from VFS when lookup,
1318 VFS will do that by itself.
1320 Severity : enhancement
1322 Description: Generic /proc file permissions
1323 Details : Set /Proc file permissions in a more generic way to enable non-
1324 root users operate on some /proc files.
1328 Description: Hitting mdc_commit_close() ASSERTION
1329 Details : Properly handle request reference release in
1330 ll_release_openhandle().
1334 Description: quota recovery deadlock during mds failover
1335 Details : This patch includes att18982, att18236, att18237 in bz14840.
1337 1. fix osts hang when mds does failover with quotaon
1338 2. prevent watchdog storm when osts threads wait for the
1343 Frequency : only patchless client
1344 Description: add workaround for race between add/remove dentry from hash
1346 Severity : enhancement
1348 Description: Allow OST glimpses to return PW locks
1352 Description: LBUG when llog conf file is full
1353 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1357 Description: Prevent import from entering FULL state when server in recovery
1361 Description: service mount cannot take device name with ":"
1362 Details : Only when device name contains ":/" will mount treat it as
1368 Description: replace ptlrpcd with the statahead thread to interpret the async
1369 statahead RPC callback
1373 Frequency : on recovery
1374 Description: I/O failures after umount during fail back
1375 Details : if client reconnected to restarted server we need join to recovery
1376 instead of find server handler is changed and process self
1377 eviction with cancel all locks.
1379 Severity : enhancement
1381 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1383 Severity : enhancement
1385 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1387 Severity : enhancement
1389 Description: Upcall on Lustre log has been dumped
1390 Details : Allow for a user mode script to be called once a Lustre log has
1391 been dumped. It passes the filename of the dumped log to the
1392 script, the location of the script can be specified via
1393 /proc/sys/lnet/debug_log_upcall.
1398 Description: avoid idr_remove called for id which is not allocated.
1399 Details : Move assigment s_dev for clustered nfs to end of initialization,
1400 to avoid problem with error handling.
1405 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1406 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1407 into CONN_USED_HASH and this prodice warning when put connection
1408 again in unused hash.
1411 Severity : enhancement
1413 Description: File striping can now be set to use an arbitrary pool of OSTs.
1415 Severity : enhancement
1417 Description: Export bytes_read/bytes_write count on OSC/OST.
1421 Description: Early reply size mismatch, MGC loses connection
1422 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1423 the connect flags are properly negotiated.
1428 Description: quotas are not honored with O_DIRECT
1429 Details : all writes with the flag O_DIRECT will use grants which leads to
1430 this problem. Now using OBD_BRW_SYNC to guard this.
1434 Description: add quota statistics
1435 Details : 1. sort out quota proc entries and proc code.
1436 2. add quota statistics
1438 Severity : enhancement
1440 Description: enable quota support for HEAD.
1444 Description: Properly propagate oinfo flags from lov to osc for statfs
1445 Details : restore missing copy oi_flags to lov requests.
1447 Severity : enhancement
1449 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1450 : l_getgroups(8), lst(8), routerstat(8)
1451 Details : included man pages for llobdstat(8), llstat(8),
1452 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1454 Severity : enhancement
1456 Description: configure's --enable-quota should check the
1457 : kernel .config for CONFIG_QUOTA
1458 Details : configure is terminated if --enable-quota is passed but
1459 : no quota support is in kernel
1463 Description: Remove portals compatibility
1464 Details : Remove portals compatibility, not interoperable with releases
1469 Description: Resolve device initialization race
1470 Details : Prevent proc handler from accessing devices added to the
1471 obd_devs array but yet be intialized.
1473 Severity : enhancement
1475 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1477 Severity : enhancement
1479 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1484 Description: Using wrong pointer in osc_brw_prep_request
1485 Details : Access to array[-1] can produce panic if kernel compiled with
1486 CONFIG_PAGE_ALLOC enabled
1490 Description: Client runs out of low memory
1491 Details : Consider only lowmem when counting initial number of llap pages
1495 Description: *optional* service tags registration
1496 Details : if the "service tags" package is installed on a Lustre node
1497 When the filesystem is mounted, a local-node service tag will
1498 be created. See http://inventory.sun.com/ for more information
1499 about the Service Tags asset management system.
1503 Description: Kernel BUG tries to release flock
1504 Details : Lustre does not destroy flock lock before last reference goes
1505 away. So always drop flock locks when client is evicted and
1506 perform unlock regardless of successfulness of speaking to MDS.
1510 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1514 Description: sanity test 65a fails if stripecount of -1 is set
1515 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1520 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1521 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1524 Severity : enhancement
1526 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1530 Description: do not process already freed flock
1531 Details : flock can possibly be freed by another thread before it reaches
1532 to ldlm_flock_completion_ast.
1536 Description: LBUG during stress test
1537 Details : Need properly lock accesses the flock deadlock detection list.
1541 Description: oops in page fault handler
1542 Details : kernel page fault handler can return two special 'pages' in error
1543 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1547 Description: timeout with invalidate import.
1548 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1549 should be handled by ptlrpcd. This produce long age waiting and
1550 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1552 Severity : enhancement
1554 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1558 Description: enable MGS and MDT services start separately
1559 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1560 a co-located MGS without starting the MGS, which is a complement
1561 to 'nosvc' mount option.
1565 Frequency : after recovery
1566 Description: precreate to many object's after del orphan.
1567 Details : del orphan st in oscc last_id == next_id and this triger growing
1568 count of precreated objects. Set flag LOW to skip increase count
1569 of precreated objects.
1573 Frequency : rare, on clear nid stats
1574 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1575 Details : when clean nid stats sometimes try destroy live entry,
1576 and this produce panic in free.
1580 Description: Stack overflow during MDS log replay
1581 ease stack pressure by using a thread dealing llog_process.
1585 Description: wait until IO finished before start new when do lock cancel.
1586 Details : VM protocol want old IO finished before start new, in this case
1587 need wait until PG_writeback is cleared until check dirty flag and
1588 call writepages in lock cancel callback.
1590 Severity : enhancement
1592 Description: using special macro for print time and cleanup in includes.
1596 Description: mds_mfd_close() ASSERTION(rc == 0)
1597 Details : In mds_mfd_close(), we need protect inode's writecount change
1598 within its orphan write semaphore to prevent possible races.
1602 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1606 Frequency : rare, on shutdown ost
1607 Description: don't hit live lock with umount ost.
1608 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1609 use shrink_dcache_sb instead.
1613 Description: don't panic with use echo client
1614 Details : echo client pass NULL as client nid pointer and this produce null
1615 pointer dereference.
1619 Description: fix build on ppc32
1620 Details : compile code with -m64 flag produce wrong object file for ppc32.
1624 Description: add message levels for liblustreapi
1628 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1629 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1630 will result in an -EAGAIN and in the exit call path we call
1631 alloc_rr() which will with increasing aggressiveness attempt to
1632 aquire precreated objects on the minimum number of required OSCs.
1636 Description: Use old size assignment to avoid deadlock
1637 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1638 the scheduling while holding a spinlock. We do not need locking
1639 for size in ll_update_inode() because size is only updated from
1640 the MDS for directories or files without objects, so there is no
1641 other place to do the update, and concurrent access to such inodes
1642 are protected by the inode lock.
1646 Description: resolve "_IOWR redefined" build error on SLES10
1650 Description: dump the memory debugging after all modules are unloaded to
1651 suppress false negative in conf_sanity test 39
1653 Severity : enhancement
1655 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1660 Description: fix SLES kernel versioning
1661 Details : the kernel version for our SLES 10 kernel did not include a "-"
1662 before the "smp" at the end. while this was not a problem in
1663 general, it did mean that software trying to use the kernel
1664 version to try to detect a vendor specific kernel would fail.
1665 this was most evident by the OFED build scripts.
1669 Description: Don't update lov_desc members until making sure they are valid
1670 Details : When updating lov_desc members via proc fs, need fix their
1671 validities before doing the real update.
1675 Description: don't put request into delay list while invalidate in flight.
1676 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1677 import in flight. this produce timeout for invalidate and sometimes
1678 can cause stale data.
1682 Frequency : on ppc only
1683 Description: not convert ost objects for directory because it's not exist.
1684 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1688 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1690 Severity : enhancement
1692 Description: Update to support 2.6.22.14 vanilla kernel.
1696 Frequency : rare, on recovery
1697 Description: read procfs can produce deadlock in some situation
1698 Details : Holding lprocfs lock which send rpc can produce block for destroy
1699 obd objects and this also block reconnect with -EALREADY. This
1700 isn't fix all lprocfs bugs - but make it rare.
1702 Severity : enhancement
1704 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1707 Frequency : frequent on X2 node
1709 Description: mdc_set_open_replay_data LBUG
1710 Details : Set replay data for requests that are eligible for replay.
1714 Description: lustre_mgs: operation 101 on unconnected MGS
1715 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1716 MGC, and late on MGC cannot successfully connect to MGS and a lot
1717 of the error messages complaining that MGS is not connected.
1721 Frequency : on network error
1722 Description: panic with double free request if network error
1723 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1724 but it's true only for synchronus enqueue, for async enqueue
1725 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1728 Severity : enhancement
1730 Description: client-side metadata stat-ahead during readdir(directory readahead)
1731 Details : perform client-side metadata stat-ahead when the client detects
1732 readdir and sequential stat of dir entries therein
1735 Frequency : on start mds
1737 Description: Implement get_info(last_id) in obdfilter.
1740 Frequency : occasional
1742 Description: Correctly check stale fid, not start epoch if ost not support SOM
1743 Details : open with flag O_CREATE need set old fid in op_fid3 because
1744 op_fid2 was overwritten with new generated fid, but mds can answer
1745 with one of these two fids and both is not stale. Setattr
1746 incorrectly started an epoch and assume will be called
1747 done_writing, but without SOM done_writing ever being called.
1750 Frequency : rare, depends on device drivers and load
1752 Description: MDS or OSS nodes crash due to stack overflow
1753 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1754 In some cases, in conjunction with device drivers that use a lot
1755 of stack the MDS (or possibly OSS) service threads could overflow
1756 the stack. One change which was identified to consume additional
1757 stack has been reworked to avoid the extra stack usage.
1760 Frequency : occasional
1762 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1763 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1764 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1765 oscc_has_objects and see OSC has no objests and call
1766 oscc_internal_create to resend create request.
1768 Severity : enhancement
1770 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1772 Severity : enhancement
1774 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1777 Frequency : very rare
1779 Description: Fix replay if there is an un-replied request and open
1780 Details : In some cases, older replay request will revert the
1781 mcd->mcd_last_xid on MDS which is used to record the client's
1782 latest sent request.
1784 Severity : enhancement
1786 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1788 Severity : enhancement
1790 Description: Add rhel5 support to HEAD.
1792 Serverity : enhancement
1794 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1799 Description: Don't allow skipping OSTs if index has been specified.
1800 Details : Don't allow skipping OSTs if index has been specified, make
1801 locking in internal create lots better.
1805 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1806 Details : access to bitfield in structure is always rounded to long
1807 and this produce problem with not atomic change any bit.
1811 Description: Lustre make rpms failed.
1812 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1813 builds Lustre rpms from tarball.
1816 Frequency : rare on shutdown ost
1818 Description: If llog cancel was not send before clean_exports phase, this can
1819 produce deadlock in llog code.
1820 Details : If llog thread has last reference to obd and call class_import_put
1821 this produce deadlock because llog_cleanup_commit_master wait when
1822 last llog_commit_thread exited, but this never success because was
1823 called from llog_commit_thread.
1827 Description: allow userland application know is lost one of stripes.
1828 Details : fill lvb_blocks with error code on ost and return it to
1829 application if error flag found.
1833 Description: NULL lov_tgts causing MDS oops
1834 Details : more safe checks for NULL lov_tgts for avoid oops.
1836 Severity : enhancement
1838 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1842 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1843 Details : If an OST is down the MDS will hang indefinitely in
1844 obd_statfs_rqset() waiting for the statfs data. While for
1845 MDS QOS usage of statfs, it should not stuck in waiting.
1847 Severity : enhancement
1849 Description: remote_acl support
1850 Details : Support ACL-based permission check for remote user.
1851 Support setfacl/getfacl for remote user with the utils
1852 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1853 the system "{s,g}etfacl" utils.
1855 Severity : enhancement
1857 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1859 Severity : enhancement
1861 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1865 Description: make mgs_setparam() handle fsname containing dash
1866 Details : fsname containing a dash does not work with lctl conf_param
1868 Severity : enhancement
1870 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1872 Severity : enhancement
1874 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1878 Frequency : rare, at shutdown
1879 Description: access already free / zero obd_namespace.
1880 Details : if client_disconnect_export was called without force flag set,
1881 and exist connect request in flight, this can produce access to
1882 NULL pointer (or already free pointer) when connect_interpret
1883 store ocd flags in obd_namespace.
1887 Frequency : only at startup
1888 Description: not alloc memory with spinlock held.
1889 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1890 if any spinlock held.
1892 Severity : enhancement
1894 Description: make lustre randomly fail allocating memory
1895 Details : Make lustre randomly failed allocating memory for testing purpose.
1897 Severity : enhancement
1899 Description: lost problems with lov objid file
1900 Details : Fixes some scability and access to not inited memory problems
1901 in work with lov objdid file.
1906 Description: lfs find does not continue on file error
1907 Details : Continue other files processing when a file/dir is absent.
1911 Description: Inconsistent usage of lustre_pack_reply()
1912 Details : Standardize the usage of lustre_pack_reply() such that it
1913 always generate a CERROR on failure.
1916 Frequency : occasional
1918 Description: MDS hang or stay in waiting lock
1919 Details : If client receive lock with CBPENDING flag ldlm need send lock
1920 cancel as separate rpc, to avoid situation when cancel request
1921 can't processed due all i/o threads stay in wait lock.
1925 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1929 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1933 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1937 Description: Update to RHEL4 latest kernel.
1939 Severity : enhancement
1941 Description: Build SLES10 patchless client fails
1942 Details : The configure was broken by run ./configure with
1943 --with-linux-obj=.... argument for patchless client. When the
1944 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1945 can't search header adequately. Use absolute path such as
1946 -I($LINUX)/include instead.
1950 Description: interrupt oig_wait produce painc on resend.
1951 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1952 private set, and this produce situation when rq_ptlrpcd_data not
1953 copyed to new allocated request and triggered LBUG on assert
1954 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1955 ptlrpcd set if request is changed and can be safety changed to use
1960 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1961 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1962 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1963 these macros and replace them with their difinitions instead.
1967 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1968 Details : Update which_patch & target file for SLES10 latest kernel.
1971 Description: Remove client patches from SLES 10 kernel.
1972 Details : This causes SLES 10 clients to behave as patchless clients
1973 even on a Lustre-patched (server) kernel.
1975 Severity : enhancement
1977 Description: self-adjustable client's lru lists
1978 Details : use adaptive algorithm for managing client cached locks lru
1979 lists according to current server load, other client's work
1980 pattern, memory activities, etc. Both, server and client
1981 side namespaces provide number of proc tunables for controlling
1984 Severity : enhancement
1986 Description: light-weight GSS support
1987 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1988 and reduce performance overhead.
1990 Severity : enhancement
1992 Description: Linux keyring support
1993 Details : Support using service of Linux keyring for Lustre GSS internal
1994 context refresh/cache mechanism.
1998 Description: Fix errors in lfs documentation
1999 Details : Fixes man pages
2003 Description: don't use GFP_* in generic Lustre code.
2004 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2008 Description: obdclass is limited by single OBD_ALLOC(idarray)
2009 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2013 Description: warnings with build patchless client with vanila 2.6.19 and up
2014 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2015 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2019 Frequency : Only for SLES
2020 Description: sanity_quota fail test_1
2021 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2022 file, which makes TSTID[2] inunique.
2026 Frequency : Only for SLES9
2027 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2028 Details : Update target/ChangeLog/which_patch .
2032 Frequency : Only for test_10 in sanity-quota.sh
2033 Description: a bug in quota test code
2034 Details : checking a test flag in a wrong place
2038 Frequency : Only for test_13 in sanity-quota.sh
2039 Description: a bug in quota test script
2040 Details : shouldn't check overfull allocated quota
2044 Description: jbd statistics
2045 Details : Port older jbd statistics patch for sles10
2047 Severity : enhancement
2049 Description: Add printing inode info into message about error in writepage.
2054 Description: reply_lock_interpret crash due to race with it and lock cancel.
2055 Details : Do not replay locks that are being cancelled. Do not reference
2056 locks by their address during replay, just by their handle.
2060 Frequency : When flocks are used.
2061 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2062 Details : Flock locks might destroy just granted lock if it could be merged
2063 with another existing flock, this is done in completion handler,
2064 so teach ldlm_cli_enquque_fini that this is a valid case for
2070 Description: Oops in read and write path when failing to allocate lock.
2071 Details : Check if lock allocation failed and return error back.
2075 Description: lstripe command fails for valid OST index
2076 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2077 instead of lov->desc.ld_active_tgt_count.
2081 Description: sanity.sh failed test 103
2082 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2087 Description: df doesn't work properly if diskfs blocksize != 4K
2088 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2092 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2093 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2095 Severity : enhancement
2097 Description: Adaptive timeouts
2098 Details : RPC timeouts adapt to changing server load and network
2099 conditions to reduce resend attempts and improve recovery time.
2103 Description: llapi_file_create() does not allow some changes
2104 Details : add llapi_file_open() that allows specifying the mode and
2105 open flags, and also returns an open file handle.
2109 Description: merge and cleanup kernel patches.
2110 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2114 Description: Add journal checksum support.(Kernel part)
2115 Details : The journal checksum feature adds two new flags i.e
2116 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2117 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2118 indicates that the commit block contains the checksum for
2119 the blocks described by the descriptor blocks. Now commit
2120 record can be sent to disk without waiting for descriptor
2121 blocks to be written to disk. This behavior is controlled
2122 using JBD2_FEATURE_ASYNC_COMMIT flag.
2126 Description: OSS needs mutliple precreate threads
2127 Details : Add ability to start more than one create thread per OSS.
2131 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2132 Details : execute lfs setstripe on client
2136 Description: mds_obd_create error creating tmp object
2137 Details : When the user sets quota on root, llog will be affected and can't
2138 create files and write files.
2141 Frequency : Always on ia64 patchless client, and possibly others.
2143 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2144 Details : This allows the patchless client to be loaded on architectures
2145 without this export.
2149 Description: RedHat Update kernel for RHEL5
2150 Details : Add the kernel config file for RHEL5.
2154 Description: RedHat Update kernel for RHEL5
2155 Details : Modify the kernel config file more closer RHEL5.
2159 Description: Build failure against Centos5 (RHEL5)
2160 Details : Define PAGE_SIZE when it isn't present.
2164 Description: test_7 on sanity-quota.sh can't be used on separate machines
2165 Details : add facet to handle it
2169 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2170 Details : with NFS, the anon dentry's parent was set to itself in
2171 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2172 corresponding dentry other than use rec->ur_name.
2174 Severity : enhancement
2176 Description: lfs setstripe enhancement
2177 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2181 Description: enable data checksumming by default
2182 Details : enable checksum by default, allow --disable-checksum
2183 configure option and "-o nochecksum" mount option. Checksums
2184 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2188 Description: lustre support for RHEL5
2189 Details : Add support for RHEL5.
2193 Description: Client eviction due to failover config
2194 Details : after a connection loss, the lustre client should attempt to
2195 reconnect to the last active server first before trying the
2196 other potential connections.
2200 Description: when mds and osts use different quota unit(32bit and 64bit),
2201 quota will be released repeatly.
2202 Details : void sending multiple quota reqs to mds, which will keep the
2203 status between the reqs.
2207 Description: osts not allocated evenly to files
2208 Details : change the condition to increase offset_idx
2211 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2213 Description: Data corruption for OSTs that are formatted larger than 2TB
2215 Details : When generating the bio request for lustre file writes the
2216 sector number would overflow a temporary variable before being
2217 used for the IO. The data reads correctly from Lustre (which
2218 will overflow in a similar manner) but other file data or
2219 filesystem metadata may be corrupted in some cases.
2223 Description: Tune the kernel for good SCSI performance.
2224 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2225 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2230 Description: rewrite ext2-derived code in obdclass/uuid.c
2231 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2232 from scratch preserving functionality.
2236 Description: "lfs find -obd UUID" prints directories
2237 Details : "lfs find -obd UUID" will return all directory names instead
2238 of just file names. It is incorrect because the directories
2239 do not reside on the OSTs.
2243 Description: use i_size_read and i_size_write in 2.6 port
2244 Details : replace inode->i_size access with i_size_read/write()
2247 Frequency : when removing large files
2249 Description: scheduling issue during removal of large Lustre files
2250 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2251 It causes scheduling issues when removing large files (17TB in the
2255 Frequency : only with liblustre clients on XT3
2257 Description: evictions taking too long
2258 Details : allow llrd to evict clients directly on OSTs
2261 Frequency : only on ppc
2263 Description: /proc/fs/lustre/devices broken on ppc
2264 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2265 We should make sure the type of 'index' is loff_t and then cast
2266 explicitly as needed below. Do not assign an explicitly cast
2270 Frequency : only for rhel5
2272 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2273 Details : Modify the target file & which_kernel.
2275 Severity : enhancement
2277 Description: omit set fsid for export NFS
2278 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2279 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2283 Frequency : Always, for kernels after 2.6.16
2284 Description: Fix warning idr_remove called for id=.. which is not allocated.
2285 Details : Last kernels save old s_dev before kill super and not allow
2286 to restore from callback - restore it before call kill_anon_super.
2290 Description: buffer overruns could theoretically occur
2291 Details : llapi_semantic_traverse() modifies the "path" argument by
2292 appending values to the end of the origin string, and a
2293 overrun may occur. Adding buffer overrun check in liblustreapi.
2297 Description: Fix error on 'ls .' at the top of the Lustre mount.
2298 Details : Don't revalidate dentry if it is a root dentry.
2302 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2303 Details : Modify vm-tunables-rhel4.patch.
2307 Description: Kernel config for 2.6.18-vanilla.
2308 Details : Modify targets/2.6-vanilla.target.in.
2309 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2310 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2311 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2312 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2316 Description: improve handling recoverable errors
2317 Details : if request processig with error which can be recoverable on server
2318 request should be resend, otherwise page released from cache and
2323 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2324 Details : Modify target file & which_patch.
2325 A flaw was found in the IA32 system call emulation provided
2326 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2327 value could be stored in the %RAX register, which could trigger an
2328 out-of-bounds system call table access. An untrusted local user
2329 could exploit this flaw to run code in the kernel
2330 (ie a root privilege escalation). (CVE-2007-4573).
2334 Description: O_DIRECT bypasses client statistics.
2335 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2336 read_ahead_stats were updated. copied the stats section from
2337 osc_send_oap_rpc() into async_internal().
2341 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2345 Description: change order of libsysio includes
2346 Details : '#include sysio.h' should always come before '#include xtio.h'
2350 Description: handle "serious error: objid * already exists" more gracefully
2351 Details : If LAST_ID value on disk is smaller than the objects existing in
2352 the O/0/d* directories, it indicates disk corruption and causes an
2353 LBUG(). If the object is 0-length, then we should use the existing
2354 object. This will help to avoid a full fsck in most cases.
2356 Severity : enhancement
2358 Description: adapt the lustre_config script to support the upgrade case
2359 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2360 server targets to 1.6 in parallel.
2364 Description: To avoid grant space > avaible space when the disk is almost
2365 full. Without this patch you might see the error "grant XXXX >
2366 available" or some LBUG about grant, when the disk is almost
2368 Details : In filter_check_grant, for non_grant cache write, we should
2369 check the left space by if (*left > ungranted + bytes), instead
2370 of (*left > ungranted), because only we are sure the left space
2371 is enough for another "bytes", then the ungrant space should be
2372 increase. In client, we should update cl_avail_grant only there
2373 is OBD_MD_FLGRANT in the reply.
2377 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2378 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2379 described in CVE-2007-4573. This problem could allow untrusted
2380 local users to gain root access.
2383 Frequency : when using O_DIRECT and quotas
2385 Description: Incorrect file ownership on O_DIRECT output files
2386 Details : block usage reported by 'lfs quota' does not take into account
2387 files that have been written with O_DIRECT.
2392 Description: touch file failed when fs is not full
2393 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2394 otherwise we can get ENOSP while fs is not full.
2398 Description: parallel lock callbacks
2399 Details : Instead of sending blocking and completion callbacks as separated
2400 requests, adding them to a set and sending in parallel.
2403 Frequency : only for Cray XT3
2404 Bugzilla : 12829/13455
2405 Description: Changing primary group doesn't change the group lustre assigns to
2407 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2408 overridden with the primary group provided by the group upcall,
2409 whereas the supplied fsgid can be trusted if it is in the list of
2410 supplementary groups returned by the group upcall.
2412 Severity : enhancement
2414 Description: Allow masking D_WARNING, D_ERROR messages from console
2415 Details : Console messages can now be disabled via lnet.printk.
2419 Description: User code with malformed file open parameter crashes client node
2420 Details : Before packing join_file req, all the related reference should be
2421 checked carefully in case some malformed flags cause fake
2422 join_file req on client.
2426 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2427 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2428 mode is equal to the granted mode. However, it should also check
2429 that LDLM_FL_CP_REQD is not set.
2433 Description: Heavy nfs access might result in deadlocks
2434 Details : After ELC code landed, it is now improper to enqueue any mds
2435 locks under och_sem, because enqueue might want to decide to
2436 cancel open locks for same inode we are holding och_sem for.
2440 Description: Client eviction while running blogbench
2441 Details : A lot of unlink operations with concurrent I/O can lead to a
2442 deadlock causing evictions. To address the problem, the number of
2443 oustanding OST_DESTROY requests is now throttled to
2444 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2445 callbacks are processed in priority.
2449 Description: enable ACLs on MDS by default
2450 Details : ACLs must be enabled on MDS by default.
2453 Frequency : PPC/PPC64 only
2455 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2456 Details : fix duplicated definitions between asm-ppc64/types.h and
2457 lustre_types.h on PPC.
2460 Frequency : PPC/PPC64 only
2462 Description: asm-ppc/segment.h does not exist
2463 Details : fix compile issue on PPC.
2467 Description: better handle error messages in extents code
2470 Frequency : RHEL4 only
2472 Description: mkfs is very slow on IA64/RHEL4
2473 Details : A performance regression has been discovered in the MPT Fusion
2474 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2475 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2476 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2478 Severity : enhancement
2480 Description: SNMP support enhancement
2481 Details : Adding total number of sampled request for an MDS node in snmp
2484 Severity : enhancement
2486 Description: Optimize ldlm waiting list processing for PR extent locks
2487 Details : When processing waiting list for read extent lock and meeting read
2488 lock that is same or wider to it that is not contended, skip
2489 processing rest of the list and immediatelly return current
2490 status of conflictness, since we are guaranteed there are no
2491 conflicting locks in the rest of the list.
2495 Description: Time out and refuse to reconnect
2496 Details : When the failover node is the primary node, it is possible
2497 to have two identical connections in imp_conn_list. We must
2498 compare not conn's pointers but NIDs, otherwise we can defeat
2499 connection throttling.
2503 Description: port llog fixes from b1_6 into HEAD
2504 Details : Port llog reference couting and some llog cleanups from b1_6
2505 (bug 10800) into HEAD, for protect from panic and access to already
2506 free llog structures.
2510 Description: Detect stride IO mode in read-ahead
2511 Details : When a client does stride read, read-ahead should detect that and
2512 read-ahead pages according to the detected stride pattern.
2516 Description: data checksumming impacts single node performance
2517 Details : add support for several checksum algorithm. Currently, only CRC32
2518 and Adler-32 are supported. The checksum type can be changed on
2519 the fly via /proc/fs/lustre/osc/*/checksum_type.
2523 Description: use adler32 for page checksums
2524 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2529 Description: build for x2 fails
2530 Details : fix compile issue on Cray systems.
2534 Description: Properly match for duplicate locks
2535 Details : Due to different lock order from skiplists code, we need to
2536 traverse entire list for now
2539 Frequency : only on PPC/SLES10
2541 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2542 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2543 be defined. Add a hack in mkfs_lustre.c to work around this compile
2548 Description: LASSERT on MDS when client holding flock lock dies
2549 Details : ldlm pool logic depends on number of granted locks equal to
2550 number of released locks which is not true for flock locks, so
2551 just exclude such locks from consideration.
2555 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2556 Details : Use fsfilt_sync() for both the whole filesystem sync and
2557 individual file sync to eliminate dangerous inode locking
2558 with I_LOCK that can lead to a deadlock.
2562 Description: performance in 1.6.3
2563 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2564 sure that 1MB requests are not fragmented by the block layer.
2566 Severity : enhancement
2568 Description: organize the server-side client stats on per-nid basis
2569 Details : Change the structure of stats under obdfilter and mds to
2577 The "uuid"s file would list the uuids of _active_ exports.
2578 And the clear entry is to clear all stats and stale nids.
2580 Severity : enhancement
2582 Description: eliminate client locks in face of contention
2583 Details : file contention detection and lockless i/o implementation
2584 for contended files.
2588 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2591 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2593 Description: MDS may lose file striping (and hence file data) in some cases
2594 Details : If there are additional extended attributes stored on the MDS,
2595 in particular ACLs, SELinux, or user attributes (if user_xattr
2596 is specified for the client mount options) then there is a risk
2597 of attribute loss. Additionally, the Lustre file striping
2598 needs to be larger than default (e.g. striped over all OSTs),
2599 and an additional attribute must be stored initially in the
2600 inode and then increase in size enough to be moved to the
2601 external attribute block (e.g. ACL growing in size) for file
2606 Description: skiplist implementation simplification
2607 Details : skiplists are used to group compatible locks on granted list
2608 that was implemented as tracking first and last lock of each
2609 lock group the patch changes that to using doubly linked lists
2613 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2614 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2615 because LASSERT on both the data supplied by a client, and the
2616 data on disk is dangerous and incorrect.
2618 Severity : enhancement
2620 Description: Slow truncate/writes to huge files at high offsets.
2621 Details : Directly associate cached pages to lock that protect those pages,
2622 this allows us to quickly find what pages to write and remove
2623 once lock callback is received.
2627 Description: more ldlm soft lockups
2628 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2629 starve other threads from the resource lock for a long time in
2630 case of long waiting queue, so change the debug level from
2631 D_OTHER to the less frequently used D_INFO.
2633 Severity : enhancement
2635 Description: add -gid, -group, -uid, -user options to lfs find
2639 Description: Hung threads in invalidate_inode_pages2_range
2640 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2641 once one is completed. As a result, some RPCs are stuck in the
2642 queue and are never sent.
2646 Description: filter threads hungs on waiting journal commit
2647 Details : Cleanup filter group llog code, then only filter group llog will
2648 be only created in the MDS/OST syncing process.
2652 Description: Procfs and llog threads access destoryed import sometimes.
2653 Details : Sync the import destoryed process with procfs and llog threads by
2654 the import refcount and semaphore.
2656 Severity : enhancement
2658 Description: openlock cache of b1_6 port to HEAD
2663 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2664 Details : If insertion of an extent fails, then discard the inode
2665 preallocation and free data blocks else it can lead to duplicate
2670 Description: don't always update ctime in ext3_xattr_set_handle()
2671 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2672 In some cases the ctime should not be updated, for example for
2673 2.0->1.8 compatibility it is necessary to delete an xattr and it
2674 should not update the ctime.
2678 Bugzilla : 15713/16362
2679 Description: Assertion in iopen_connect_dentry in 1.6.3
2680 Details : looking up an inode via iopen with the wrong generation number can
2681 populate the dcache with a disconneced dentry while the inode
2682 number is in the process of being reallocated. This causes an
2683 assertion failure in iopen since the inode's dentry list contains
2684 both a connected and disconnected dentry.
2688 Description: assertion failure in ldlm_handle2lock()
2689 Details : fix a race between class_handle_unhash() and class_handle2object()
2690 introduced in lustre 1.6.5 by bug 13622.
2695 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2696 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2701 Description: oops in sd_iostats_seq_show()
2702 Details : unloading/reloading the scsi low level driver triggers a kernel
2703 bug when trying to access the sd iostat file.
2708 Description: Kernel panics during QLogic driver reload
2709 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2710 patch, causing memory corruption.
2715 Description: journal_dev option does not work in b1_6
2716 Details : pass mount option during pre-mount.
2718 Severity : enhancement
2720 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2721 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2722 extent information of a file. It can be used to map logical blocks
2723 in a file to physical blocks in the block device.
2727 Description: LDLM soft lockups - improvement
2728 Details : It is be possible to send the lock handle along with each read
2729 or write request because the client is already doing a lock match
2730 itself so there isn't any reason the OST should have to re-do that
2734 Frequency : only on Cray X2
2736 Description: X2 build failures
2737 Details : fix build failures on Cray X2.
2741 Description: xid & resent requests
2742 Details : Initialize RPC XID from clock at startup (randomly if clock is
2745 Severity : enhancement
2747 Description: Add lustre_start utility to start or stop multiple Lustre servers
2752 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2753 Details : In case of memory pressure, list_del() can be called twice on
2754 req->rq_history_list, causing a kernel oops.
2758 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2759 Details : fix stack overflow in the distributed lock manager by defering
2760 export eviction after a failed AST to the elt thread instead of
2761 handling it in the dlm interpret routine.
2765 Description: Convert some comments to new format.
2766 Details : Update documenting comments to match doxygen conventions.
2770 Description: Grammar fixes.
2771 Details : A couple of trivial spelling fixes.
2775 Description: OSD_COUNTERS-mandatory
2776 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2777 to simplify the code.
2781 Description: simplify cmm_device freeing logic.
2782 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2786 Description: Add lockdep support to dt_object_operations locking interface.
2787 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2788 parameter indicating lock ordering. Update mdd code to use new
2793 Description: Introduce failloc constants for lockless IO tests.
2794 Details : Add two new failloc constants to test lockless IO. Only one of
2795 them in implemented---another is checked in yet to be landed
2800 Description: Add lockdep support for inode mutex.
2801 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2802 in the situations where Lustre has to lock more than one inode
2807 Description: Add optional invariants checking support.
2808 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2809 configure switch. This macro is to be used for consistency and
2810 sanity checks that are too expensive to be left in `production'
2815 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2816 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2821 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2822 Details : Use spin_lock_nested() in (the only) situation where more than
2823 one ldlm_lock is locked simultaneously. Also, fix possible
2824 dead-lock in ldlm_lock_change_resource() by enforcing particular
2829 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2830 Details : Instead of specifying each ldlm_lock call-back through separate
2831 parameter, wrap them into struct ldlm_callback_suite.
2835 Description: Kill join_lru obd method and its callers.
2836 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2837 in memory---a requirement implemented through ->o_join_lru() obd
2838 method in HEAD. Remove this method and its users.
2842 Description: Add asynchronous ldlm ENQUEUE completion handler.
2843 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2844 case that stock ldlm_completion_ast() cannot handle as it waits
2845 until lock is granted. Introduce new ldlm_completion_ast_async()
2846 for this. Also comment ldlm_completion_ast().
2850 Description: ldlm_error <-> errno conversion.
2851 Details : Add functions to map (rather arbitrary) between LDLM error codes
2852 and standard errno values. CLIO needs this to prevent LDLM specific
2853 constants from escaping ldlm and osc.
2857 Description: Kill unused ldlm_handle2lock_ns() function.
2858 Details : Kill unused ldlm_handle2lock_ns() function.
2862 Description: Add lu_ref support to ldlm_lock
2863 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2864 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2865 and ldlm_resource. LDLM interface has to be changed, because code that
2866 releases a reference on a lock, has to "know" what reference this is.
2867 In the most frequent case
2869 lock = ldlm_handle2lock(handle);
2871 LDLM_LOCK_PUT(lock);
2873 no changes are required. When any other reference (received _not_ from
2874 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2875 instead of LDLM_LOCK_PUT().
2877 Arguably, changes are pervasive, and interface requires some discipline
2878 for proper use. On the other hand, it was very instrumental in finding
2879 a few leaked lock references.
2883 Description: Add ldlm_lock_addref_try().
2884 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2885 attempts to addref a lock that might be being canceled
2890 Description: Add ldlm_weigh_callback().
2891 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2892 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2893 hard-coded `number of pages' logic.
2897 Description: Add lockdep annotations to llog code.
2898 Details : Use appropriately tagged _nested() locking calls in the places
2899 where llog takes more than one ->lgh_lock lock.
2903 Description: Add loi_kms_set().
2904 Details : Wrap kms updates into a helper function.
2908 Description: Constify instances of struct lsm_operations.
2909 Details : Constify instances of struct lsm_operations.
2913 Description: lu_conf support.
2914 Details : On a server, a file system object is uniquely identified
2915 by a fid, which is sufficient to locate and load all object
2916 state (inode). On a client, on the other hand, more data are
2917 necessary instantiate an object. Change lu_object_find() and
2918 friends to take additional `lu_conf' argument describing object.
2919 Typically this includes layout information.
2923 Description: lu_context fixes.
2924 Details : Introduce new lu_context functions that are needed on the client
2925 side, where some system threads (ptlrpcd) are shared by multiple
2926 modules, and so cannot be stopped during module shutdown.
2930 Description: Add start and stop methods to lu_device_type_operations.
2931 Details : Introduce two new methods in lu_device_type_operations, that are
2932 invoked when first instance of a given type is created and last one
2933 is destroyed respectively. This is need by CLIO.
2937 Description: Add lu_ref support to struct lu_device.
2938 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2939 track leaked references.
2943 Description: Introduce lu_kmem_descr.
2944 Details : lu_kmem_descr and its companion interface allow to create
2945 and destroy a number of kmem caches at once.
2949 Description: Fix lu_object finalization race.
2950 Details : Fix a race between lu_object_find() finding an object and its
2951 concurrent finalization. This race is (most likely) not possible
2952 on the server, but might happen on the client.
2956 Description: Introduce lu_ref interface.
2957 Details : lu_ref is a debugging module allowing to track references to
2958 a given object. It is quite cpu expensive, and has to be
2959 explicitly enabled with --enable-lu_ref. See usage description
2964 Description: Factor lu_site procfs stats into a separate function.
2965 Details : Separate lu_site stats printing code into a separate function
2966 that can be reused on a client.
2970 Description: Constify instances of struct {lu,dt,md}_device_operations.
2971 Details : Constify instances of struct {lu,dt,md}_device_operations.
2975 Description: Introduce struct md_site and move meta-data specific parts of
2976 struct lu_site here.
2977 Details : Move md-specific fields out of struct lu_site into special struct
2978 md_site, so that lu_site can be used on a client.
2982 Description: Kill mdd_lov_destroy().
2983 Details : Remove unused mdd code.
2987 Description: Add st_block checking to multistat.c.
2988 Details : Add st_block checking to multistat.c.
2992 Description: Add lu_ref support to struct obd_device.
2993 Details : Add lu_ref tracking to obd_device.
2997 Description: Kill obd_set_fail_loc().
2998 Details : Remove unused code.
3002 Description: Add special type for ptlrpc_request interpret functions.
3003 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
3004 there. Actual usage will be in CLIO.
3008 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
3009 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
3010 doesn't work with lockdep.
3014 Description: Add rwv.c test program.
3015 Details : New testing program exercising readv(2) and writev(2) (Qian).
3019 Description: Add sendfile.c test program.
3020 Details : New testing program exercising sendfile(2) (Jay).
3024 Description: Ratelimit a message that can be very frequent.
3025 Details : Ratelimit a memory allocation failure message that can
3030 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3031 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3035 Description: Liblustre build fixes.
3036 Details : Liblustre build fixes.
3040 Description: libcfs: add cfs_{need,cond}_resched() interface.
3041 Details : libcfs: add cfs_{need,cond}_resched() definition and
3042 implementations for Linux, NT, and liblustre.
3044 Severity : enhancement
3046 Description: More exported tunables for mballoc
3047 Details : Add support for tunable preallocation window and new tunables for
3048 large/small requests
3052 Description: Detect corruption of block bitmap and checking for preallocations
3053 Details : Checks validity of on-disk block bitmap. Also it does better
3054 checking of number of applied preallocations. When corruption is
3055 found, it turns filesystem readonly to prevent further corruptions.
3059 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3060 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3061 window to 0. If it does hit cache miss, and read-pattern is still
3062 stride-io mode, does not reset the stride window, but also does not
3063 increase the stride window length in this case.
3067 Frequency : only for big-endian servers
3068 Description: Check if system is big-endian while mounting fs with extents feature
3069 Details : Mounting a filesystem with extents feature will fail on big-endian
3070 systems since ext3-based ldiskfs is not supported on big-endian
3071 systems. This can be over-riden with "bigendian_extents" mount option.
3073 Severity : enhancement
3075 Description: The root squash functionality
3076 Details : A security feature, which is to prevent users from being able
3077 to mount lustre on their desktop, run as root, and delete
3078 all of the files in the filesystem. The goal is accomplished by
3079 remapping user id (UID) and group id (GID) of the root user to
3080 a UID and GID specified by the system administartor via Lustre
3081 configuration management server (MGS). The functionality also
3082 allows to specify sets of clients for which the remapping does
3087 Description: Excessive recovery window
3088 Details : With AT enabled, the recovery window can be excessively long (6000+
3089 seconds). To address this problem, we no longer use
3090 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3091 timeout no longer depends on the service time, it is set to
3092 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3093 time via pb_service_time.
3097 Description: Watchdog triggered on MDS failover
3098 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3099 the OSTs know that the MDS "UUID" can be reused for the same export
3100 from a different NID, so we do not need to wait for the export to be
3104 Frequency : rare, only if using MMP with Linux RAID
3106 Description: MMP doesn't work with Linux RAID
3107 Details : While using HA for Lustre servers with Linux RAID, it is possible
3108 that MMP will not detect multiple mounts. To make this work we
3109 need to unplug the device queue in RAID when the MMP block is being
3110 written. Also while reading the MMP block, we should read it from
3111 disk and not the cached one.
3113 Severity : enhancement
3115 Description: open file using fid
3116 Details : A file can be opened using just its fid, like
3117 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3120 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3123 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3124 Details : add proper locking for creation of jbd_4k slab cache
3128 Description: MMP check in ext3_remount() fails without displaying any error
3129 Details : When multiple mount protection fails during remount, proper error
3132 Severity : enhancement
3134 Description: Allow stripe size to be up to 4G-64k
3135 Details : Fix math logic to allow large stripe sizes.
3139 Description: add check for >8TB ldiskfs filesystems
3140 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3141 Don't allow >8TB ldiskfs filesystems to be mounted without
3142 force_over_8tb mount option
3147 Descriptoin: OSS complains "too many blocks" and crashes
3148 Details : In filter_commitrw_write():
3149 1. lquota_chkquota();
3150 2. obdo_from_inode();
3151 3. lquota_pending_commit();
3152 "2." has changed uid and gid in this case, then caused
3153 lquota_chkquota() and lquota_pending_commit() to use
3154 different uid and gid. That is the root of the bug.
3158 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3159 Details : Detect and print out the 'stripe count = -1' case properly.
3163 Description: Out or order replies might be lost on replay
3164 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3165 request with tid smaller than one currently being added, add it
3166 to the start, not end of the list.
3168 --------------------------------------------------------------------------------
3170 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3172 * Support for kernels:
3173 2.6.5-7.283 (SLES 9),
3174 2.6.9-55.EL (RHEL 4),
3175 2.6.16.46-0.14 (SLES 10),
3176 2.6.18.8 vanilla (kernel.org)
3177 * Client support for unpatched kernels:
3178 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3179 2.6.16 - 2.6.22 vanilla (kernel.org)
3180 * Due to recently discovered recovery problems, we do not recommend
3181 using patchless RHEL 4 clients with this or any earlier release.
3182 * Recommended e2fsprogs version: 1.39.cfs8
3183 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3184 * Starting with this release, the ldiskfs backing filesystem required
3185 by Lustre is now in its own package, lustre-ldiskfs. This package
3186 should be installed. It is versioned separately from Lustre and
3187 may be released separately in future.
3191 Description: block reactivating mgc import until all deactivates complete
3192 Details : Fix race when failing back MDT/MGS to itself (testing)
3195 Frequency : at statup only
3197 Description: mds_lov_synchronize race leads to various problems
3198 Details : simultaneous MDT->OST connections at startup can cause the
3199 sync to abort, leaving the OSC in a bad state.
3201 Severity : enhancement
3203 Description: add optional extra BUILD_VERSION info
3204 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3205 to override the lustre version.
3208 Frequency : 2.6.18 servers only
3210 Description: ll_kern_mount() doesn't release the module reference
3211 Details : The ldiskfs module reference count never drops down to 0
3212 because ll_kern_mount() doesn't release the module reference.
3217 Description: server LBUG when using old ost_num_threads parameter
3218 Details : Accept the old ost_num_threads parameter but warn that it
3219 is deprecated, and fix an off-by-one error that caused an LBUG.
3224 Description: Transient SCSI error results in persistent IO issue
3225 Details : iobuf->dr_error is not reinitialized to 0 between two
3229 Frequency : sometimes when underlying device returns I/O errors
3231 Description: OSTs not going read-only during write failures
3232 Details : OSTs are not remounted read-only when the journal commit threads
3233 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3234 instead of the ext3 wrappers.
3238 Description: poor connect scaling with increasing client count
3239 Details : Don't run filter_grant_sanity_check for more than 100 exports
3240 to improve scaling for large numbers of clients.
3243 Frequency : SLES10 only
3245 Description: sanity-quota.sh quotacheck failed: rc = -22
3246 Details : Quotas cannot be enabled on SLES10.
3249 Frequency : liblustre clients only
3251 Description: getdirentries does not give error when run on compute nodes
3252 Details : getdirentries does not fail when the size specified as an argument
3253 is too small to contain at least one entry
3255 Severity : enhancement
3257 Description: Add LNET router traceability for debug purposes
3258 Details : If a checksum failure occurs with a router as part of the
3259 IO path, the NID of the last router that forwarded the bulk data
3260 is printed so it can be identified.
3265 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3266 Details : Due to a race condition, liblustre clients were occasionally
3267 evicted incorrectly.
3269 Severity : enhancement
3271 Description: lfs setstripe use optional parameters instead of postional
3274 Severity : enhancement
3276 Description: Nanosecond timestamp support for ldiskfs
3277 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3278 resolution timestamps. There is not yet support for this at
3279 the Lustre filesystem level.
3282 Frequency : during server recovery
3284 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3285 Details : request with rq_no_resend flag not awake l_wait_event if they get
3289 Frequency : nfs export on patchless client
3291 Description: connectathon hang when test nfs export over patchless client
3292 Details : Disconnected dentry cannot be found with lookup, so we do not need
3293 to unhash it or make it invalid
3296 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3297 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3298 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3303 Description: open req refcounting wrong on reconnect
3304 Details : If reconnect happened between getting open reply from server and
3305 call to mdc_set_replay_data in ll_file_open, we will schedule
3306 replay for unreferenced request that we are about to free.
3307 Subsequent close will crash in variety of ways.
3308 Check that request is still eligible for replay in
3309 mdc_set_replay_data().
3314 Description: disable writes to filesystem when reading health_check file
3315 Details : the default for reading the health_check proc file has changed
3316 to NOT do a journal transaction and write to disk, because this
3317 can cause reads of the /proc file to hang and block HA state
3318 checking on a healthy but otherwise heavily loaded system. It
3319 is possible to return to the previous behaviour during configure
3320 with --enable-health-write.
3322 Severity : enhancement
3324 Description: 64-bit inode version
3325 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3326 to the inode. This will be required for version-based recovery.
3331 Description: MDS fails to start if a duplicate client export is detected
3332 Details : in some rare cases it was possible for a client to connect to
3333 an MDS multiple times. Upon recovery the MDS would detect this
3334 and fail during startup. Handle this more gracefully.
3336 Severity : enhancement
3338 Description: Add -o localflock option to simulate old noflock
3340 Details : This will achieve local-only flock/fcntl locks
3346 Description: log_commit_thread vs filter_destroy race leads to crash
3347 Details : Take import reference before releasing llog record semaphore
3352 Description: Wrong request locking in request set processing
3353 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3354 imp_delayed_list, in this place should be used imp_lock.
3357 Frequency : when reconnection
3359 Description: Grant Leak when osc reconnect to OST
3360 Details : When osc reconnect ost, OST(filter) should check whether it
3361 should grant more space to client by comparing fed_grant and
3362 cl_avail_grant, and return the granted space to client instead
3363 of "new granted" space, because client will call osc_init_grant
3364 to update the client grant space info.
3367 Frequency : when client reconnect to OST
3369 Description: Grant Leak when osc do resend and replay bulk write
3370 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3371 bulk write request, because the grant info while be sync between
3372 OSC and OST when reconnect, and we should ignore the grant info
3373 these of resend/replay write req.
3378 Description: Grant space more than available left space sometimes.
3379 Details : When then OST is about to be full, if two bulk writing from
3380 different clients came to OST. Accord the avaliable space of the
3381 OST, the first req should be permitted, and the second one
3382 should be denied by ENOSPC. But if the seconde arrived before
3383 the first one is committed. The OST might wrongly permit second
3384 writing, which will cause grant space > available space.
3387 Frequency : when client is evicted
3389 Description: Grant might be wrongly erased when osc is evicted by OST
3390 Details : when the import is evicted by server, it will fork another
3391 thread ptlrpc_invalidate_import_thread to invalidate the
3392 import, where the grant will be set to 0. While the original
3393 thread will update the grant it got when connecting. So if
3394 the former happened latter, the grant will be wrongly errased
3395 because of this race.
3400 Description: Checking Stale with correct fid
3401 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3402 to check whether it is stale, because sometimes, we want the
3403 enqueue happened anyway, and op_data.fid2 will not be initialized.
3405 Severity : enhancement
3407 Description: update patchless client
3408 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3411 Frequency : only with 2.4 kernel
3413 Description: random memory corruption
3414 Details : size of struct ll_inode_info is to big for union inode.u and this
3415 can be cause of random memory corruption.
3420 Description: Memory leak in recovery
3421 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3422 It should also check obd_fail before fsfilt_start, otherwise if
3423 fsfilt_start return -EROFS,(failover mds during mds recovery).
3424 then the req will return with repmsg->transno = 0 and rc = EROFS.
3425 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3426 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3427 be freed no matter whether fsfilt_commit success or not.
3430 Frequency : only with huge count clients
3432 Description: Prevents from taking the superblock lock in llap_from_page for
3434 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3435 for llap_from_page call in ll_removepage prevents from taking the
3436 superblock lock for a soon died page.
3441 Description: Not check open intent error before release open handle
3442 Details : in some rare cases, the open intent error is not checked before
3443 release open handle, which may cause
3444 ASSERTION(open_req->rq_transno != 0), because it tries to release
3445 the failed open handle.
3450 Description: Set cat log bitmap only after create log success.
3451 Details : in some rare cases, the cat log bitmap is set too early. and it
3452 should be set only after create log success.
3456 Description: Accessing a block bevice can re-enable I/O when Lustre is
3457 tearing down a device.
3458 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3463 Description: service threads may hog cpus when there are a lot of requests
3464 Details : Insert cond_resched to give other threads a chance to use some CPU
3469 Description: the cat log was not initialized in recovery
3470 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3471 unlink log on mds will not be initialized until mds post
3472 recovery. And also in mds post recovery, the unlink log will
3473 initialization will be done asynchronausly, so there will be race
3474 between add unlink log and unlink log initialization.
3478 Description: brw_stats were being printed incorrectly
3479 Details : brw_stats were being printed as log2 but all of them were not
3480 recorded as log2. Also remove some code duplication arising from
3481 filter_tally_{read,write}.
3485 Frequency : rare, only in recovery.
3486 Description: ASSERTION(req->rq_type != LI_POISON) failed
3487 Details : imp_lock should be held while iterating over imp_sending_list for
3488 prevent destroy request after get timeout in ptlrpc_queue_wait.
3492 Description: replay-single.sh test 52 fails
3493 Details : A lock's skiplist need to be cleanup when it being unlinked
3494 from its resource list.
3498 Description: Short directio read returns full requested size rather than
3500 Details : Direct I/O operations should return actual amount of bytes
3501 transferred rather than requested size.
3503 Severity : enhancement
3505 Description: metadata RPC reduction (e.g. for rm performance)
3506 Details : decrease the amount of synchronous RPC between clients and servers
3507 by canceling conflicing lock before the operation on the client
3508 and packing thier handles into the main operation RPC to server.
3510 Severity : enhancement
3512 Description: Async OSC create to avoid the blocking unnecessarily.
3513 Details : If an OST has no remaining object, system will block on the
3514 creation when it needs to create a new object on this OST. Now,
3515 ways use pre-created objects when available, instead of blocking on
3516 an empty osc while others are not empty. If we must block, we
3517 block for the shortest possible period of time.
3521 Description: improve handling recoverable errors
3522 Details : if request processig with error which can be recoverable on server
3523 request should be resend, otherwise page released from cache and
3526 Severity : enhancement
3528 Description: refine locking for avoid write wrong info into lov_objid file
3529 Details : fix possible races with add new target and write/update data in
3532 --------------------------------------------------------------------------------
3534 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3539 Frequency : on some architectures
3541 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3542 Details : Uninitialized flags sometimes cause configuration commands to
3546 Frequency : patchless clients only
3548 Description: missing __iget() symbol export
3549 Details : The __iget() symbol export is missing. To avoid the need for
3550 this on patchless clients the deathrow inode reaper is turned
3551 off, and we depend on the VM to clean up old inodes. This
3552 dependency was during via the fix for bug 12181.
3554 --------------------------------------------------------------------------------
3556 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3558 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3559 INTEROPERATE with older versions automatically. In many cases a
3560 special upgrade step is needed. Please read the
3561 user documentation before upgrading any part of a live system.
3562 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3563 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3564 * WARNING: Lustre configuration and startup changes are required with
3565 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3569 Severity : enhancement
3571 Description: MountConf
3572 Details : Lustre configuration is now managed via mkfs and mount
3573 commands instead of lmc and lconf. New obd types (MGS, MGC)
3574 are added for dynamic configuration management. See
3575 https://mail.clusterfs.com/wikis/lustre/MountConf for
3578 Severity : enhancement
3580 Description: dynamic OST addition
3581 Details : OSTs can now be added to a live filesystem
3583 Severity : enhancement
3585 Description: startup order invariance
3586 Details : MDTs and OSTs can be started in any order. Clients only
3587 require the MDT to complete startup.
3589 Severity : enhancement
3591 Description: parallel, asynchronous orphan cleanup
3592 Details : orphan cleanup is now performed in separate threads for each
3593 OST, allowing parallel non-blocking operation.
3595 Severity : enhancement
3597 Description: optimized stripe assignment
3598 Details : stripe assignments are now made based on ost space available,
3599 ost previous usage, and OSS previous usage, in order to try
3600 to optimize storage space and networking resources.
3602 Severity : enhancement
3604 Description: Permanently set tunables
3605 Details : All writable /proc/fs/lustre tunables can now be permanently
3606 set on a per-server basis, at mkfs time or on a live system.
3608 Severity : enhancement
3610 Description: Lustre message v2
3611 Details : Add lustre message format v2.
3613 Severity : enhancement
3615 Description: client OST exclusion list
3616 Details : Clients can be started with a list of OSTs that should be
3617 declared "inactive" for known non-responsive OSTs.
3621 Description: SPEC SFS validation failure on NFS v2 over lustre.
3622 Details : Changes the blocksize for regular files to be 2x RPC size,
3623 and not depend on stripe size.
3625 Severity : enhancement
3627 Description: Multiple MD RPCs in flight.
3628 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3629 To avoid overly-overloading MDS, introduce a limit on number of
3630 MDS RPCs in flight for a single client and add /proc controls
3631 to adjust this limit.
3633 Severity : enhancement
3635 Description: client read/write statistics
3636 Details : Add client read/write call usage stats for performance
3637 analysis of user processes.
3638 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3639 file access. extents_stats shows chunk size distribution.
3640 extents_stats_per_process show chunk size distribution per
3643 Severity : enhancement
3645 Description: mds statistics
3646 Details : Add detailed mds operations statistics in
3647 /proc/fs/lustre/mds/*/stats.
3651 Description: Failure of copying files with lustre special EAs.
3652 Details : Client side always return success for setxattr call for lustre
3653 special xattr (currently only "trusted.lov").
3658 Description: Refcount LNET uuids
3659 Details : The global LNET uuid list grew linearly with every startup;
3660 refcount repeated list entries instead of always adding to
3663 Severity : enhancement
3665 Description: Dynamic service threads
3666 Details : Within a small range, start extra service threads
3667 automatically when the request queue builds up.
3670 Frequency : mixed-endian client/server environments
3672 Description: mixed-endian crashes
3673 Details : The new msg_v2 system had some failures in mixed-endian
3676 Severity : enhancement
3678 Description: Easy OST removal
3679 Details : OSTs can be permanently deactivated with e.g. 'lctl
3680 conf_param lustre-OST0001.osc.active=0'
3682 Severity : enhancement
3684 Description: MGS proc entries
3685 Details : Added basic proc entries for the MGS showing what filesystems
3688 Severity : enhancement
3690 Description: provide MGS failover
3691 Details : Added config lock reacquisition after MGS server failover.
3693 Severity : enhancement
3695 Description: add Linux 2.4 support
3696 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3700 Description: a large application tries to do I/O to the same resource and dies
3701 in the middle of it.
3702 Details : Check the req->rq_arrival time after the call to
3703 ost_brw_lock_get(), but before we do anything about
3704 processing it & sending the BULK transfer request. This
3705 should help move old stale pending locks off the queue as
3706 quickly as obd_timeout.
3709 Frequency : when an incorrect nid is specified during startup
3711 Description: ptlrpc connect to non-existant node causes kernel crash
3712 Details : LNET can't be re-entered from an event callback, which
3713 happened when we expire a message after the export has been
3714 cleaned up. Instead, hand the zombie cleanup off to another
3717 Severity : enhancement
3719 Description: plain/inodebits lock performance improvement
3720 Details : Grouping plain/inodebits in granted list by their request modes
3721 and bits policy, thus improving the performance of search through
3725 Frequency : only if OST filesystem is corrupted
3727 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3728 Details : for a short time RPCs with bulk IO are in the replay list,
3729 but replay of bulk IOs is unimplemented. If the OST filesystem
3730 is corrupted due to disk cache incoherency and then replay is
3731 started it is possible to trip an assertion. Avoid putting
3732 committed RPCs into the replay list at all to avoid this issue.
3735 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3737 Description: System hang on startup
3738 Details : This bug allowed the liblustre (e.g. catamount) client to
3739 return to the app before handling all startup RPCs. This
3740 could leave the node unresponsive to lustre network traffic
3741 and manifested as a server ptllnd timeout.
3743 Severity : enhancement
3745 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3746 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3747 boolean to control whether to print peer debug info when a
3748 client's RPC times out.
3751 Frequency : only for kernels with patches from Lustre below 1.4.3
3753 Description: Remove old rdonly API
3754 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3757 Frequency : only for devices with external journals
3759 Description: Set external device read-only also
3760 Details : During a commanded failover stop, we set the disk device
3761 read-only while the server shuts down. We now also set any
3762 external journal device read-only at the same time.
3765 Frequency : when upgrading from 1.4 while trying to change parameters
3767 Description: The wrong (new) MDC name was used when setting parameters for
3768 upgraded MDT's. Also allows changing of OSC (and MDC)
3769 parameters if --writeconf is specified at tunefs upgrade time.
3772 Frequency : when setting specific ost indicies
3774 Description: QOS code breaks on skipped indicies
3775 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3776 created with --index need not be sequential.
3778 Severity : enhancement
3780 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3781 Details : The uninit_groups feature works in conjunction with the kernel
3782 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3783 up the pass1 processing of e2fsck. This is a read-only feature
3784 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3785 mount filesystems that have had this feature enabled.
3787 Severity : enhancement
3789 Description: Improve multi-block allocation algorithm to avoid fragmentation
3790 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3791 allocation locality and avoid filesystem fragmentation.
3793 ------------------------------------------------------------------------------
3795 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3797 * Support for kernels:
3798 2.4.21-47.0.1.EL (RHEL 3)
3799 2.6.5-7.283 (SLES 9)
3800 2.6.9-42.0.10.EL (RHEL 4)
3801 2.6.12.6 vanilla (kernel.org)
3802 2.6.16.27-0.9 (SLES 10)
3803 * Recommended e2fsprogs version: 1.39.cfs5
3805 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3809 Frequency : occasional, depends on client load and configuration
3810 Bugzilla : 12181, 12203
3811 Description: data loss for recently-modified files
3813 Details : In some cases it is possible that recently written or created
3814 files may not be written to disk in a timely manner (this should
3815 normally be within 30s unless client IO load is very high).
3816 The problem appears as zero-length files or files that are a
3817 multiple of 1MB in size after a client crash or client eviction
3818 that are missing data at the end of the file.
3820 This problem is more likely to be hit on clients where files are
3821 repeatedly created and unlinked in the same directory, clients
3822 have a large amount of RAM, have many CPUs, the filesystem has
3823 many OSTs, the clients are rebooted frequently, and/or the files
3824 are not accessed by other nodes after being written.
3826 The presence of the problem can be detected by looking at
3827 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3828 smaller than the second (nr_unused) then dirty files will not
3829 be flushed automatically to disk. "sync; sleep 10" should be
3830 run several times on the node before unmounting it to update
3831 Lustre (this is also safe to run on nodes without this problem).
3833 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3834 that can cause this same problem, so customers using that kernel
3835 also need to update the kernel in addition to Lustre. In order
3836 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3838 It is normal that files written just before a client crash (less
3839 than 30s) may not yet have been flushed to disk, even for local
3843 Frequency : frequent on thin XT3 nodes
3845 Description: UUID collision on thin XT3 Linux nodes
3846 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3847 randomly, since we relied on an insufficiently-seeded PRNG.
3852 Description: OSS hangs after "All ost request buffers busy"
3853 Details : A deadlock between quota and journal operations caused OSS
3854 hangs after printing "All ost request buffers busy."
3857 Frequency : always on liblustre builds
3859 Description: Cleanup compiler warnings on liblustre
3862 Frequency : always on liblustre builds on XT3
3864 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3868 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3869 > 32000 subdirectories
3870 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3871 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3872 subdirectory count crosses 32000. This will aid e2fsck to
3873 correctly handle more than 32000 subdirectories.
3876 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3878 Description: System hang on startup
3879 Details : This bug allowed the liblustre (e.g. catamount) client to
3880 return to the app before handling all startup RPCs. This
3881 could leave the node unresponsive to lustre network traffic
3882 and manifested as a server ptllnd timeout.
3884 Severity : enhancement
3886 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3887 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3888 boolean to control whether to print peer debug info when a
3889 client's RPC times out.
3894 Description: make O_SYNC working on 2.6 kernels
3895 Details : 2.6 kernels use different method for mark pages for write,
3896 so need add a code to lustre for O_SYNC work.
3901 Description: Failure to close file and release space on NFS
3902 Details : Put inode details into lock acquired in ll_intent_file_open.
3903 Use mdc_intent_lock in ll_intent_open to properly
3904 detect all kind of errors unhandled by mdc_enqueue.
3909 Description: proc file read during shutdown sometimes raced obd removal,
3911 Details : Add lock to prevent obd access after proc file removal.
3914 Frequency : Only for files larger than 4GB on 32-bit clients.
3916 Description: improperly doing page alignment of locks
3917 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3918 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3923 Description: RPCs being resent when they shouldn't be
3924 Details : Some RPCs that should not be resent are being resent. This
3925 can cause inconsistencies in the RPC state machine. Do not
3926 resend such requests.
3929 Frequency : rare, only with NFS export
3931 Description: Crash on NFS re-export node
3932 Details : under very unusual load conditions an assertion is hit in
3933 ll_intent_file_open()
3936 Frequency : only if OST filesystem is corrupted
3938 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3939 Details : for a short time RPCs with bulk IO are in the replay list,
3940 but replay of bulk IOs is unimplemented. If the OST filesystem
3941 is corrupted due to disk cache incoherency and then replay is
3942 started it is possible to trip an assertion. Avoid putting
3943 committed RPCs into the replay list at all to avoid this issue.
3948 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3949 Details : Large single O_DIRECT read and write calls can fail to allocate
3950 a sufficiently large buffer to process the request. In case of
3951 allocation failure the allocation is retried with a smaller
3952 buffer and broken into smaller requests.
3954 Severity : enhancement
3956 Description: Add -o localflock option to simulate old noflock behaviour.
3957 Details : This will achieve local-only flock/fcntl locks coherentness.
3962 Description: versioning check is incomplete
3963 Details : Checking the version difference of client vs. server, report
3964 error if the gap is too big.
3969 Description: add support PG_writeback bit
3970 Details : add support for PG_writeback bit for Lustre, for more carefull
3971 work with page cache in 2.6 kernel. This also fix some deadlocks
3972 and remove hack for work O_SYNC with 2.6 kernel.
3974 Severity : enhancement
3976 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3977 Details : The uninit_groups feature works in conjunction with the kernel
3978 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3979 up the pass1 processing of e2fsck. This is a read-only feature
3980 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3981 mount filesystems that have had this feature enabled.
3983 Severity : enhancement
3985 Description: Improve multi-block allocation algorithm to avoid fragmentation
3986 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3987 allocation locality and avoid filesystem fragmentation.
3989 ------------------------------------------------------------------------------
3991 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3993 * Support for kernels:
3994 2.6.9-42.0.3.EL (RHEL 4)
3995 2.6.5-7.276 (SLES 9)
3996 2.4.21-47.0.1.EL (RHEL 3)
3997 2.6.12.6 vanilla (kernel.org)
3998 2.6.16.21-0.8 (SLES10)
3999 * Recommended e2fsprogs version: 1.39.cfs2-0
4001 * The backwards-compatible /proc/sys/portals symlink has been removed
4002 in this release. Before upgrading, please ensure that you change
4003 any configuration scripts or /etc/sysctl.conf files that access
4004 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4005 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4006 in advance of the upgrade on any system running Lustre 1.4.6 or
4007 newer, since /proc/sys/lnet was added in that version.
4008 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4012 Frequency : only when quota is used
4014 Description: avoid scanning export list for quota master
4015 Details : Change the algorithms to avoid scanning export list in order
4016 to improve the efficiency.
4019 Frequency : MDS failover only, very rarely
4021 Description: "went back in time" messages on mds failover
4022 Details : The greatest transno may be lost when the current operation
4023 finishes with an error (transno==0) and the client's last_rcvd
4024 record is over-written. Save the greatest transno in the
4025 mds_last_transno for this case.
4028 Frequency : always for specific kernels and striping counts
4030 Description: client may get "Matching packet too big" without ACL support
4031 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4032 when trying to access files in certain configurations. The
4033 clients should in fact be denied when mounting because they do
4034 not understand ACLs.
4037 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4039 Description: many clients connecting with IO in progress causes connect timeouts
4040 Details : Avoid synchronous journal commits to avoid delays caused by many
4041 clients connecting/disconnecting when bulk IO is in progress.
4042 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4043 OST_IO_PORTAL to avoid delays behind potentially many pending
4047 Frequency : occasionally with multiple writers to a single file
4049 Description: shared writes to file may result in wrong size reported by stat()
4050 Details : Allow growing of kms when extent lock is cancelled
4053 Frequency : always with random mmap IO to multi-striped file
4055 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4056 Details : Only if the hole is at the end of OST object so that kms is too
4057 small. Fix is to increase kms accordingly in ll_nopage.
4060 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4062 Description: writes to a missing object would leak memory on the OST
4063 Details : If there is an inconsistency between the MDS and OST filesystems,
4064 such that the MDS references an object that doesn't exist, writes
4065 to that object will leak memory due to incorrect cleanup in the
4066 error handling path, eventually running out of memory on the OST.
4071 Description: Creating too long symlink causes lustre errors
4072 Details : Check symlink and name lengths before sending requests to MDS.
4075 Frequency : only if flock is enabled (not on by default)
4077 Description: posix locks not released on fd closure on 2.6.9+
4078 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4079 kernels, this caused such locks not to be released on fd close and
4080 then assertions on fs unmount about still used locks.
4083 Frequency : MDS failover only, very rarely
4085 Description: clients may get ASSERTION(granted_lock != NULL)
4086 Details : When request was taking a long time, and a client was resending
4087 a getattr by name lock request. The were multiple lock requests
4088 with the same client lock handle and
4089 mds_getattr_name->fixup_handle_for_resent_request found one of the
4090 lock handles but later failed with ASSERTION(granted_lock != NULL).
4095 Description: handle->h_buffer_credits > 0, assertion failure
4096 Details : h_buffer_credits is zero after truncate, causing assertion
4097 failure. This patch extends the transaction or creates a new
4101 Frequency : NFS re-export or patchless client
4102 Bugzilla : 11179, 10796
4103 Description: Crash on NFS re-export node (__d_move)
4104 Details : We do not want to hash the dentry if we don't have a lock.
4105 But if this dentry is later used in d_move, we'd hit uninitialised
4106 list head d_hash, so we just do this to init d_hash field but
4107 leave dentry unhashed.
4110 Frequency : NFS re-export or patchless client
4112 Description: NFS exports has problem with symbolic link
4113 Details : lustre client didn't properly install dentry when re-exported
4114 to NFS or running patchless client.
4117 Frequency : NFS re-export or patchless client
4119 Description: Various nfs/patchless fixes.
4120 Details : fixes reuse disconected alias for lookup process - this fixes
4121 warning "find_exported_dentry: npd != pd",
4122 fix permission error with open files at nfs.
4123 fix apply umask when do revalidate.
4126 Frequency : occasional
4128 Description: Crash on NFS re-export node
4129 Details : calling clear_page() on the wrong pointer triggered oops in
4130 generic_mapping_read().
4133 Frequency : rarely, using O_DIRECT IO
4135 Description: unaligned directio crashes client with LASSERT
4136 Details : check for unaligned buffers before trying any requests.
4139 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4141 Description: stale data returned from RAID cache
4142 Details : If only a small amount of IO is done to the RAID device before
4143 reading it again it is possible to get stale data from the RAID
4144 cache instead of reading it from disk.
4147 Frequency : always for sles10 kernel
4149 Description: sles10 support
4150 Details : ll_follow_link: compile fixes and using of nd_set_link
4151 under newer kernels.
4154 Frequency : depends on arch, kernel and compiler version, always on sles10
4157 Description: recursive or deep enough symlinks cause stack overflow
4158 Details : getting rid of large stack-allocated variable in
4162 Frequency : depends on hardware
4164 Description: lustre write performance loss in the SLES10 kernel
4165 Details : the performance loss is caused by using of write barriers in the
4166 ext3 code. The SLES10 kernel turns barrier support on by
4167 default. The fix is to undo that change for ldiskfs.
4169 ------------------------------------------------------------------------------
4171 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4173 * Support for kernels:
4174 2.6.9-42.0.3EL (RHEL 4)
4175 2.6.5-7.276 (SLES 9)
4176 2.4.21-47.0.1.EL (RHEL 3)
4177 2.6.12.6 vanilla (kernel.org)
4181 Frequency : quota enabled and large files being deleted
4183 Description: releasing more than 4GB of quota at once hangs OST
4184 Details : If a user deletes more than 4GB of files on a single OST it
4185 will cause the OST to spin in an infinite loop. Release
4186 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4191 Description: statfs data retrieved from /proc may be stale or zero
4192 Details : When reading per-device statfs data from /proc, in the
4193 {kbytes,files}_{total,free,avail} files, it may appear
4194 as zero or be out of date.
4197 Frequency : systems with MD RAID1 external journal devices
4199 Description: lconf's call to blkid is confused by RAID1 journal devices
4200 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4201 of returning all block devices that match the journal UUID.
4204 Frequency : always, for aggregate stripe size over 4GB
4206 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4207 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4208 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4211 Frequency : always if "lfs find" used on a local file/directory
4213 Description: "lfs find" segfaults if used on a local file/directory
4214 Details : The case where a directory component was not specified wasn't
4215 handled correctly. Handle this properly.
4218 Frequency : always on ppc64
4220 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4221 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4222 perform endianness conversion.
4225 Frequency : rarely (truncate to non-zero file size after write under load)
4226 Bugzilla : 10730, 10687
4227 Description: Files padded with zeros to next 4K multiple
4228 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4229 it is possible that files that are truncated to a non-zero size
4230 immediately after being written are filled with zero bytes beyond
4231 the truncated size. No file data is lost.
4233 Severity : enhancement
4235 Description: Allow recovery/failover for liblustre clients.
4236 Details : liblustre clients were unaware of failover configurations until
4239 Severity : enhancement
4241 Description: user file locks should fail when not mounting with flock option
4242 Details : Set up an error-returning stub in ll_file_operations.lock field
4243 to prevent incorrect behaviour when client is mounted without
4244 flock option. Also, set up properly f_op->flock field for
4248 Frequency : always on ia64
4250 Description: "lfs df" loops on printing out MDS statfs information
4251 Details : The obd_ioctl_data was not initialized and in some systems
4252 this caused a failure during the ioctl that did not return
4253 an error. Initialize the struct and return an error on failure.
4256 Frequency : SLES 9 only
4258 Description: Error of copying files with lustre special EAs as root
4259 Details : Client side always return success for setxattr call for lustre
4260 special xattr (currently only "trusted.lov").
4263 Frequency : rarely on clusters with both ia64+i386 clients
4265 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4266 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4267 clients are doing concurrent writes to the same file it is
4268 possible that the ia64 clients may LASSERT because the OST
4269 extent locks are not PAGE_SIZE aligned. Ensure that grown
4270 locks are always aligned on the request boundary.
4273 Frequency : specific use, occasional
4275 Description: Overwriting in use executable truncates on-disk binary image
4276 Details : If one node attempts to overwrite an executable in use by
4277 another node, we now correctly return ETXTBSY instead of
4278 truncating the file.
4280 Severity : enhancement
4282 Description: Async OSC create to avoid the blocking unnecessarily.
4283 Details : If an OST has no remaining object, system will block on the creating
4284 when need to create a new object on this OST. Now, ways use
4285 pre-created objects when available, instead of blocking on an
4286 empty osc while others are not empty. If we must block, we block
4287 for the shortest possible period of time.
4292 Description: chmod on Lustre root is propagated to other clients
4293 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4294 invalid by the follow_mount time.
4297 Frequency : liblustre clients only
4299 Description: Race in 'instant cancel' lock handling could lead to such locks
4300 never to be granted in case of SMP MDS
4301 Details : Do not destroy not yet granted but cbpending locks in
4305 Frequency : replay/resend of open
4307 Description: non null lock assetion failure in mds_intent_policy
4308 Details : Trying to replay/resend lockless open requests resulted in
4309 mds_open() returning 0 with no lock. Now it sets a flag if
4310 a lock is going to be returned.
4312 Severity : enhancement
4314 Description: Checksum enhancements
4315 Details : New checksum enhancements allow for resending RPCs that failed
4318 Severity : enhancement
4320 Description: Tunables on number of dirty pages in cacche
4321 Details : Allow to set limit on number of dirty pages cached.
4326 Description: client crash on unmount - lock still has references
4327 Details : In some error handling cases it was possible to leak a lock
4328 reference on a client while accessing a file. This was not
4329 harmful to the client during operation, but would cause the
4330 client to crash when the filesystem is unmounted.
4333 Frequency : specific case, rare
4335 Description: ETXTBSY on mds though file not in use by client
4336 Details : ETXTBSY is no longer incorrectly returned when attempting to
4337 chmod or chown a directory that the user previously tried to
4338 execute or a currently-executing binary.
4341 Frequency : extremely rare except on liblustre-based clients
4343 Description: Lustre space not freed when files are deleted
4344 Details : Clean up open-unlinked files after client eviction. Previously
4345 the unlink was skipped and the files remained as orphans.
4350 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4351 Details : In some cases it was possible to send a blocking callback to
4352 a client doing a glimpse, even though that client didn't get
4353 a lock granted. When the glimpse lock is cancelled on the OST
4354 the freed lock is left on the waiting list and corrupted the list.
4357 Frequency : all core dumps
4359 Description: Broke core dumps to lustre
4360 Details : Negative dentry may be unhashed if parent does not have UPDATE
4361 lock, but some callers, e.g. do_coredump, expect dentry to be
4362 hashed after successful create, hash it in ll_create_it.
4364 ------------------------------------------------------------------------------
4366 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4368 * Support for kernels:
4369 2.6.9-42.0.2.EL (RHEL 4)
4370 2.6.5-7.276 (SLES 9)
4371 2.4.21-40.EL (RHEL 3)
4372 2.6.12.6 vanilla (kernel.org)
4376 Frequency : always on RHEL 3
4378 Description: Number of open files grows over time
4379 Details : The number of open files grows over time, whether or not
4380 Lustre is started. This was due to a filp leak introduced
4381 by one of our kernel patches.
4383 ------------------------------------------------------------------------------
4385 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4387 * Support for kernels:
4388 2.6.9-42.EL (RHEL 4)
4389 2.6.5-7.276 (SLES 9)
4390 2.4.21-40.EL (RHEL 3)
4391 2.6.12.6 vanilla (kernel.org)
4396 Bugzilla : 5719, 9635, 9792, 9684
4397 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4398 Details : If a server is under heavy load and cannot reply to new
4399 connection requests before the client resends the (re)connect,
4400 the connection handling code can behave badly if two service
4401 threads are concurrently handing separate (re)connections from
4402 the same client. Add better locking to the connection handling
4403 code, and ensure that only a single connection will be processed
4404 for a given client UUID, even if the lock is dropped.
4406 Severity : enhancement
4408 Description: add TCP zero-copy support to kernel
4409 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4410 sends if the hardware supports scatter-gather and checksumming.
4411 This allows socklnd to do client-write and server-read more
4412 efficiently and reduce CPU utilization from skbuf copying.
4415 Frequency : only if NFS exporting from client
4417 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4418 Details : If setting attributes on a file created under NFS that had
4419 never been opened it would be possible to oops the client
4420 if the file had no objects.
4423 Frequency : always for liblustre
4425 Description: liblustre client does MDS+OSTs setattr RPC for each write
4426 Details : When doing a write from a liblustre client, the client
4427 incorrectly issued an RPC to the MDS and each OST the file was
4428 striped over in order to update the timestamps. When writing
4429 with small chunks and many clients this could overwhelm the MDS
4430 with RPCs. In all cases it would slow down the write because
4431 these RPCs are unnecessary.
4433 Severity : enhancement
4435 Description: allow number of MDS service threads to be changed at module load
4436 Details : It is now possible to change the number of MDS service threads
4437 running. Adding "options mds mds_num_threads={N}" to the MDS's
4438 /etc/modprobe.conf will set the number of threads for the next
4439 time Lustre is restarted (assuming the "mds" module is also
4440 reloaded at that time). The default number of threads will
4441 stay the same, 32 for most systems.
4446 Description: OST crash if filesystem is unformatted or corrupt
4447 Details : If an OST is started on a device that has never been formatted
4448 or if the filesystem is corrupt and cannot even mount then the
4449 error handling cleanup routines would dereference a NULL pointer.
4454 Description: NULL pointer deref in llap_from_page.
4455 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4456 incorrect) mapping if the page was truncated/reclaimed while it was
4457 searched for. Check for this condition and skip such pages when
4458 doing readahead. Introduce extra check to llap_from_page() to
4459 verify page->mapping->host is non-NULL (so page is not anonymous).
4462 Frequency : Sometimes when using sys_sendfile
4464 Description: "page not covered by a lock" warnings from ll_readpage
4465 Details : sendfile called ll_readpage without right page locks present.
4466 Now we introduced ll_file_sendfile that does necessary locking
4467 around call to generic_file_sendfile() much like we do in
4471 Frequency : with certain MDS communication failures at client mount time
4473 Description: NULL pointer deref after failed client mount
4474 Details : a client connection request may delayed by the network layer
4475 and not be sent until after the PTLRPC layer has timed out the
4476 request. If the client fails the mount immediately it will try
4477 to clean up before the network times out the request. Add a
4478 reference from the request import to the obd device and delay
4479 the cleanup until the network drops the request.
4482 Frequency : occasionally during client (re)connect
4484 Description: assertion failure during client (re)connect
4485 Details : processing a client connection request may be delayed by the
4486 client or server longer than the client connect timeout. This
4487 causes the client to resend the connection request. If the
4488 original connection request is replied in this interval, the
4489 client may trip an assertion failure in ptlrpc_connect_interpret()
4490 which thought it would be the only running connect process.
4493 Frequency : only with obd_echo servers and clients that are rebooted
4495 Description: kernel BUG accessing uninitialized data structure
4496 Details : When running an obd_echo server it did not start the ping_evictor
4497 thread, and when a client was evicted an uninitialized data
4498 structure was accessed. Start the ping_evictor in the RPC
4499 service startup instead of the OBD startup.
4501 Severity : enhancement
4502 Bugzilla : 10193 (patchless)
4503 Description: Remove dependency on various unexported kernel interfaces.
4504 Details : No longer need reparent_to_init, exit_mm, exit_files,
4505 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4508 Frequency : rare (only users of deprecated and unsupported LDAP config)
4510 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4511 Details : LDAP apparently contains 'lustreName' attributes instead of
4512 'name'. A simple remapping of the name is sufficient.
4515 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4517 Description: waiting_locks_callback trips kernel BUG if client is evicted
4518 Details : Running with the dump_on_timeout debug flag turned on makes
4519 it possible that the waiting_locks_callback() can try to dump
4520 the Lustre kernel debug logs from an interrupt handler. Defer
4521 this log dumping to the expired_lock_main() thread.
4523 Severity : enhancement
4525 Description: Support NFS exporting on 2.6 kernels.
4526 Details : Implement non-rawops metadata methods for NFS server to use without
4527 changing NFS server code.
4530 Frequency : very rare (synthetic metadata workload only)
4532 Description: two racing renames might cause an MDS thread to deadlock
4533 Details : Running the "racer" program may cause one MDS thread to rename
4534 a file from being the source of a rename to being the target of
4535 a rename at exactly the same time that another thread is doing
4536 so, and the second thread has already enqueued these locks after
4537 doing a lookup of the target and is trying to relock them in
4538 order. Ensure that we don't try to re-lock the same resource.
4541 Frequency : only very large systems with liblustre clients
4543 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4544 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4545 all OSTs in parallel. This allows the eviction of stale liblustre
4546 clients to proceed much faster than if they were done in series,
4547 and also offers similar improvements for other set_info RPCs.
4552 Description: excessive CPU usage during initial read phase on client
4553 Details : During the initial read phase on a client, it would agressively
4554 retry readahead on the file, consuming too much CPU and impacting
4555 performance (since 1.4.5.8). Improve the readahead algorithm
4556 to avoid this, and also improve some other common cases (read
4557 of small files in particular, where "small" is files smaller than
4558 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4563 Description: MDS crash when receiving packet with unknown intent.
4564 Details : Do not LBUG in unknown intent case, just return -EFAULT
4566 Severity : enhancement
4567 Bugzilla : 9293, 9385
4568 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4569 Details : Do not serialize getattr (non-intent version) and statfs.
4572 Frequency : occasional, when OST network is overloaded/intermittent
4574 Description: client evicted by OST after bulk IO timeout
4575 Details : If a client sends a bulk IO request (read or write) the OST
4576 may evict the client if it is unresposive to its data GET/PUT
4577 request. This is incorrect if the network is overloaded (takes
4578 too long to transfer the RPC data) or dropped the OST GET/PUT
4579 request. There is no need to evict the client at all, since
4580 the pinger and/or lock callbacks will handle this, and the
4581 client can restart the bulk request.
4584 Frequency : Always when mmapping file with no objects
4586 Description: client crashes when mmapping file with no objects
4587 Details : Check that we actually have objects in a file before doing any
4588 operations on objects in ll_vm_open, ll_vm_close and
4594 Description: Request leak when working with deleted CWD
4595 Details : Introduce advanced request refcount tracking for requests
4596 referenced from lustre intent.
4598 Severity : Enhancement
4600 Description: Cache open file handles on client.
4601 Details : MDS now will return special lock along with openhandle, if
4602 requested and client is allowed to hold openhandle, even if unused,
4603 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4604 closing files for every read/write openration.
4606 Severity : Enhancement
4608 Description: Cache open negative dentries on client when possible.
4609 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4610 negative dentries on lock revocation.
4615 Description: Remounting a client read-only wasn't possible with a zconf mount
4616 Details : It wasn't possible to remount a client read-only with llmount.
4618 Severity : enhancement
4619 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4620 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4621 MPICH is included to add Lustre-specific ADIO interfaces.
4622 This is based closely on the UFS ADIO layer and only differs
4623 in file creation, in order to allow the OST striping to be set.
4624 This is user-contributed code and not supported by CFS.
4629 Description: extended inode attributes (immutable, append-only) work improperly
4630 when 2.4 and 2.6 kernels are used on client/server or vice versa
4631 Details : Introduce kernel-independent values for these flags.
4633 Severity : enhancement
4636 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4637 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4638 granularity. Now, Fractional MB granularity is be supported,
4639 this is very useful for low memory system.
4641 Severity : enhancement
4643 Description: Getattr by fid
4644 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4645 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4646 have locks and avoids getattr rpc requests for it on every stat.
4649 Frequency : Always, for filesystems larger than 2TB
4651 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4652 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4653 filesystems larger than 2TB to be mounted with mballoc enabled.
4656 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4658 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4659 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4660 filesystems larger than 2TB is mounted then the kernel will
4661 silently corrupt the start of the filesystem. CONFIG_LBD is
4662 enabled for all CFS-supported kernels, but the possibility of
4663 this happening with a modified kernel config exists.
4665 Severity : enhancement
4667 Description: add client O_DIRECT support for 2.6 kernels
4668 Details : It is now possible to do O_DIRECT reads and writes to files
4669 in the Lustre client mountpoint on 2.6 kernel clients.
4671 Severity : enhancement
4673 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4674 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4675 parallel, not waiting for response from every OST before sending
4676 a rpc to the next OST.
4681 Description: setattr vs write race when updating file timestamps
4682 Details : Client processes that update a file timestamp into the past
4683 right after writing to the file (e.g. tar) it is possible that
4684 the updated file modification time can be reset to the current
4685 time due to a race between processing the setattr and write RPC.
4687 Severity : enhancement
4689 Description: Bring 'lfs find' closer in line with regular Linux find.
4690 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4691 -print0 options and obtains all the needed info through the lustre
4694 Severity : enhancement
4696 Description: support up to 1024 configured devices on one node
4697 Details : change obd_dev array from statically allocated to dynamically
4698 allocated structs as they are first used to reduce memory usage
4703 Description: Flush dirty partially truncated pages during truncate
4704 Details : Immediatelly flush partially truncated pages in filter_setattr,
4705 this way we completely avoid having any pages in page cache on OST
4706 and can retire ugly workarounds during writes to flush such pages.
4711 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4712 Details : protect inode from truncation within vfs_unlink() context
4713 just take a reference before calling vfs_unlink() and release it
4714 when parent's i_sem is free.
4719 Description: last_id value checked outside lock on OST caused LASSERT failure
4720 Details : If there were multiple MDS->OST object precreate requests in
4721 flight, it was possible that the OST's last object id was checked
4722 outside a lock and incorrectly tripped an assertion. Move checks
4723 inside locks, and discard old precreate requests.
4726 Frequency : always, if extents are used on OSTs
4728 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4729 Details : OSTs using the extents format would not zero the high 16 bits of
4730 the index physical block number. This is not a problem for any
4731 OST filesystems smaller than 16TB, and no kernels support ext3
4732 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4733 new/modified files) and can be fixed for existing filesystems
4734 with e2fsprogs-1.39-cfs1.
4739 Description: import connection selection may be incorrect if timer wraps
4740 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4741 import connections to be ignored if the 32-bit jiffies counter
4742 wraps. Use a 64-bit jiffies counter.
4745 Frequency : very large clusters immediately after boot
4747 Description: LNET request buffers exhausted under heavy short-term load
4748 Details : If a large number of client requests are generated on a service
4749 that has previously never seen so many requests it is possible
4750 that the request buffer growth cannot keep up with the spike in
4751 demand. Instead of dropping incoming requests, they are held in
4752 the LND until the RPC service can accept more requests.
4755 Frequency : Sometimes during replay
4757 Description: Assertion failure in ll_local_open after replay.
4758 Details : If replay happened on an open request reply before we were able
4759 to set replay handler, reply will become not swabbed tripping the
4760 assertion in ll_local_open. Now we set the handler right after
4761 recognising of open request
4764 Frequency : very rare
4766 Description: kernel reports "badness in vsnprintf"
4767 Details : Reading from the "recovery_status" /proc file in small chunks
4768 may cause a negative length in lprocfs_obd_rd_recovery_status()
4769 call to vsnprintf() (which is otherwise harmless). Exit early
4770 if there is no more space in the output buffer.
4772 Severity : enhancement
4774 Description: clear OBD RPC statistics by writing to them
4775 Details : It is now possible to clear the OBD RPC statistics by writing
4776 to the "stats" file.
4781 Description: Client mtime is not the same on different clients after utimes
4782 Details : In some cases, the client was using the utimes() syscall on
4783 a file cached on another node. The clients now validate the
4784 ctime from the MDS + OSTs to determine which one is right.
4789 Description: Inability to activate failout mode
4790 Details : lconf script incorrectly assumed that in pythong string's numeric
4791 value is used in comparisons.
4794 Frequency : always with multiple stripes per file
4796 Description: Inefficient object allocation for mutli-stripe files
4797 Details : When selecting which OSTs to stripe files over, for files with
4798 a stripe count that divides evenly into the number of OSTs,
4799 the MDS is always picking the same starting OST for each file.
4800 Return the OST selection heuristic to the original design.
4805 Description: mount failures may take full timeout to return an error
4806 Details : Under some heavy load conditions it is possible that a
4807 failed mount can wait for the full obd_timeout interval,
4808 possibly several minutes, before reporting an error.
4809 Instead return an error as soon as the status is known.
4811 Frequency : quota enabled and large files being deleted
4813 Description: releasing more than 4GB of quota at once hangs OST
4814 Details : If a user deletes more than 4GB of files on a single OST it
4815 will cause the OST to spin in an infinite loop. Release
4816 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4821 Description: statfs data retrieved from /proc may be stale or zero
4822 Details : When reading per-device statfs data from /proc, in the
4823 {kbytes,files}_{total,free,avail} files, it may appear
4824 as zero or be out of date.
4827 Frequency : systems with MD RAID1 external journal devices
4829 Description: lconf's call to blkid is confused by RAID1 journal devices
4830 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4831 of returning all block devices that match the journal UUID.
4834 Frequency : always, for aggregate stripe size over 4GB
4836 Description: assertion fails when trying to use 4GB stripe size
4837 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4838 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4841 Frequency : always on ppc64
4843 Description: the first write on an ext3 filesystem with mballoc got stuck
4844 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4845 perform endianness conversion.
4847 ------------------------------------------------------------------------------
4849 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4851 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4852 INTEROPERATE with older versions automatically. Please read the
4853 user documentation before upgrading any part of a live system.
4854 * WARNING: Lustre networking configuration changes are required with
4855 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4858 * Support for kernels:
4859 2.6.9-22.0.2.EL (RHEL 4)
4860 2.6.5-7.244 (SLES 9)
4861 2.6.12.6 vanilla (kernel.org)
4864 Severity : enhancement
4865 Bugzilla : 7981/8208
4866 Description: Introduced Lustre Networking (LNET)
4867 Details : LNET is new networking infrastructure for Lustre, it includes
4868 a reorganized network configuration mode (see the user
4869 documentation for full details) as well as support for routing
4870 between different network fabrics. Lustre Networking Devices
4871 (LNDs) for the supported network fabrics have also been
4872 created for this new infrastructure.
4874 Severity : enhancement
4875 Description: Introduced Access control lists
4876 Details : clients can set ACLs on files and directories in order to have
4877 more fine-grained permissions than the standard Unix UGO+RWX.
4878 The MDS must be started with the "-o acl" mount option.
4880 Severity : enhancement
4881 Description: Introduced filesystem quotas
4882 Details : Administrators may now establish per-user quotas on the
4885 Severity : enhancement
4887 Description: Configuration change for the XT3
4888 The PTLLND is now used to run Lustre over Portals on the XT3
4889 The configure option(s) --with-cray-portals are no longer used.
4890 Rather --with-portals=<path-to-portals-includes> is used to
4891 enable building on the XT3. In addition to enable XT3 specific
4892 features the option --enable-cray-xt3 must be used.
4897 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4898 Details : A race between allocating a new llcd and re-getting the llcd_lock
4899 allowed another thread to grab newly-allocated llcd.
4901 Severity : enhancement
4903 Description: 2.6 OST async journal commit and locking fix to improve performance
4904 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4905 2.6 kernels are now async as they already were in 2.4 kernels so
4906 that they can commit concurrently with the network bulk transfer.
4907 For block-allocated files the filter allocation semaphore is held
4908 to avoid filesystem fragmentation during allocation. BKL lock
4909 removed for 2.6 xattr operations where it is no longer needed.
4914 Description: lconf incorrectly determined whether two IP networks could talk
4915 Details : In some more complicated routing and multiple-network
4916 configurations, lconf will avoid trying to make a network
4917 connection to a disjoint part of the IP space. It was doing the
4918 math incorrectly for one set of cases.
4923 Description: Fix for potential infinite loop processing records in an llog.
4924 Details : If an llog record is corrupted/zeroed, it is possible to loop
4925 forever in llog_process(). Validate the llog record length
4926 and skip the remainder of the block on error.
4929 Frequency : occasional (liblustre only)
4931 Description: liblustre could not open files whose last component is a symlink
4932 Details : sysio_path_walk() would incorrectly pass the open intent to
4933 intermediate path components.
4936 Frequency : rare (liblustre only with non-standard tuning)
4937 Bugzilla : 7201 (7350)
4938 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4939 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4940 after it has been released, only do this for referenced locks
4942 Severity : enhancement
4944 Description: specifying an (invalid) directory default stripe_size of -1
4945 would reset the directory default striping
4946 Details : stripe_size -1 was used internally to signal directory stripe
4947 removal, now use "all default" to signal dir stripe removal
4948 as a directory striping of "all default" is not useful
4951 Frequency : common for large clusters running liblustre clients
4953 Description: doing an ls when liblustre clients are running is slow
4954 Details : sending a glimpse AST to a liblustre client waits for every AST
4955 to time out, as liblustre clients will not respond. Since they
4956 cannot cache data we refresh the OST lock LVB from disk instead.
4958 Severity : enhancement
4960 Description: doing an ls at the same time as file IO can be slow
4961 Details : enqueue and other "small" requests can be blocked behind many
4962 large IO requests. Create a new OST IO portal for non-IO
4963 requests so they can be processed faster.
4966 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4968 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4969 created file with mode 000
4970 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4971 allow an ftruncate() as a non-root user to a file with mode 000.
4972 Lustre now respects this flag to disable mode checks when
4973 truncating a file owned by the user
4976 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4979 Description: Revoking locks from clients that went dead or catatonic might take
4981 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4982 cancellation of such locks instant on servers without waiting for
4983 any reply from clients. Clients drops these locks when cancel
4984 notification from server is received without replying.
4987 Frequency : liblustre-only, when liblustre client dies or becomes busy
4989 Description: Doing ls on Linux clients can take a long time with active
4991 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4992 granting such locks to it in the first place if possible. Locks
4993 are taken by proxy on the OST during the read or write and
4994 dropped immediately afterward. Add connect flags handling, do
4995 not grant locks to liblustre clients for glimpse ASTs.
4997 Severity : enhancement
4999 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
5000 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5001 sequential reads and behaves badly with applications doing random
5002 reads. Improve it by reading ahead at least read region, and
5003 avoiding excessive large RPC for small reads.
5005 Severity : enhancement
5007 Description: Creating more than 1000 files for a single job may cause a load
5008 imbalance on the OSTs if there are also a large number of OSTs.
5009 Details : qos_prep_create() uses an OST index reseed value that is an
5010 even multiple of the number of available OSTs so that if the
5011 reseed happens in the middle of the object allocation it will
5012 still utilize the OSTs as uniformly as possible.
5017 Description: OST or MDS may oops in ping_evictor_main()
5018 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5019 but doesn't restart at beginning of obd_exports_timed list
5022 Severity : enhancement
5024 Description: improve by-nid export eviction on the MDS and OST
5025 Details : allow multiple exports with the same NID to be evicted at one
5026 time without re-searching the exports list.
5029 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5031 Description: MDS may oops in groups_free()
5032 Details : in rare race conditions a newly allocated group_info struct is
5033 freed again, and this can be NULL. The 2.4 compatibility code
5034 for groups_free() checked for a NULL pointer, but 2.6 did not.
5037 Frequency : common for liblustre clients doing little filesystem IO
5038 Bugzilla : 9352, 7313
5039 Description: server may evict liblustre clients accessing contended locks
5040 Details : if a client is granted a lock or receives a completion AST
5041 with a blocking AST already set it would not reply to the AST
5042 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5045 Frequency : lfs setstripe, only systems with more than 160 OSTs
5047 Description: unable to set striping with a starting offset beyond OST 160
5048 Details : llapi_create_file() incorrectly limited the starting stripe
5049 index to the maximum single-file stripe count.
5052 Frequency : LDAP users only
5054 Description: lconf did not handle in-kernel recovery with LDAP properly
5055 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5057 Severity : enhancement
5059 Description: bind OST threads to NUMA nodes to improve performance
5060 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5061 node and do their allocations there to localize memory access
5063 Severity : enhancement
5065 Description: llmount can determine client NID directly from Myrinet (GM)
5066 Details : the client NID code from gmnalnid was moved directly into
5067 llmount, removing the need to use this or specifying the
5068 client NID explicitly when mounting GM clients with zeroconf
5071 Frequency : if client is started with down MDS
5073 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5074 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5075 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5076 ptlrpc_import_delayed_req() because we didn't check if the
5077 request was interrupted, and we also didn't break out of the
5078 event loop if there was no timeout
5083 Description: data loss during non-page-aligned writes to a single file from
5084 both multiple nodes and multiple threads on one node at same time
5085 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5086 inconsistency led to false short-reads, that were cached and later
5087 used by ->prepare_write() to fill in partially written page,
5088 leading to data loss.
5091 Frequency : always, if lconf --abort_recovery used
5093 Description: lconf --abort_recovery fails with 'Operation not supported'
5094 Details : lconf was attempting to abort recovery on the MDT device and not
5097 Severity : enhancement
5099 Description: remove cleanup logs
5100 Details : replace lconf-generated cleanup logs with lustre internal
5101 cleanup routines. Eliminates the need for client-cleanup and
5104 Severity : enhancement
5106 Description: add support for EAs (user and system) on lustre filesystems
5107 Details : it is now possible to store extended attributes in the Lustre
5108 client filesystem, and with the user_xattr mount option it
5109 is possible to allow users to store EAs on their files also
5111 Severity : enhancement
5113 Description: Add possibility (config option) to show minimal available OST free
5115 Details : When compiled with --enable-mindf configure option, statfs(2)
5116 (and so, df) will return least minimal free space available from
5117 all OSTs as amount of free space on FS, instead of summary of
5118 free spaces of all OSTs.
5120 Severity : enhancement
5122 Description: do not expand extent locks acquired on OST-side
5123 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5124 by server: they are not cached anyway.
5127 Frequency : when mmap is used/binaries executed from Lustre
5129 Description: Unmmap pages before throwing them away from read cache.
5130 Details : llap_shrink cache now attempts to unmap pages before discarding
5131 them (if unmapping failed - do not discard). SLES9 kernel has
5132 extra checks that trigger if this unmapping is not done first.
5137 Description: lconf didn't resolve symlinks before checking to see whether a
5138 given mountpoint was already in use
5141 Frequency : when migrating failover services
5142 Bugzilla : 6395, 9514
5143 Description: When migrating a subset of services from a node (e.g. failback
5144 from a failover service node) the remaining services would
5145 time out and evict clients.
5146 Details : lconf --force (implied by --failover) sets the global obd_timeout
5147 to 5 seconds in order to quickly disconnect, but this caused
5148 other RPCs to time out too quickly. Do not change the global
5149 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5151 Severity : enhancement
5152 Frequency : if MDS is started with down OST
5153 Bugzilla : 9439,5706
5154 Description: Allow startup/shutdown of an MDS without depending on the
5155 availability of the OSTs.
5156 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5157 Add appropriate locking and lov-osc refcounts for safe
5158 cleaning. Add osc abort_inflight calls in case the
5159 synchronize never started.
5162 Frequency : occasional (Cray XT3 only)
5164 Description: root not authorized to access files in CRAY_PORTALS environment
5165 Details : The client process capabilities were not honoured on the MDS in
5166 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5167 been accessed by an authorized user then root was able to access
5168 the file on the local client also. The root user capabilities
5169 are now allowed on the MDS, as this environment has secure UID.
5172 Frequency : occasional
5174 Description: ldiskfs "too long searching" message happens too often
5175 Details : A debugging message (otherwise harmless) prints too often on
5176 the OST console. This has been reduced to only happen when
5177 there are fragmentation problems on the filesystem.
5182 Description: Division by zero in statfs when all OSCs are inactive
5183 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5184 lov_statfs divides by value returned by lov_get_stripecnt().
5188 Bugzilla : 9489, 3273
5189 Description: First write from each client to each OST was only 4kB in size,
5190 to initialize client writeback cache, which caused sub-optimal
5191 RPCs and poor layout on disk for the first written file.
5192 Details : Clients now request an initial cache grant at (re)connect time
5193 and so that they can start streaming writes to the cache right
5194 away and always do full-sized RPCs if there is enough data.
5195 If the OST is rebooted the client also re-establishes its grant
5196 so that client cached writes will be honoured under the grant.
5201 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5202 Details : Now I/O RPCs go to different portal number and (presumably) fast
5203 lock requests (and glimses) and other RPCs get their own service
5204 threads pool that should be able to service those RPCs
5207 Severity : enhancement
5209 Description: Ability to exchange lustre version between client and servers and
5210 issue warnings at client side if client is too old. Also for
5211 liblustre clients there is ability to refuse connection of too old
5213 Details : New 'version' field is added to connect data structure that is
5214 filled with version info. That info is later checked by server and
5218 Frequency : rare, liblustre only.
5219 Bugzilla : 9296, 9581
5220 Description: Two simultaneous writes from liblustre at offset within same page
5221 might proceed at the same time overwriting eachother with stale
5223 Details : I/O lock withing llu_file_prwv was released too early, before data
5224 actually was hitting the wire. Extended lock-holding time until
5225 server acknowledges receiving data.
5228 Frequency : extremely rare. Never observed in practice.
5230 Description: avoid generating lustre_handle cookie of 0.
5231 Details : class_handle_hash() generates handle cookies by incrementing
5232 global counter, and can hit 0 occasionaly (this is unlikely, but
5233 not impossible, because initial value of cookie counter is
5234 selected randonly). Value of 0 is used as a sentinel meaning
5235 "unassigned handle" --- avoid it. Also coalesce two critical
5236 sections in this function into one.
5238 Severity : enhancement
5240 Description: allow liblustre clients to delegate truncate locking to OST
5241 Details : To avoid overhead of locking, liblustre client instructs OST to
5242 take extent lock in ost_punch() on client's behalf. New connection
5243 flag is added to handle backward compatibility.
5245 Severity : enhancement
5246 Bugzilla : 4928, 7341, 9758
5247 Description: allow number of OST service threads to be specified
5248 Details : a module parameter allows the number of OST service threads
5249 to be specified via "options ost ost_num_threads={N}" in the
5250 OSS's /etc/modules.conf or /etc/modprobe.conf.
5254 Bugzilla : 6146, 9635, 9895
5255 Description: servers crash with bad pointer in target_handle_connect()
5256 Details : In rare cases when a client is reconnecting it was possible that
5257 the connection request was the last reference for that export.
5258 We would temporarily drop the export reference and get a new
5259 one, but this may have been the last reference and the export
5260 was just destroyed. Get new reference before dropping old one.
5262 Severity : enhancement
5263 Frequency : if client is started with failover MDS
5265 Description: Allow multiple MDS hostnames in the mount command
5266 Details : Try to read the configuration from all specified MDS
5267 hostnames during a client mount in case the "primary"
5270 Severity : enhancement
5272 Description: Stop sending data to evicted clients as soon as possible.
5273 Details : Check if the client we are about to send or are sending data to
5274 was evicted already. (Check is done every second of waiting,
5275 for which l_wait_event interface was extended to allow checking
5276 of exit condition at specified intervals).
5279 Frequency : rare, normally only when NFS exporting is done from client
5281 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5283 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5286 Frequency : always, if srand() is called before liblustre initialization
5288 Description: Liblustre uses system PRNG disturbing its usage by user application
5289 Details : Introduce internal to lustre fast and high-quality PRNG for
5290 lustre usage and make liblustre and some other places in generic
5291 lustre code to use it.
5293 Severity : enhancement
5294 Bugzilla : 9477, 9557, 9870
5295 Description: Verify that the MDS configuration logs are updated when xml is
5296 Details : Check if the .xml configuration logs are newer than the config
5297 logs stored on the MDS and report an error if this is the case.
5298 Request --write-conf, or allow starting with --old_conf.
5300 Severity : enhancement
5302 Description: Handle symlinks in the path when checking if Lustre is mounted.
5303 Details : Resolve intermediate symlinks when checking if a client has
5304 mounted a filesystem to avoid duplicate client mounts.
5309 Description: lconf can hit an error exception but still return success.
5310 Details : The lconf command catches the Command error exception at the top
5311 level script context and will exit with the associated exit
5312 status, but doesn't ensure that this exit status is non-zero.
5317 Description: failure of ptlrpc thread startup can cause oops
5318 Details : Starting a ptlrpc service thread can fail if there are a large
5319 number of threads or the server memory is very fragmented.
5320 Handle this without oopsing.
5323 Frequency : always, only if liblustre and non-default acceptor port was used
5325 Description: liblustre cannot connect to servers with non-default acceptor port
5326 Details : tcpnal_set_default_params() was not called and was therefore
5327 ignoring the environment varaible TCPNAL_PORT, as well as other
5328 TCPNAL_ environment variables
5333 Description: two objects could be created on the same OST for a single file
5334 Details : If an OST is down, in some cases it was possible to create two
5335 objects on a single OST for a single file. No problems other
5336 than potential performance impact and spurious error messages.
5340 Bugzilla : 5681, 9562
5341 Description: Client may oops in ll_unhash_aliases
5342 Details : Client dcache may become inconsistent in race condition.
5343 In some cases "getcwd" can fail if the current directory is
5349 Description: Inode refcounting problems in NFS export code
5350 Details : link_raw functions used to call d_instantiate without obtaining
5351 extra inode reference first.
5355 Bugzilla : 9942, 9903
5356 Description: Referencing freed requests leading to crash, memleaks with NFS.
5357 Details : We used to require that call to ll_revalidate_it was always
5358 followed by ll_lookup_it. Also with revalidate_special() it is
5359 possible to call ll_revalidate_it() twice for the same dentry
5360 even if first occurence returned success. This fix changes semantic
5361 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5362 reference on a request referred from the intent.
5363 ll_intent_release() then releases such a request.
5366 Frequency : rare, normally benchmark loads only
5368 Description: unlinked inodes were kept in memory on the client
5369 Details : If a client is repeatedly creating and unlinking files it
5370 can accumulate a lot of stale inodes in the inode slab cache.
5371 If there is no other client load running this can cause the
5372 client node to run out of memory. Instead flush old inodes
5373 from client cache that have the same inode number as a new inode.
5376 Frequency : SLES9 2.6.5 kernel and long filenames only
5377 Bugzilla : 9969, 10379
5378 Description: utime reports stale NFS file handle
5379 Details : SLES9 uses out-of-dentry names in some cases, which confused
5380 the lustre dentry revalidation. Change it to always use the
5384 Frequency : rare, unless heavy write-truncate concurrency is continuous
5385 Bugzilla : 4180, 6984, 7171, 9963, 9331
5386 Description: OST becomes very slow and/or deadlocked during object unlink
5387 Details : filter_destroy() was holding onto the parent directory lock
5388 while truncating+unlinking objects. For very large objects this
5389 may block other threads for a long time and slow overall OST
5390 responsiveness. It may also be possible to get a lock ordering
5391 deadlock in this case, or run out of journal credits because of
5392 the combined truncate+unlink. Solution is to do object truncate
5393 first in one transaction without parent lock, and then do the
5394 final unlink in a new transaction with the parent lock. This
5395 reduces the lock hold time dramatically.
5398 Frequency : rare, 2.4 kernels only
5400 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5401 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5402 before unmounting their filesystem in order to avoid deadlock.
5403 The kernel_locked() function in 2.4 kernels only checks whether
5404 the kernel lock is held, not whether it is this process that is
5405 holding it as 2.6 kernels do.
5410 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5411 Details : The client ptlrpc code may be trying to reconnect to a down
5412 server before a previous connection attempt has timed out.
5413 Increase the reconnect interval to be longer than the connection
5414 timeout interval to avoid sending duplicate connections to
5418 Frequency : echo_client brw_test command
5420 Description: fix echo_client to work with OST preallocated code
5421 Details : OST preallocation code (5137) didn't take echo_client IO path
5422 into account: echo_client calls filter methods outside of any
5423 OST thread and, hence, there is no per-thread preallocated
5424 pages and buffers to use. Solution: hijack pga pages for IO. As
5425 a byproduct, this avoids unnecessary data copying.
5429 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5430 Description: Client can oops in mdc_commit_close() after open replay
5431 Details : It was possible for the MDS to return an open request with no
5432 transaction number in mds_finish_transno() if the client was
5433 evicted, but without actually returning an error. Clients
5434 would later try to replay that open and may trip an assertion
5435 Simplify the client close codepath, and always return an error
5436 from the MDS in case the open is not successful.
5439 Frequency : rare, 2.6 OSTs only
5441 Description: OST may deadlock under high load on fragmented files
5442 Details : If there was a heavy load and highly-fragmented OST filesystems
5443 it was possible to have all the OST threads deadlock waiting on
5444 allocation of biovecs, because the biovecs were not released
5445 until the entire RPC IO was completed. Instead, release biovecs
5446 as soon as they are complete to ensure forward IO progress.
5448 Severity : enhancement
5450 Description: Support for specifying external journal device at mount
5451 Details : If an OST or MDS device is formatted with an external journal
5452 device, this device major/minor is stored in the ext3 superblock
5453 and may not be valid for failover. Allow detecting and
5454 specifying the external journal at mount time.
5459 Description: Mounting an MDS with pending unlinked files may cause oops
5460 Details : target_finish_recovery() calls mds_postrecov() which returned
5461 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5462 considers this an error and immediately begins cleaning up the
5463 lov, just after starting the mds_lov process
5465 Severity : enhancement
5467 Description: Implement 'lfs df' to report actual free space on per-OST basis
5468 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5469 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5470 usage of objects; '-h' to report in human readable format.
5472 ------------------------------------------------------------------------------
5474 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5481 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5482 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5483 at startup during orphan unlinking. Instead, skip buddy update
5484 before setup, as it will be regenerated after recovery is complete.
5489 Description: If an OST is inactive, its locks might reference stale inodes.
5490 Details : lov_change_cbdata() must iterate over all namespaces, even if
5491 they are inactive to clear inode references from the lock.
5493 Severity : enhancement
5494 Frequency : occasional, if non-standard max_dirty_mb used
5496 Description: Client will block write RPCs if not enough grant
5497 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5498 then the client will block writes while waiting for another RPC
5499 to complete instead of consuming its dirty limit. With change
5500 we get improved performance when max_dirty_mb is small.
5502 Severity : enhancement
5503 Bugzilla : 3389, 6253
5504 Description: Add support for supplementary groups on the MDS.
5505 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5506 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5507 lookups for user supplementary groups into a cache.
5511 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5512 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5513 for liblustre O_EXCL is enforced by MDS.
5517 Description: readdir never returns NULL in liblustre.
5518 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5519 dentry in struct dirent.
5523 Frequency : liblustre only, depends on application IO pattern
5524 Description: liblustre clients evicted if not contacting servers
5525 Details : Don't put liblustre clients into the ping_evictor list, so
5526 they will not be evicted by the pinger ever.
5528 Severity : enhancement
5530 Description: Add ability to evict clients by NID from MDS.
5531 Details : By echoing "nid:$NID" string into
5532 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5533 $NID would be instantly evicted from this MDS and from all active
5534 OSTs connected to it.
5538 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5539 Details : lookup_it_finish() used to query file size from OSTs that was not
5544 Description: service threads change working directory to that of init
5545 Details : Starting lustre service threads may pin the working directory
5546 of the parent thread, making that filesystem busy. Threads
5547 now change to the working directory of init to avoid this.
5551 Frequency : during shutdown only
5552 Description: shutdown with a failed MDS or OST can cause unmount to hang
5553 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5558 Frequency : 2.6 only
5559 Description: chmod/chown may include an extra supplementary group
5560 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5561 supplementary group and if none is specified this is used.
5564 Bugzilla : 5479 (6816)
5566 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5567 Details : If lookup is in progress on a file that is unlinked we might try
5568 to revalidate the inode and fail in revalidate after lookup is
5569 complete and ll_file_open() enqueues the open again but
5570 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5573 Frequency : always, if lconf --abort_recovery used
5575 Description: lconf --abort_recovery fails with 'Operation not supported'
5576 Details : lconf was attempting to abort recovery on the MDT device and not
5579 ------------------------------------------------------------------------------
5581 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5586 Frequency : rare (only unsupported configurations with a node running as an
5588 Bugzilla : 6514, 5137
5589 Description: Mounting a Lustre file system on a node running as an OST could
5591 Details : OSTs now preallocates memory needed to write out data at
5592 startup, instead of when needed, to avoid having to
5593 allocate memory in possibly low memory situations.
5594 Specifically, if the file system is mounted on on OST,
5595 memory pressure could force it to try to write out data,
5596 which it needed to allocate memory to do. Due to the low
5597 memory, it would be unable to do so and the node would
5598 become unresponsive.
5600 Severity : enhancement
5602 Description: Addition of lconf --service command line option
5603 Details : lconf now accepts a '--service <arg>' option, which is
5604 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5606 Severity : enhancement
5608 Description: Failover mode is now the default for OSTs.
5609 Details : By default, OSTs will now run in failover mode. To return to
5610 the old behaviour, add '--failout' to the lmc line for OSTs.
5612 Severity : enhancement
5614 Description: Health checks are now provided for MDS and OSTs
5615 Details : Additional detailed health check information on MSD and OSTs
5616 is now provided through the procfs health_check value.
5619 Frequency : occasional, depends on IO load
5621 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5622 after numerous create/delete cycles
5623 Details : The ext3 inode allocation policy would not allocate new inodes
5624 very well on the OSTs because there are no new directories
5625 being created. Instead we look for groups with free space if
5626 the parent directories are nearly full.
5631 Description: Network or server problems during mount may cause partially
5632 mounted clients instead of returning an error.
5633 Details : The config llog parsing code may overwrite the error return
5634 code during mount error handling, returning success instead
5640 Description: MDS can fail to allocate large reply buffers
5641 Details : After long uptimes the MDS can fail to allocate large reply
5642 buffers (e.g. zconf client mount config records) due to memory
5643 fragmentation or consumption by the buffer cache. Preallocate
5644 some large reply buffers so that these replies can be sent even
5645 under memory pressure.
5649 Frequency : rare (liblustre)
5650 Description: fsx running with liblustre complained that using truncate() to
5651 extend the file doesn't work. This patch corrects that issue.
5652 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5653 ATTR_SIZE and lsm use in llu_setattr_raw.
5657 Frequency : rare, only 2.6 kernels
5658 Description: Unusual file access patterns on the MDS may result in inode
5659 data being lost in very rare circumstances.
5660 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5661 code in the 2.6 kernel caused the inode and/or EA data not to
5662 be read from disk, causing single-file corruption.
5666 Frequency : rare, only 2.6 filesystems using extents
5667 Description: Heavy concurrent write and delete load may cause data corruption.
5668 Details : It was possible under high-load situations to have an extent
5669 metadata block in the block device cache from a just-unlinked
5670 file overwrite a newly-allocated data block. We now unmap any
5671 metadata buffers that alias just-allocated data blocks.
5675 Frequency : filesystems with default stripe_count larger than 77
5676 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5677 Details : lconf specifies an inode size of 4096 bytes when the default
5678 stripe_count is larger than 77. This conflicts with the default
5679 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5680 this case to avoid pinning too much memory for large EAs.
5682 ------------------------------------------------------------------------------
5684 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5689 Frequency : rare (extremely heavy IO load with hundreds of clients)
5691 Description: Client is evicted, gets IO error writing to file
5692 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5693 caused clients to be evicted for AST timeouts. The fixes in
5694 bug 5192 mean we no longer need to have such short AST timeouts
5695 so ldlm_timeout has been increased.
5698 Frequency : occasional during --force or --failover shutdown under load
5699 Bugzilla : 5949, 4834
5700 Description: Server oops/LBUG if stopped with --force or --failover under load
5701 Details : a collection of import/export refcount and cleanup ordering
5702 issues fixed for safer force cleanup
5705 Frequency : only filesystems larger than 120 OSTs
5706 Bugzilla : 5990, 6223
5707 Description: lfs getstripe would oops on a very large filesystem
5708 Details : lov_getconfig used kfree on vmalloc'd memory
5711 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5712 Bugzilla : 6242, 6243
5713 Description: reading from files that had been truncated to a non-zero size
5714 but never opened returned no data
5715 Details : ll_file_read() reads zeros from no-object files to EOF
5720 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5721 Details : There's little harm in aborting MDS/OSS recovery and letting it
5722 try again, so I removed the LASSERT and return an error instead.
5724 Severity : enhancement
5726 Description: New debugging infrastructure for tracking down data corruption
5727 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5728 (b) cover more of the client-side code path, and (c) try to narrow
5729 down where problems occurred
5733 Bugzilla : 3819, 4364, 4397, 6313
5734 Description: Racing close and eviction MDS could cause assertion in mds_close
5735 Details : It was possible to get multiple mfd references during close and
5736 client eviction, leading to one thread referencing a freed mfd.
5738 Severity: : enhancement
5739 Bugzilla : 3262, 6359
5740 Description: Attempts to reconnect to servers are now more aggressive.
5741 Details : This builds on the enhanced upcall-less recovery that was added
5742 in 1.4.2. When trying to reconnect to servers, clients will
5743 now try each server in the failover group every 10 seconds. By
5744 default, clients would previously try one server every 25 seconds.
5749 Description: After recovery, certain operations trigger a failed
5750 assertion on a client.
5751 Details : Failing over an mds, using lconf -d --failover, while a
5752 client was doing a readdir() call would cause the client to
5753 LBUG after recovery completed and the readdir() was resent.
5755 Severity : enhancement
5757 Description: Default groups are now added by lconf
5758 Details : You can now run lconf --group <servicename> without having to
5759 manually add groups with lmc.
5762 Frequency : occasional
5764 Description: Nodes with an elan id of 0 trigger a failed assertion
5767 Frequency : always when accessing e.g. tty/console device nodes
5769 Description: tty and some other devices nodes cannot be used on lustre
5770 Details : file's private_data field is used by device data and lustre
5771 values in there got lost. New field was added to struct file to
5772 store fs-specific private data.
5775 Frequency : when exporting Lustre via NFS
5777 Description: NFSD failed occasionally when looking up a path component
5778 Details : NFSD is looking up ".." which was broken in ext3 directories
5779 that had grown large enough to become hashed.
5782 Frequency : Clusters with multiple interfaces not on the same subnet
5784 Description: Nodes will repeatedly try to reconnect to an interface which it
5785 cannot reach and report an error to the log.
5786 Details : Extra peer list entries will be created by lconf with some peers
5787 unreachable. lconf now validates the peer before adding it.
5790 Frequency : Only if a default stripe is set on the filesystem root.
5792 Description: Setting a default stripe on the filesystem root prevented the
5793 filesystem from being remounted.
5794 Details : The client was sending extra request flags in the root getattr
5795 request and did not allocate a reply buffer for the dir EA.
5798 Frequency : occasional, higher if lots of files are accessed by one client
5799 Bugzilla : 6159, 6097
5800 Description: Client trips assertion regarding lsm mismatch/magic
5801 Details : While revalidating inodes the VFS looks up inodes with ifind()
5802 and in rare cases can find an inode that is being freed.
5803 The ll_test_inode() code will free the lsm during ifind()
5804 when it finds an existing inode and then the VFS later attaches
5805 this free lsm to a new inode.
5809 Bugzilla : 6422, 7030
5810 Description: MDS deadlock between mkdir and client eviction
5811 Details : Creating a new file via mkdir or mknod (starting a transaction
5812 and getting the ns lock) can deadlock with client eviction
5813 (gets ns lock and trying to finish a synchronous transaction).
5816 Frequency : occasional
5817 Description: While starting a server, the fsfilt_ext3 module could not be
5819 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5820 kernels. Previously, lconf would still use the ext3 name
5821 when trying to load modules. Now, it will correctly use
5822 ext3 on 2.4 and ldiskfs on 2.6.
5824 Severity : enhancement
5825 Description: The default stripe count has been changed to 1
5826 Details : The interpretation of the default stripe count (0, to lfs
5827 or lmc) has been changed to mean striping across a single
5828 OST, rather than all available. For general usage we have
5829 found a stripe count of 1 or 2 works best.
5831 Severity : enhancement
5832 Description: Add support for compiling against Cray portals.
5833 Details : Conditional compiling for some areas that are different
5837 Frequency : occasional
5838 Bugzilla : 6409, 6834
5839 Description: Creating files with an explicit stripe count may lead to
5840 a failed assertion on the MDS
5841 Details : If some OSTs are full or unavailable, creating files may
5842 trigger a failed assertion on the MDS. Now, Lustre will
5843 try to use other servers or return an error to the
5847 Frequency : occasional
5849 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5850 caused lock timeout during readahead (since 1.4.2).
5851 Details : Processes doing readahead might match a lock that hasn't been
5852 granted yet if there are overlapping and conflicting lock
5853 requests. The readahead process waits on ungranted lock
5854 (original lock is CBPENDING), while OST waits for that process
5855 to cancel CBPENDING read lock and eventually evicts client.
5857 Severity : enhancement
5859 Description: Initial enabling of flock support for clients
5860 Details : Implements fcntl advisory locking and file status functions.
5861 This feature is provided as an optional mount flag (default
5862 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5863 locking are implemented yet, and those that are are not guaranteed
5864 to be completely correct in production environments.
5865 mount -t lustre -o [flock|noflock] ...
5868 Frequency : occasional
5870 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5871 assertion in the ext3 JBD (journaling) layer.
5872 Details : The b_committed_data struct is protected by the big kernel lock
5873 in 2.4 kernels, serializing journal_commit_transaction() and
5874 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5875 access to this struct. In 2.6 kernels there is finer grained
5876 locking to improve SMP performance of the JBD layer.
5880 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5885 Description: merge and cleanup kernel patches.
5888 -----------------------------------------------------------------------------
5890 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5892 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5893 versions of Lustre. Please update all servers and clients to
5894 version 1.4.2 or later at the same time. You must also run
5895 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5896 to update the configuration logs.
5898 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5899 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5900 - swab llog records of type '0' so we get proper header size/idx (5861)
5901 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5902 - fix rename of one directory over another leaking an inode (5953)
5903 - avoid SetPageDirty on 2.6 (5981)
5904 - don't re-add just-being-destroyed locks to the waiting list (5653)
5905 - when creating new directories, inherit the parent's custom
5906 striping settings if present parent (3048)
5907 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5908 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5909 - don't hold client locks on temporary worklist from l_lru (5666)
5910 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5911 - automatically evict dead clients (5921)
5912 - Update file size properly in create+truncate+fstat case (6196)
5913 - Do not unhash mountpoint dentries, do not allow removal of
5915 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5916 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5917 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5918 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5919 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5920 - Correctly handle reads of files with no objects (6243)
5921 - lctl recover will also mark a device active if deactivate used (5933)
5923 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5924 - allow --write-conf on an MDS with different nettype than client (5619)
5925 - don't write config llogs to MDS for mounts not from that MDS (5617)
5926 - lconf should create multiple TCP connections from a client (5201)
5927 - init scripts are now turned off by default; run chkconfig --on
5928 lustre and chkconfig --on lustrefs to use them
5929 - upcalls are no longer needed for clients to recover to failover
5931 - add --abort-recovery option to lconf to abort recovery on device
5933 - add support for an arbitrary number of OSTs (3026)
5934 - Quota support protocol changes.
5935 - forward compatibility changes to wire structs (6007)
5936 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5937 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5938 - improved "lustre status" script
5939 - initialize blocksize for non-regular files (6062)
5940 - added --disable-server and --disable-client configure options (5782)
5941 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5942 - Vanilla 2.4.29 support
5943 - increase maximum number of obd devices to 520 (6242)
5944 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5945 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5946 SLES 9 2.6.5 kernels
5947 - limit stripes per file to 160 (the maximum EA size) (6093)
5949 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5952 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5953 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5954 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5955 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5956 - small ext3 extents cleanups and fixes (5733)
5957 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5958 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5959 - avoid needless client->OST connect, fix handle mismatch (5317)
5960 - fix DLM error path that led to out-of-sync client, long delays (5779)
5961 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5962 - fix several locking issues related to i_size (5492,5624,5654,5672)
5963 - don't move pending lock onto export if it is already evicted (5683)
5964 - fix kernel oops when creating .foo in unlinked directory (5548)
5965 - fix deadlock in obdfilter statistics vs. object create (5811)
5966 - use time_{before,after} to avoid timer jiffies wrap (5882)
5967 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5968 - Do not leak request if server was not able to process it (5154)
5969 - If mds_open unable to find parent dir, make that negative lookup(5154)
5970 - don't create new directories with extent-mapping (5909, 5936)
5972 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5973 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5974 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5975 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5976 - support for RHEL4 kernel on i686 (5773)
5977 - provide error messages when incompatible logs are encountered (5898)
5979 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5980 * version 1.4.0.10 (1.4.1 release candidate 1)
5982 - don't keep a lock reference when lock is not granted (4238)
5983 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5984 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5985 - drop import inflight refcount on signal_completed_replay error (5255)
5986 - unlock page after async write error during send (3677)
5987 - handle missing objects in filter_preprw_read properly (5265)
5988 - no transno return for symlink open, don't save no-trasno open (3440)
5989 - don't try to complete elan receive that already failed (4012)
5990 - free RPC server reply state on error (5406)
5991 - clean up thread from ptlrpc_start_thread() on error (5160)
5992 - readahead could read extra page into cache that wasn't ejected (5388)
5993 - prevent races in class_attach/setup/cleanup/detach (5260)
5994 - don't dereference de->d_inode after l_dput of de (5458)
5995 - use "int" for stripe value returned from lock_to_stripe (5544)
5996 - mballoc allocation and error-checking fixes in 2.6 (5504)
5997 - block device patches to fix I/O request sizes in 2.6 (5482)
5998 - look up hostnames for IB nals (5602)
5999 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
6000 - don't start multiple acceptors for the same port (5277)
6001 - fix incorrect LASSERT in mds_getattr_name (5635)
6002 - export a proc file for general "ping" checking (5628)
6003 - fix "lfs check" to not block when the MDS is down (5628)
6005 - service request history (4965)
6006 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6007 - create an "evict_client" /proc entry on OSTs, like the MDS has
6008 - fix mount usage message, return errors per mount(8) (5168)
6009 - change grep [] to grep "[]" in tests so they work in more UMLs
6010 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6011 - remove ancient LOV_MAGIC_V0 EA support (5047)
6012 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6013 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6014 - allow readahead allocations to fail when low on memory (5383)
6015 - mmap locking landed again, after considerable improvement (2828)
6016 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6018 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6021 - send OST transaction number in read/write reply to free req (4966)
6022 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6023 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6024 - the watchdog thread now runs as interruptible (5246)
6025 - flock/lockf fixes (but it's still disabled, pending 5135)
6026 - don't use EXT3 constants in llite code (5094)
6027 - memory shortage at startup could cause assertion (5176)
6029 - reorganization of lov code
6030 - single portals codebase
6032 - add extents/mballoc support (5025)
6033 - direct I/O reads in the obdfilter (4048)
6034 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6036 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6039 - send OST transaction number in read/write reply to free req (4966)
6040 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6041 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6042 - the watchdog thread now runs as interruptible (5246)
6043 - handle missing objects in filter_preprw_read properly (5265)
6044 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6045 - drop import inflight refcount on signal_completed_replay error (5255)
6046 - unlock page after async write error during send (3677)
6047 - return original error code on reconstructed replies (3761)
6048 - no transno return for symlink open, don't save no-trasno open (3440)
6050 - add pid to ldlm debugging output (4922)
6051 - bump the watchdog timeouts -- we can't handle 30sec yet
6052 - extra debugging for orphan dentry/inode bug (5259)
6054 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6057 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6058 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6059 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6060 - fix NULL dereference in /proc/sys/portals/routes (4827)
6061 - allow failed mdc_close() operations to be interrupted (4561)
6062 - stop precreate on OST before MDS would time out on it (4778)
6063 - don't send partial-page writes before EOF from client (4410)
6064 - discard client grant for sub-page writes on large-page clients (4520)
6065 - don't free dentries not owned by NFS code, check generation (4806)
6066 - fix lsm leak if mds_create_objects() fails (4801)
6067 - limit debug_daemon file size, always print CERROR messages (4789)
6068 - use transno after validating reply (3892)
6069 - process timed out requests if import state changes (3754)
6070 - update mtime on OST during writes, return in glimpse (4829)
6071 - add mkfsoptions to LDAP (4679)
6072 - use ->max_readahead method instead of zapping global ra (5039)
6073 - don't interrupt __l_wait_event() during strace
6075 - add software watchdogs to catch hung threads quickly (4941)
6076 - make lustrefs init script start after nfs is mounted
6077 - fix CWARN/ERROR duplication (4930)
6078 - return async write errors to application if possible (2248)
6079 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6080 - print NAL number in %x format (4645)
6081 - update barely-supported suse-2.4.21-171 series (4842)
6082 - support for sles 9 %post scripts
6083 - support for building 2.6 kernel-source packages
6084 - support for sles km_* packages
6086 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6089 - ignore -ENOENT errors in osc_destroy (3639)
6090 - notify osc create thread that OSC is being cleaned up (4600)
6091 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6092 - reconstruct ost_handle() like mds_handle() (4657)
6093 - create a new thread to do import eviction to avoid deadlock (3969)
6094 - let lconf resolve symlinked-to devices (4629)
6095 - don't unlink "objects" from directory with default EA (4554)
6096 - hold socknal file ref over connect in case target is down (4394)
6097 - allow more than 32000 subdirectories in a single directory (3244)
6098 - fix blocks count for O_DIRECT writes (3751)
6099 - OST returns ENOSPC from object create when no space left (4539)
6100 - don't send truncate RPC if file size isn't changing (4410)
6101 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6102 - bind to privileged port in socknal and tcpnal (3689)
6104 - rate limit CERROR/CWARN console message to avoid overload (4519)
6105 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6106 - basic mmap support (3918)
6107 - kernel patch series update from b1_4 (4711)
6109 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6112 - avoid crash during MDS cleanup with OST shut down (2775)
6113 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6114 - don't use bad inodes on the MDS (3744)
6115 - dynamic object preallocation to improve recovery speed (4236)
6116 - don't hold spinlock over lock dumping or change debug flags (4401)
6117 - don't zero obd_dev when it is force cleaned (3651)
6118 - print grants to console if they go negative (4431)
6119 - "lctl deactivate" will stop automatic recovery attempts (3406)
6120 - look for existing locks in ldlm_handle_enqueue() (3764)
6121 - don't resolve lock handle twice in recovery avoiding race (4401)
6122 - revalidate should check working dir is a directory (4134)
6124 - don't always mark "slow" obdfilter messages as errors (4418)
6126 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6129 - don't close LustreDB during write_conf until it is done (3860)
6130 - fix typo in lconf for_each_profile (3821)
6131 - allow dumping logs from multiple threads at one time (3820)
6132 - don't allow multiple threads in OSC recovery (3812)
6133 - fix debug_size parameters (3864)
6134 - fix mds_postrecov to initialize import for llog ctxt (3121)
6135 - replace config semaphore with spinlock (3306)
6136 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6137 - don't allow enqueue to complete on a destroyed export (3822)
6138 - down write_lock before checking llog header bitmap (3825)
6139 - recover from lock replay timeout (3764)
6140 - up llog sem before sending rpc (3652)
6141 - reduce ns lock hold times when setting kms (3267)
6142 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6143 - fix NULL deref and obd_dev leak on setup error (3312)
6144 - replace some LBUG about llog ops with error handling (3841)
6145 - don't match INVALID dentries from d_lookup and spin (3784)
6146 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6147 - fix invalid assertion in ptlrpc_set_wait (3880)
6149 - add libwrap support for the TCP acceptor (3996)
6150 - add /proc/sys/portals/routes for non-root route listing (3994)
6151 - allow setting MDS UUID in .xml (2580)
6152 - print the stack of a process that LBUGs (4228)
6154 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6157 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6158 - make sure to unset replay flag from failed open requests (3440)
6159 - if default stripe count is 0, use OST count for inode size (3636)
6160 - update parent mtime/ctime on client for create/unlink (2611)
6161 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6162 - free recovery state on server during a forced cleanup (3571)
6163 - unregister_reply for resent reqs (3063)
6164 - loop back devices mounting and status check on 2.6 (3563)
6165 - fix resource-creation race that can provoke i_size == 0 (3513)
6166 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6167 - more debugging for page-accounting assertion (3746)
6168 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6169 - avoid deadlock after precreation failure (3758)
6170 - fix race and lock order deadlock in orphan handling (3450, 3750)
6171 - add validity checks when grabbing inodes from l_ast_data (3599)
6173 - add /proc/.../recovery_status to obdfilter (3428)
6174 - lightweight CDEBUG infrastructure, debug daemon (3668)
6175 - change default OSC RPC parameters to be better on small clusters
6176 - turn off OST read cache for files smaller than 32MB
6177 - install man pages and include them in rpms (3100)
6178 - add new init script for (un)mounting lustre filesystems (2593)
6179 - run chkconfig in %post for init scripts (3701)
6180 - drop scimac NAL (unmaintained)
6182 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6185 - clean kiobufs before and after use (3485)
6186 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6187 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6188 - change init script to fail more clearly if not run as root (1528)
6189 - allow clients to reconnect during replay (1742)
6190 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6191 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6192 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6193 - fix cond_resched (3554)
6194 - only evict unfinished clients after recovery (3515)
6195 - allow bulk resend, prevent data loss (3570)
6196 - dynamic ptlrpc request buffer allocation (2102)
6197 - don't allow unlinking open directory if it isn't empty (2904)
6198 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6199 - remove extraneous obd dereference causing LASSERT failure (3334)
6200 - don't use get_cycles() when creating temp. files on the mds (3156)
6201 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6202 - handle EEXIST for set-stripe, set proper directory name (3336)
6204 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6205 - fix ksocknal_fmb_callback() error messages (2918)
6207 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6210 - don't copy lvb into (possibly NULL) reply on error (2983)
6211 - don't deref dentry after dput, don't free lvb on error (2922)
6212 - use the kms to determine writeback rpc length (2947)
6213 - increment oti_logcookies when osc is inactive (2948)
6214 - update client's i_blocks count via lvb messages (2543)
6215 - handle intent open/close of special files properly (1557)
6216 - mount MDS with errors=remount-ro, like obdfilter (2009)
6217 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6218 - don't use cancelling-locks' kms values (2947)
6219 - use highest lock extent for kms, not last one (2925)
6220 - don't dereference ERR_PTR() dentry in error handling path (3107)
6221 - fix thread race in portals_debug_dumplog() (3122)
6222 - create lprocfs device entries at setup instead of at attach (1519)
6223 - common AST error handler, don't evict client on completion race (3145)
6224 - zero nameidata in detach_mnt in 2.6 (3118)
6225 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6226 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6227 - initialize RPC timeout timer earlier for 2.6 (3219)
6228 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6229 - print nal/nid for unknown nid (3258)
6230 - additional checks for oscc recovery before doing precreate (3284)
6231 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6232 - don't crash in mdc_close for bad permissions on open (3285)
6233 - zero i_rdev for non-device files (3147)
6234 - clear page->private before handing to FS, better assertion (3119)
6235 - tune the read pipeline (3236)
6236 - fix incorrect decref of invalidated dentry (2350)
6237 - provide read-ahead stats and refine rpc in flight stats (3328)
6238 - don't hold journal transaction open across create RPC (3313)
6239 - update atime on MDS at close time (3265)
6240 - close LDAP connection when recovering to avoid server load (3315)
6241 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6242 - don't leak open file on MDS after open resend (3325)
6243 - serialize filter_precreate and filter_destroy_precreated (3329)
6244 - loop device shouldn't call sync_dev() for nul device (3092)
6245 - clear page cache after eviction (2766)
6246 - resynchronize MDS->OST in background (2824)
6247 - refuse to mount the same filesystem twice on same mountpoint (3394)
6248 - allow llmount to create routes for mounting behind routers (3320)
6249 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6250 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6251 - fix rare problem with rename on htree directories (3417)
6253 - allow default OST striping configuration per directory (1414)
6254 - fix compilation for qswnal for 2.6 kernels (3125)
6255 - increase maximum number of MDS request buffers for large systems
6256 - change liblustreapi to be useful for external progs like lfsck (3098)
6257 - increase local configuration timeout for slow disks (3353)
6258 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6260 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6263 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6264 - don't overwrite extent policy data in reply if lock was blocked (2901)
6265 - drop filter export grants atomically with removal from device (2663)
6266 - del obd_self_export from work_list in class_disconnect_exports (2908)
6267 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6268 - swab reply message in mdc_close, other PPC fixes (2464)
6269 - fix destroying of named logs (2325)
6270 - overwrite old logs when running lconf --write_conf (2264)
6271 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6272 - fix race in target_handle_connect (2898)
6273 - mds_reint_create() should take same inode create lock (2926)
6274 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6275 - don't close files for self_export to avoid uninitialized obd (2936)
6276 - allow MDS with the same name as client node (2939)
6277 - hold dentry reference for closed log files for unlink (2325)
6278 - reserve space for all logs during transactions (2059)
6279 - don't evict page beyond end of stripe extent (2925)
6280 - don't oops on a deleted current working directory (2399)
6281 - handle hard links to targets without a parent properly (2517)
6282 - don't dereference NULL lock when racing during eviction (2867)
6283 - don't grow lock extents when lots of conflicting locks (2919)
6285 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6288 - account for cache space usage on clients to avoid data loss (974)
6289 - lfsck support in lustre kernel code (2349)
6290 - reduce journal credits needed for BRW writes (2370)
6291 - orphan handling to avoid losing space on client/server crashes
6292 - ptlrpcd can be blocked, stopping ALL progress (2477)
6293 - use lock value blocks to assist in proper KMS, faster stat (1021)
6294 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6295 - recovery for initial connections (2355)
6296 - fixes for mds_cleanup_orphans (1934)
6297 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6298 - block all file creations until orphan recovery completes (1901)
6299 - client remove rq_connection from request struct (2423)
6300 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6301 - recovery timer race (2670)
6302 - mdc_close recovey bug (2532)
6303 - ptlrpc cleanup bug (2710)
6304 - mds timeout on local locks (2588)
6305 - namespace lock held during RPCs (2431)
6306 - handle interrupted sync write properly (2503)
6307 - don't try to handle a message that hasn't been replied to (2699)
6308 - client assert failure during cleanup after abort recovery (2701)
6309 - leak mdc device after failed mount (2712)
6310 - ptlrpc_check_set allows timedout requests to complete (2714)
6311 - wait for inflight reqs when ptlrpcd finishes (2710)
6312 - make sure unregistered services are removed from the srv_list
6313 - reset bulk XID's when resending them (caught by 1138 test)
6314 - unregister_bulk after timeout
6315 - fix lconf error (2694)
6316 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6317 - readahead locks pages, leaves pending causing memory pressure (2673)
6318 - increase OST request buffers to 4096 on large machines (2729)
6319 - fix up permission of existing directories in simple_mkdir (2661)
6320 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6321 - don't assign transno to errored transactions (2742)
6322 - don't delete objects on OST if given a bogus objid from MDS (2751)
6323 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6324 - if rq_no_resend, then timeout request after recovery (2432)
6325 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6326 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6327 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6328 - MDS recovery shouldn't delete objects that it has given out (2730)
6329 - if enqueue arrives after completion, don't clobber LVB (2819)
6330 - don't unlock pages twice when trigger_group_io returns error (2814)
6331 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6332 - don't write pages to disk if there was an error (1450)
6333 - don't ping imports that have recovery disabled (2676)
6334 - take buffered bytes into account when balancing socknal conn (2817)
6335 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6336 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6337 - remove little-endian swabbing of llog records (1987)
6338 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6339 - retry reposting request buffers if they fail (1191)
6340 - grow extent at grant time to avoid granting a revoked lock (2809)
6341 - lock revoke doesn't evict page if covered by a second lock (2765)
6342 - disable VM readahead to avoid reading outside lock extents (2805)
6344 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6345 - updated kernel patches for hp-2.4.20 kernel (2681)
6347 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6350 - fix truncated write corruption (2366)
6351 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6353 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6354 - ensure len(uuid) < 37 in lmc (1171)
6355 - fix ia64 OOPS in llog_test (2255)
6356 - zero end of page at obdfilter for partial page writes (2648)
6357 - don't leave stale dentries around after renames (bug 2428)
6358 - fix timeouts when evicting a client with a single lock held (2642)
6359 - set deadline for the initial HELLO message to drain (2634)
6360 - print out dotted-quad IP addresses in the socknal (2302)
6362 - additional debugging for MDS client eviction problem (2443)
6363 - fix mkfsoptions support for osts (2603, 2604)
6365 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6368 - add series for the vanilla 2.6.0 kernel
6369 - add series for the vanilla 2.4.24 kernel
6370 - add series for a cray x86/64 UL kernel drop
6371 - fix xattr patches for the vanilla 2.4.19 series
6373 - generate true UUIDs in lmc (1171)
6374 - have portals stack dumping break in UML (2466)
6375 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6376 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6377 - wake callback waiting threads on client eviction (2460)
6378 - Add --ptldebug and --subsystem to lmc (1719)
6379 - update assertion to allow safe interrupt allocation
6380 - set rq_no_resend for cancel requests (2432)
6381 - recalculate ptlrpcd timeout after resend (2494)
6382 - call vfs_rmdir when removing pending directories (2368)
6383 - fix renaming a file to itself (2429)
6384 - lmc creates a default one-stripe lov (2454)
6385 - expand procfs space to handle large clusters (2326)
6386 - increase UML stack to avoid overflow
6387 - update lconf's list of debug and subsystem masks
6388 - fix lfs find --obd (2510)
6389 - /proc tunable for disabling filter read caching (2591)
6390 - stop rpm packages from altering slapd.conf (2301)
6391 - disable nagle in the socknal under 0conf (2578)
6392 - choose mds inode size based on stripe count (2572)
6393 - fix kernel-source rpm problems (2516)
6395 - add --disable-doc to avoid pdf generation (2421)
6396 - update documentation, tests, type-os, comments
6397 - avoid format warnings on ia64
6398 - remove the TOE NAL
6399 - tiny code cleanups by removing unused fields
6401 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6404 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6405 - remove the most blatant lies from BUILDING file (2371)
6406 - change default debug level to reasonable production setting
6407 - reduce client side cache size to reduce cache flush time
6408 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6409 - make TCP zerocopy and pinger support enabled by default (2476)
6410 - sync writes completed after process exits caused crashes (2319)
6411 - maintain correct mount count on the MDS (2356)
6412 - backout 1557, because 2316 wasn't really fixed
6413 - better file I/O statistics gathering in /proc
6414 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6415 - another kernel patch to fix zero-copy TCP function export
6416 - don't take duplicate lock when processing re-sent getattr (2420)
6417 - lctl uses obd_self_export instead of creating new conn (2353)
6418 - MDS/OST recovery case which requires object creation asserted (2425)
6419 - move lfs from /usr/sbin to /usr/bin in packages
6420 - fix race between mds_client_add and mds_client_free (2417)
6421 - use kmalloc instead of slabs in portals (2430)
6422 - don't create duplicate records when a failover MDS is present (2442)
6423 - remove unnecessary mount age check (2332)
6424 - don't remove directory inodes from locks prematurely (2451)
6425 - don't break if MDS service name is the same as hostname (2103)
6426 - fix races in client write RPC generation when cache full (2482)
6428 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6431 - remove now-unused request->rq_obd (278)
6432 - if an allocation fails, print out how much memory we've used (1933)
6433 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6434 - add the "configurable stack size" patch to most series files (1256)
6435 - ability to write large log records, for 100+ OST configs (2306)
6436 - fix NULL deref when filter_prep fails (2314)
6437 - fix operator precedence error in filter_sync
6438 - dynamic allocation of socknal TX descriptors (2315)
6439 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6440 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6441 - ensure that CWARN messages in Portals always get to the syslog
6442 - __init/__exit are not for prototype decls (ldlm_init/exit)
6443 - x86-64 compile warning fixes
6444 - fix gateway LMC keyword conflict (2318)
6445 - fix MDS lock inversions in getattr/reint paths (1844)
6446 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6447 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6448 - DLM race condition prevented some lock evictions (2328)
6449 - ENOMEM detection and retry on socknal sends (2230)
6450 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6451 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6452 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6453 - better error messages when a client is rejected during recovery (1505)
6454 - avoid cancelling locks which were never granted, after failure (2330)
6455 - fix i_sem/journal inversion in mds_client_add (2333)
6456 - fix truncate/getattr lock cycle deadlock (2334)
6457 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6458 - fix two rare exit paths which could leak an l_lock() ref (2321)
6459 - fencepost error in MDS/OST orphan recovery (2226)
6460 - make log record alignment 8 bytes (1988)
6461 - lstripe now fails when requested offset > ost_count (2237)
6462 - ensure that all kernel series have a complete list.h (1607)
6463 - fix crashes in special-file operations (2316)
6464 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6466 - allow configurable automake binary, for testing new versions
6467 - small update to the lfs documentation
6469 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6471 * fix negative export reference count in fsfilt_sync (2312)
6473 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6474 * release candidate 0.9.1
6476 - orphans are moved into the PENDING directory for possible recovery
6477 - replayed opens now open by fid for orphan/rename safety (1042)
6478 - last close of an orphan inode generates a transno (683)
6479 - chdir() and mount() now pin the directory entry (1020)
6480 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6481 - discard very old requests without processing them (1502)
6482 - remove some common, well-understood CERRORs (1505)
6483 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6484 - clear "grant" flags in OST replies until OST grant code lands (1644)
6485 - fix read performance by not clobbering i_blksize on client (1598)
6486 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6487 - make LRU size a /proc tunable, clears locks when reduced (707)
6488 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6489 - server locks take references on exports now (1558)
6490 - build fixes for 2.4.20-rh trees (1663)
6491 - return an error from lov_create if all OSCs are inactive (1751)
6492 - fix import levels when a reconnect happens without a timeout (1597)
6493 - exit early from mds_open if we get a lookup error (1749)
6494 - partial page read at EOF wouldn't wait for disk before sending (1642)
6495 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6496 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6497 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6498 - write/truncate lock inversion (1639)
6499 - Don't auto-load obdclass, portals modules during cleanup (1495)
6500 - fix timestamps from jumping to "now" (1763)
6501 - extra journal assertions (1648)
6502 - add an extra multiunlink test (1771)
6503 - fix read_record/write_record API (1776)
6504 - fix leak of offset_extent, possible incorrect i_size later (1772)
6505 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6506 - Debugging for the kqswnal_get_idle_tx problems (1820)
6507 - Allow recovery to be attempted multiple times (1536)
6508 - Write out MDS last_rcvd file after it is first created (1600)
6509 - Fix tx_descriptor leak in failed transmit situations (1827)
6510 - ext3 journaling fixes for assertion failure after IO error (1871)
6511 - class_export_put() on freed export after completion AST error (1896)
6512 - Fix revalidate looping in VFS (1322)
6513 - Don't access a freed export during MDS_REINT timeout (1521)
6514 - Add open-unlink recovery support on the MDS (1673,1764)
6515 - Return an error if no MDS data was read from last_rcvd (1946)
6516 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6517 - Don't setup a disk device that doesn't match exported UUID (317)
6518 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6519 - avoid committing NULL handle in force close
6520 - local.sh is now a one-stripe LOV configuration
6521 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6522 - fix ext3 htree duplicate directory entry corruption (1516)
6523 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6524 - update to new LOV EA format (2097)
6525 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6527 2003-06-15 Phil Schwan <phil@clusterfs.com>
6530 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6531 - per-import/export recovery handling (958, 931, 959)
6532 - multiple last-rcvd slots, for serving multiple FSes (949)
6533 - connections are again shared between multiple imp/exports (963, 964)
6534 - "umount -f" would hang if any requests needed to be sent (393, 978)
6535 - avoid pinning large req buffer by copying for queued messages (989)
6536 - add "uuid" to "lctl device" command to help upcalls (991)
6537 - "open" RPCs with transnos would confuse recovery counters (1037)
6538 - do proper endian conversion of all wire messages (288, 340, 891)
6539 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6540 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6541 - fix ptlrpc_connection leak in target_handle_connect (1174)
6542 - fix import refcounting bug in OST and MDS cleanup (1134)
6543 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6544 - fix very unlikely obd_types race condition (501)
6545 - remove osc_open hack for echo_client (1187)
6546 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6547 - a failure in read_inode2 leads to deadlock (1139)
6548 - cancel ack-locks as soon as transaction is committed (1072)
6549 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6550 - make sure to commitrw after any preprw to avoid deadlock (1162)
6551 - failing to execute a file in a lustre FS would lock inode (1203)
6552 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6553 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6554 - obd_uuid2tgt would walk off the end of the list (1255)
6555 - on IA64 the osc would give portals incorrect bulk size (1258)
6556 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6557 - fix lock inversion caused by new llite matching code (1282)
6558 - limit the number of dirty pages on a client to 10MB (1286)
6559 - timed out locks were not being corrected cancelled (1289)
6560 - fix O_DIRECT above 4GB on IA-32 (1292)
6561 * major user-visible changes
6562 - fail out/fail over policy now controlled by the upcall (993)
6564 - add OBD_PING to check server availability and failure (954)
6565 - lustre messages are now sent in sending host order (288, 340, 891)
6566 - add eadatalen to MDS getattr reply (340)
6567 - OST read replies may contain second buffer, with per-page status (593)
6569 2003-03-11 Phil Schwan <phil@clusterfs.com>
6572 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6573 - failed open()s could cause deadlock; fixed (867, 869)
6574 - stop cancelling OST locks when files are closed (481)
6575 - overlapping XID spaces caused network corruption (851, 853)
6576 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6577 - setattr_raw added, to do single-RPC, server-side setattrs
6578 - lmc/lconf syntax change for OST UUIDs
6579 - fix crashy race condition between ptlrpc_free_req and osc_close
6580 - don't use request in mdc_enqueue if we hit a timeout (889)
6581 - don't set the inode i_size for regular files from the MDS (896)
6582 - handle out of order completion AST (842)
6583 - don't LBUG if a lock request times out after receiving AST (913)
6584 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6585 - if a bad lock AST arrives, send an error instead of dropping entirely
6586 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6587 - fix leak in bulk IO when only partially completed (899, 900, 926)
6588 - fix O_DIRECT for ia64 (55)
6589 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6590 - C-z after timeout could hang a process forever; fixed (977)
6592 - client-side I/O cache (678, 924, 929, 941, 970)
6594 - READPAGE and SETATTRs which don't take server-side locks get
6597 2003-02-11 Phil Schwan <phil@clusterfs.com>
6600 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6601 - Fix fsfilt_extN_readpage() to read a full page of directory
6602 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6603 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6604 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6605 - Most intent operations were removed, in favour of a new RPC mode
6606 that does a single RPC to the server and bypasses most of the VFS
6607 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6608 - Aggressively cancel local locks on DLM servers
6609 - mds_reint_unlink sends EA to the client if it's the last nlink.
6610 client uses that EA to unlink OST objects.
6611 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6612 - recursive symlinks were fixed (439)
6613 - fixed NULL deref in DEBUG_REQ
6614 - filter_update_lastobjid no longer calls sync, which annoyed extN
6615 - fixed multi-client small-writes to a single file problem (445)
6616 - fixed mtime updates during file writes (607)
6617 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6618 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6619 - fixed Linux OST size reporting problem (444, 656)
6620 - OST now updates object mtime with writes or setattr (607, 619)
6621 - client verifies file size before zeroing page past EOF (445)
6622 - OST now writes last allocated objid to disk with allocation (108)
6623 - LOV on echo now works (409)
6625 - mds_reint_unlink sends a new buffer, with the EA included. this
6626 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6627 set if a regular file was being unlinked, and it was the last link
6628 - use PtlGet from the target for bulk writes (315)
6629 - OST now updates object mtime with writes or setattr (607, 619)
6630 - LDLM now has a grant-time callback to revalidate locked items, if
6632 - Many MDS operations were reorganized to combat race conditions
6634 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6635 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6637 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6640 - Fully reactivate OST imports after reconnection (512, others)
6641 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6642 - More graceful error handling for truncating on dead OST (515)
6643 - Don't error out unless we're actually accessing dead stripes (474)
6644 - Fix garbage sizes when stripes are missing (410)
6645 - LRU counters were broken, causing constant lock purge (433, 432)
6646 - garbage on read from stripes with failed OSTs (441)
6647 - mark OSCs as active before reconnecting during recovery (438)
6648 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6649 - lfind did not preserve OST order in output (443)
6650 - symlinks cause hung clients, incorrect data (439)
6651 - stop dereferencing request after dropping refcount (457)
6652 - don't LASSERT(spin_is_locked) on non-SMP (455)
6653 - fixes for many rename() bugs
6654 - fstat didn't correctly synchronize attributes (399)
6655 - server must handle lock cancellation during blocking AST prep (487)
6656 - bulk descriptors were free()d too soon (511)
6657 - fix paths in lconf, which would load incorrect modules (451, 507)
6658 - fix confusing lconf 'host not found' error message (386)
6659 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6660 - fix race condition in mdc_blocking_ast() for inode access (526)
6661 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6662 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6663 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6664 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6665 - fix lstripe in several ways (kernel side) (527)
6666 - fix request leak in ldlm_cli_enqueue (262)
6667 - incorrect OSC was marked inactive after OST failure
6668 - call mds_fs_cleanup before unmounting filesystem (524)
6669 - fix races between taking ns_lock and ldlm_lock_change_resource
6670 - fix races updating LOV export open file list
6671 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6672 - fix recovery NULL deref in ldlm_cli_cancel_unused
6673 - fix some DLM races by using new hash table for lock handles (419)
6674 - permit the client to specify desired inodes, at replay
6675 - duplicate requests when we queue them for replay reintegration
6676 - fix last_rcvd offset calculation
6677 - sync after each recovered transaction, so we always make progress
6678 - never, not always, ERESTART requests without transnos
6679 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6681 - skip replay if the MDS says that the client is already connected
6682 - don't check for a recovery-enabled export to match lctl's UUID
6683 - don't INC_USE_COUNT for phantom exports
6684 - don't crash when cleaning up phantom exports (567)
6685 - don't double-finish or set replay data for errored mdc_open requests
6686 - abort requests when they time out, so we don't get old replies
6687 - send/receive replies for AST messages again
6688 - if the client says that it doesn't have the lock, cancel it on the
6690 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6691 mark it as destroyed, it will all work out when decref is called
6692 - fix module use counts (22, 581)
6694 - ASTs now expect a reply (server cancels lock on error reply)
6696 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6699 - fix many simultaneous client startup (392)
6700 - fix dentry->d_it clobbering
6701 - credentials weren't being shipped for readdir/getattr operations
6702 - remove invalid assertions triggered during some concurrent MD
6704 - proper Lustre versions added (336, 389)
6705 - fix memory leak for create error case (398)
6706 - fix LOV locking bug that would get cli/srv out of sync
6707 - fix echo client over LOV (409)
6708 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6709 - fix double-O_EXCL intent crash (424)
6710 - avoid sending multiple lock CANCELs (352)
6712 - MDS can do multi-client recovery (modulo bugs in new code)
6714 - many updates, edits, cleanups
6716 2002-11-18 Phil Schwan <phil@clusterfs.com>
6719 - fix null d_it dereference (346)
6720 - fix full OST/dbench hang (333)
6721 - fix permission problem with file removal (286)
6722 - fix removal of OSCs from LOV when they fail
6723 - fix NULL deref during bulk timeout (214)
6724 - fix problems related to multiple filesystems on one MDS (241)
6725 - fixed serious subtle metadata locking bugs
6726 - free locks on clients when inodes are removed due to memory
6728 - fix inode pointer in lock data (285)
6729 - partial support for multiple MDS on a single host (241)
6730 - data locks weren't cancelled at clear_inode time (290, 311)
6731 - intent locks could lead to unbounded lock growth (205)
6732 - added a maximum lock count, an LRU list, and a flusher
6733 - fix multiple rename (365)
6734 - properly abstracted the echo client
6735 - OSC locked 1 byte too many; fixed
6736 - rewrote brw callback code:
6737 - fixed recovery bugs related to LOVs (306)
6738 - fixed too-many-pages-in-one-write crash (191)
6739 - fixed (again) crash in sync_io_timeout (214)
6740 - probably fixed callback-related race (385)
6742 - Add capability to MDS protocol
6743 - LDLM cancellations and callbacks on different portals
6745 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6748 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6749 - defer open object destruction to close time (601981, 138)
6750 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6751 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6752 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6753 - don't call obd_setattr in ll_file_release for destroyed objects
6754 * protocol change to lustre_msg: move |version| and add |flags|
6755 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6756 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6757 * added replay of create, unlink, link and rename operations during
6758 MDS failover; recovery should be much more robust now
6759 * remove failed OSCs from LOVs (only lov_create uses this so far)
6760 * the lustre-HOWTO was brought (more) up to date (582544)
6762 2002-10-23 Phil Schwan <phil@clusterfs.com>
6765 - in-use dentries weren't being reused properly (617851)
6766 - prevent multiple LDLM setup (599178)
6767 - fix LOV size calculations for truncate (617853)
6768 - fix client handling of MDS intent errors (POSIX)
6769 - fix permission bug in lovstripe.c test (624321)
6770 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6771 - truncate past end of file could corrupt data
6772 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6773 - a race in recovery could return ETIMEDOUT to apps (623947)
6774 - building outside the source directory was fixed
6775 * the lustre-HOWTO was brought (more) up to date (582544)
6776 * major progress was made on recovery functionality
6778 2002-10-10 Phil Schwan <phil@clusterfs.com>
6781 - recovery deadlock fix
6782 - rm -rf causes LBUG fix (617817)
6783 - file open by multiple tasks fix (618962)
6784 - directory permissions bugs (602707 and 620007)
6785 - journal_stop fixed with locking (611313)
6786 - O_APPEND failures resolved (618273, perhaps 614459)
6787 - lconf PATH fix (619770)
6788 - IA64 build fix (621450)
6789 - RPC buffer sizes scale with amount of memory
6791 2002-10-01 Phil Schwan <phil@clusterfs.com>
6794 - locks would be cancelled without throwing away data pages,
6795 resulting in inconsistent data (605627)
6796 - inode attributes were not always being refreshed (605627, 612449)
6797 - lconf now continues to cleanup after lctl reports an error
6798 - MDS now enforces user permissions (602707)
6799 - lprocfs cleanup fixed, but not yet enabled (614157)
6800 - fixed infinite server hang, should a client not respond to an AST
6801 - avoid going into recovery if user calls readlink() with a buffer
6802 that's too small (613941)
6803 - AST RPCs no longer require replies (614867) -- this may be changed
6804 - don't crash server if client sends an IOV that's too big (611336)
6805 - fixed lock conversion deadlock (611892)
6806 - fixed the following of symlinks (614622)
6807 * recovery: the server can remove locks from a client that dies, other
6808 clients can make progress
6809 * more extN patch fixes
6810 * compile-time configurable ptlrpc buffer allocations
6812 - collaborative read cache document
6813 - Lustre Lite Performance CDR document-in-progress
6815 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6818 - fix typo in patch-2.4.18
6820 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6823 - clear ptlrpc request each time in handle_incoming_request()
6824 - unlink of files now destroys the object on the OST
6826 2002-09-19 Peter Braam <braam@clusterfs.com>
6828 * add hard link support
6829 * change obdfile creation method
6830 * kernel patch changed
6832 2002-09-19 Peter Braam <braam@clusterfs.com>
6835 - stack overflow bug in extN fixed
6837 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6839 * documentation updates
6840 - add man pages for config tools
6841 - update tests/README to describe testing with new config tools
6842 - finish metadata API descriptions
6843 * bug fixes and cleanups
6844 - statfs workaround for 16TB limit
6845 - LOV stripe allocation improved, can stripe on subset of OSTs
6846 - LOV file size/IO offset was wrong for files > 4GB in size
6847 - object EA data was being dropped, caused files to be unreadable
6848 - memory overflow with non-LOV OST caused memory corruption
6849 - fixed regression tests to work with new config tools, obdfilter
6850 - fixed bug when directory size became larger than 1 block
6851 - fixed bug (for single client case) when PWD was deleted
6852 - invalidate local directory pages when doing intent-based ops
6853 - avoid LDLM oops when lock callback contained bad data
6855 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6857 * documentation updates
6858 * bug fixes and cleanups
6859 - configuration tools
6862 - 64-bit compile warnings
6863 - 64-bit internal statfs data
6865 * test_brw on persistent OST devices
6867 * lprocfs (disabled)
6869 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6871 * documentation updates
6872 * bug fixes and cleanups
6873 * configuration tools
6875 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6878 * many small fixes to 0_5_4
6879 * io/network handling
6880 * thinkos in MDS operations
6882 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6885 * crucial basic fixes to 0.5.3
6886 * IOR, Iozone work over Elan
6889 2002-08-07 Phil Schwan <phil@clusterfs.com>
6890 * version 0_5_3, our first alpha
6891 * we use the new Portals iovs
6892 * documentation updates
6893 * bug fixes and cleanups
6894 * small changes in the DLM wire protocol
6896 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6897 * version 0_5_1 with some initial stability,
6898 * locking on MD and file I/O.
6899 * documentation updates
6900 * several bug fixes since 0.5.0
6901 * small changes in wire protocol
6903 2002-07-18 Phil Schwan <phil@clusterfs.com>
6905 * delivered as Lustre Light Alpha
6906 * fixed a crash after handling invalid MDS requests
6907 * fixed directory pages for architectures with non-4k pages sizes
6909 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6910 * release version v0_4_4
6911 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6913 * Fixes a number of interruption problems with OST operations.
6914 * Update documentation for portals header changes
6915 * Move all wire protocol structs/defines to lustre_idl.h
6916 * Fixes symlink length bug.
6917 * Add tcpdump to repository.
6919 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6920 * release version v0_4_3
6921 * Fixes statfs for inodes on extN.
6922 * Fixes bug in runtests which would delete /etc/hosts.
6923 * Use 64-bit object IDs wherever possible (not into VFS though)
6924 Remove ost_get_info, which is unused by lustre, and out of date.
6926 2002-07-03 Peter Braam <braam@clusterfs.com>
6927 * release version v0_4_2 Fixes a lookup error (type not passed)
6928 * move forward to head of Portals
6929 * move forward to latest Lustre kernel
6931 2002-06-25 Peter Braam <braam@clusterfs.com>
6932 * release version v0_4_1. Hopefully stable on single node use.