3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server primary kernels built and tested during release cycle:
6 3.10.0-957.27.2.el7 (RHEL7.6)
7 3.10.0-1062.18.1.el7 (RHEL7.7)
8 3.10.0-1127.19.1.el7 (RHEL7.8)
9 3.10.0-1160.25.1.el7 (RHEL7.9)
10 * Server known to build on the following kernels (others may also work):
11 2.6.32-696.18.7.el6 (RHEL6.9)
12 3.10.0-862.14.4.el7 (RHEL7.5)
13 4.4.120-92.70 (SLES12 SP2)
14 4.4.180-94.97 (SLES12 SP3)
15 4.4.0-142 (Ubuntu 16.04)
16 4.15.0-45 (Ubuntu 18.04)
17 vanilla linux 4.6.7 (ZFS only)
18 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
19 * Client primary unpatched kernels built and tested during release cycle:
20 3.10.0-957.27.2.el7 (RHEL7.6)
21 3.10.0-1062.18.1.el7 (RHEL7.7)
22 3.10.0-1127.19.1.el7 (RHEL7.8)
23 3.10.0-1160.25.1.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 4.12.14-122.66 (SLES12 SP5)
28 4.15.0-45 (Ubuntu 18.04)
29 * Client known to build on unpatched kernels (others may also work):
30 2.6.32-696.18.7.el6 (RHEL6.9)
31 2.6.32-754.6.3.el6 (RHEL6.10)
32 3.10.0-862.14.4.el7 (RHEL7.5)
33 4.14.0-49.13.1.el7a (RHEL7.5)
34 4.14.0-115.2.2.el7a (RHEL7.6)
35 4.18.0-147.3.1.el8 (RHEL8.1)
36 4.18.0-193.28.1.el8 (RHEL8.2)
37 4.18.0-240.22.1.el8 (RHEL8.3)
38 4.4.120-92.70 (SLES12 SP2)
39 4.4.0-142 (Ubuntu 16.04)
40 vanilla linux 4.15.0 (ZFS only)
41 * Recommended e2fsprogs version: 1.45.6.wc1 or newer
42 * Recommended ZFS / SPL version: 0.7.13
43 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
44 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
45 stack overflow. For more information, please refer to bugzilla 17630.
46 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
47 the re-exporting nodes to be mounted with "32bitapi" mount option
48 * Note for Ubuntu 14.04.5:
49 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
50 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
54 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
55 for currently supported client and server kernel versions.
56 * Server primary kernels built and tested during release cycle:
57 3.10.0-957.27.2.el7 (RHEL7.6)
58 3.10.0-1062.18.1.el7 (RHEL7.7)
59 3.10.0-1127.19.1.el7 (RHEL7.8)
60 3.10.0-1160.15.2.el7 (RHEL7.9)
61 * Server known to build on the following kernels (others may also work):
62 2.6.32-696.18.7.el6 (RHEL6.9)
63 3.10.0-862.14.4.el7 (RHEL7.5)
64 4.4.120-92.70 (SLES12 SP2)
65 4.4.180-94.97 (SLES12 SP3)
66 4.4.0-142 (Ubuntu 16.04)
67 4.15.0-45 (Ubuntu 18.04)
68 vanilla linux 4.6.7 (ZFS only)
69 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
70 * Client primary unpatched kernels built and tested during release cycle:
71 3.10.0-957.27.2.el7 (RHEL7.6)
72 3.10.0-1062.18.1.el7 (RHEL7.7)
73 3.10.0-1127.19.1.el7 (RHEL7.8)
74 3.10.0-1160.15.2.el7 (RHEL7.9)
75 4.18.0-80.11.2.el8 (RHEL8.0)
76 4.4.180-94.97 (SLES12 SP3)
77 4.12.14-95.54 (SLES12 SP4)
78 4.12.14-122.57 (SLES12 SP5)
79 4.15.0-45 (Ubuntu 18.04)
80 * Client known to build on unpatched kernels (others may also work):
81 2.6.32-696.18.7.el6 (RHEL6.9)
82 2.6.32-754.6.3.el6 (RHEL6.10)
83 3.10.0-862.14.4.el7 (RHEL7.5)
84 4.14.0-49.13.1.el7a (RHEL7.5)
85 4.14.0-115.2.2.el7a (RHEL7.6)
86 4.18.0-147.3.1.el8 (RHEL8.1)
87 4.18.0-193.28.1.el8 (RHEL8.2)
88 4.18.0-240.1.1.el8 (RHEL8.3)
89 4.4.120-92.70 (SLES12 SP2)
90 4.4.0-142 (Ubuntu 16.04)
91 vanilla linux 4.15.0 (ZFS only)
92 * Recommended e2fsprogs version: 1.45.6.wc1 or newer
93 * Recommended ZFS / SPL version: 0.7.13
94 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
95 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
96 stack overflow. For more information, please refer to bugzilla 17630.
97 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
98 the re-exporting nodes to be mounted with "32bitapi" mount option
99 * Note for Ubuntu 14.04.5:
100 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
101 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
105 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
106 for currently supported client and server kernel versions.
107 * Server primary kernels built and tested during release cycle:
108 3.10.0-957.27.2.el7 (RHEL7.6)
109 3.10.0-1062.18.1.el7 (RHEL7.7)
110 3.10.0-1127.8.2.el7 (RHEL7.8)
111 * Server known to build on the following kernels (others may also work):
112 2.6.32-696.18.7.el6 (RHEL6.9)
113 3.10.0-862.14.4.el7 (RHEL7.5)
114 4.4.120-92.70 (SLES12 SP2)
115 4.4.180-94.97 (SLES12 SP3)
116 4.4.0-142 (Ubuntu 16.04)
117 4.15.0-45 (Ubuntu 18.04)
118 vanilla linux 4.6.7 (ZFS only)
119 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
120 * Client primary unpatched kernels built and tested during release cycle:
121 3.10.0-957.27.2.el7 (RHEL7.6)
122 3.10.0-1062.18.1.el7 (RHEL7.7)
123 3.10.0-1127.8.2.el7 (RHEL7.8)
124 4.18.0-80.11.2.el8 (RHEL8.0)
125 4.4.180-94.97 (SLES12 SP3)
126 4.12.14-95.45 (SLES12 SP4)
127 4.12.14-122.20 (SLES12 SP5)
128 4.15.0-45 (Ubuntu 18.04)
129 * Client known to build on unpatched kernels (others may also work):
130 2.6.32-696.18.7.el6 (RHEL6.9)
131 2.6.32-754.6.3.el6 (RHEL6.10)
132 3.10.0-862.14.4.el7 (RHEL7.5)
133 4.14.0-49.13.1.el7a (RHEL7.5)
134 4.14.0-115.2.2.el7a (RHEL7.6)
135 4.18.0-147.8.1.el8 (RHEL8.1)
136 4.18.0-193.1.2.el8 (RHEL8.2)
137 4.4.120-92.70 (SLES12 SP2)
138 4.4.0-142 (Ubuntu 16.04)
139 vanilla linux 4.15.0 (ZFS only)
140 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
141 * Recommended ZFS / SPL version: 0.7.13
142 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
143 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
144 stack overflow. For more information, please refer to bugzilla 17630.
145 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
146 the re-exporting nodes to be mounted with "32bitapi" mount option
147 * Note for Ubuntu 14.04.5:
148 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
149 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
153 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
154 for currently supported client and server kernel versions.
155 * Server primary kernels built and tested during release cycle:
156 3.10.0-957.27.2.el7 (RHEL7.6)
157 3.10.0-1062.9.1.el7 (RHEL7.7)
158 * Server known to build on the following kernels (others may also work):
159 2.6.32-696.18.7.el6 (RHEL6.9)
160 3.10.0-862.14.4.el7 (RHEL7.5)
161 4.4.120-92.70 (SLES12 SP2)
162 4.4.180-94.97 (SLES12 SP3)
163 4.4.0-142 (Ubuntu 16.04)
164 4.15.0-45 (Ubuntu 18.04)
165 vanilla linux 4.6.7 (ZFS only)
166 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
167 * Client primary unpatched kernels built and tested during release cycle:
168 3.10.0-957.27.2.el7 (RHEL7.6)
169 3.10.0-1062.9.1.el7 (RHEL7.7)
170 4.18.0-80.11.2.el8 (RHEL8.0)
171 4.4.180-94.97 (SLES12 SP3)
172 4.12.14-95.29 (SLES12 SP4)
173 4.15.0-45 (Ubuntu 18.04)
174 * Client known to build on unpatched kernels (others may also work):
175 2.6.32-696.18.7.el6 (RHEL6.9)
176 2.6.32-754.6.3.el6 (RHEL6.10)
177 3.10.0-862.14.4.el7 (RHEL7.5)
178 4.14.0-49.13.1.el7a (RHEL7.5)
179 4.14.0-115.2.2.el7a (RHEL7.6)
180 4.18.0-147.3.1.el8 (RHEL8.1)
181 4.4.120-92.70 (SLES12 SP2)
182 4.4.0-142 (Ubuntu 16.04)
183 vanilla linux 4.15.0 (ZFS only)
184 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
185 * Recommended ZFS / SPL version: 0.7.13
186 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
187 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
188 stack overflow. For more information, please refer to bugzilla 17630.
189 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
190 the re-exporting nodes to be mounted with "32bitapi" mount option
191 * Note for Ubuntu 14.04.5:
192 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
193 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
197 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
198 for currently supported client and server kernel versions.
199 * Server primary kernels built and tested during release cycle:
200 3.10.0-957.27.2.el7 (RHEL7.6)
201 3.10.0-1062.1.1.el7 (RHEL7.7)
202 * Server known to build on the following kernels (others may also work):
203 2.6.32-696.18.7.el6 (RHEL6.9)
204 3.10.0-862.14.4.el7 (RHEL7.5)
205 4.4.120-92.70 (SLES12 SP2)
206 4.4.180-94.97 (SLES12 SP3)
207 4.4.0-142 (Ubuntu 16.04)
208 4.15.0-45 (Ubuntu 18.04)
209 vanilla linux 4.6.7 (ZFS only)
210 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
211 * Client primary unpatched kernels built and tested during release cycle:
212 3.10.0-957.27.2.el7 (RHEL7.6)
213 3.10.0-1062.1.1.el7 (RHEL7.7)
214 4.18.0-80.7.1.el8 (RHEL8.0)
215 4.4.180-94.97 (SLES12 SP3)
216 4.12.14-95.29 (SLES12 SP4)
217 4.15.0-45 (Ubuntu 18.04)
218 * Client known to build on unpatched kernels (others may also work):
219 2.6.32-696.18.7.el6 (RHEL6.9)
220 2.6.32-754.6.3.el6 (RHEL6.10)
221 3.10.0-862.14.4.el7 (RHEL7.5)
222 4.14.0-49.13.1.el7a (RHEL7.5)
223 4.14.0-115.2.2.el7a (RHEL7.6)
224 4.4.120-92.70 (SLES12 SP2)
225 4.4.0-142 (Ubuntu 16.04)
226 vanilla linux 4.15.0 (ZFS only)
227 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
228 * Recommended ZFS / SPL version: 0.7.13
229 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
230 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
231 stack overflow. For more information, please refer to bugzilla 17630.
232 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
233 the re-exporting nodes to be mounted with "32bitapi" mount option
234 * Note for Ubuntu 14.04.5:
235 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
236 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
240 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
241 for currently supported client and server kernel versions.
242 * Server known to build on the following kernels (others may also work):
243 2.6.32-696.18.7.el6 (RHEL6.9)
244 3.10.0-862.14.4.el7 (RHEL7.5)
245 3.10.0-957.10.1.el7 (RHEL7.6)
246 4.4.120-92.70 (SLES12 SP2)
247 4.4.162-94.69 (SLES12 SP3)
248 4.4.0-142 (Ubuntu 16.04)
249 4.15.0-45 (Ubuntu 18.04)
250 vanilla linux 4.6.7 (ZFS only)
251 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
252 * Client known to build on unpatched kernels (others may also work):
253 2.6.32-696.18.7.el6 (RHEL6.9)
254 2.6.32-754.6.3.el6 (RHEL6.10)
255 3.10.0-862.14.4.el7 (RHEL7.5)
256 3.10.0-957.10.1.el7 (RHEL7.6)
257 4.14.0-49.13.1.el7a (RHEL7.5)
258 4.14.0-115.2.2.el7a (RHEL7.6)
259 4.4.120-92.70 (SLES12 SP2)
260 4.4.162-94.69 (SLES12 SP3)
261 4.12.14-95.13 (SLES12 SP4)
262 4.4.0-142 (Ubuntu 16.04)
263 4.15.0-45 (Ubuntu 18.04)
264 vanilla linux 4.15.0 (ZFS only)
265 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
266 * Recommended ZFS / SPL version: 0.7.13
267 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
268 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
269 stack overflow. For more information, please refer to bugzilla 17630.
270 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
271 the re-exporting nodes to be mounted with "32bitapi" mount option
272 * Note for Ubuntu 14.04.5:
273 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
274 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
279 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
280 for currently supported client and server kernel versions.
281 * Server known to build on the following kernels (others may also work):
282 2.6.32-696.18.7.el6 (RHEL6.9)
283 3.10.0-862.14.4.el7 (RHEL7.5)
284 3.10.0-957.10.1.el7 (RHEL7.6)
285 4.4.120-92.70 (SLES12 SP2)
286 4.4.162-94.69 (SLES12 SP3)
287 4.4.0-142 (Ubuntu 16.04)
288 4.15.0-45 (Ubuntu 18.04)
289 vanilla linux 4.6.7 (ZFS only)
290 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
291 * Client known to build on unpatched kernels (others may also work):
292 2.6.32-696.18.7.el6 (RHEL6.9)
293 2.6.32-754.6.3.el6 (RHEL6.10)
294 3.10.0-862.14.4.el7 (RHEL7.5)
295 3.10.0-957.10.1.el7 (RHEL7.6)
296 4.14.0-49.13.1.el7a (RHEL7.5)
297 4.14.0-115.2.2.el7a (RHEL7.6)
298 4.4.120-92.70 (SLES12 SP2)
299 4.4.162-94.69 (SLES12 SP3)
300 4.12.14-95.13 (SLES12 SP4)
301 4.4.0-142 (Ubuntu 16.04)
302 4.15.0-45 (Ubuntu 18.04)
303 vanilla linux 4.15.0 (ZFS only)
304 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
305 * Recommended ZFS / SPL version: 0.7.13
306 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
307 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
308 stack overflow. For more information, please refer to bugzilla 17630.
309 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
310 the re-exporting nodes to be mounted with "32bitapi" mount option
311 * Note for Ubuntu 14.04.5:
312 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
313 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
317 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
318 for currently supported client and server kernel versions.
319 * Server known to build on the following kernels (others may also work):
320 2.6.32-696.18.7.el6 (RHEL6.9)
321 3.10.0-862.14.4.el7 (RHEL7.5)
322 3.10.0-957.el7 (RHEL7.6)
323 4.4.120-92.70 (SLES12 SP2)
324 4.4.162-94.69 (SLES12 SP3)
325 4.4.0-131 (Ubuntu 16.04)
326 4.15.0-32 (Ubuntu 18.04)
327 vanilla linux 4.6.7 (ZFS only)
328 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
329 * Client known to build on unpatched kernels (others may also work):
330 2.6.32-696.18.7.el6 (RHEL6.9)
331 2.6.32-754.6.3.el6 (RHEL6.10)
332 3.10.0-862.14.4.el7 (RHEL7.5)
333 3.10.0-957.el7 (RHEL7.6)
334 4.14.0-49.13.1.el7a (RHEL7.5)
335 4.14.0-115.2.2.el7a (RHEL7.6)
336 4.4.120-92.70 (SLES12 SP2)
337 4.4.162-94.69 (SLES12 SP3)
338 4.4.0-131 (Ubuntu 16.04)
339 4.15.0-32 (Ubuntu 18.04)
340 vanilla linux 4.15.0 (ZFS only)
341 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
342 * Recommended ZFS / SPL version: 0.7.9
343 * Tested with ZFS / SPL version: 0.7.9
344 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
345 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
346 stack overflow. For more information, please refer to bugzilla 17630.
347 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
348 the re-exporting nodes to be mounted with "32bitapi" mount option
349 * Note for Ubuntu 14.04.5:
350 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
351 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
353 --------------------------------------------------------------------------------
356 2018-04-03 Intel Corporation
358 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
359 for currently supported client and server kernel versions.
360 * Server known to build on patched kernels:
361 2.6.32-431.29.2.el6 (RHEL6.5)
362 2.6.32-504.30.3.el6 (RHEL6.6)
363 2.6.32-573.26.1.el6 (RHEL6.7)
364 2.6.32-642.15.1.el6 (RHEL6.8)
365 2.6.32-696.18.7.el6 (RHEL6.9)
366 3.10.0-693.21.1.el7 (RHEL7.4)
367 3.0.101-0.47.71 (SLES11 SP3)
368 3.0.101-107 (SLES11 SP4)
369 3.12.74-60.64.40 (SLES12 SP1)
370 4.4.114-92.64 (SLES12 SP2)
371 4.4.114-94.11 (SLES12 SP3)
372 3.13.0-101 (Ubuntu 14.04, ZFS only)
373 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
374 4.4.0-116 (Ubuntu 16.04)
375 vanilla linux 4.6.7 (ZFS only)
376 * Client known to build on unpatched kernels:
377 2.6.32-431.29.2.el6 (RHEL6.5)
378 2.6.32-504.30.3.el6 (RHEL6.6)
379 2.6.32-573.26.1.el6 (RHEL6.7)
380 2.6.32-642.15.1.el6 (RHEL6.8)
381 2.6.32-696.18.7.el6 (RHEL6.9)
382 3.10.0-693.21.1.el7 (RHEL7.4)
383 3.0.101-0.47.71 (SLES11 SP3)
384 3.0.101-107 (SLES11 SP4)
385 3.12.74-60.64.40 (SLES12 SP1)
386 4.4.114-92.64 (SLES12 SP2)
387 4.4.114-94.11 (SLES12 SP3)
388 3.13.0-101 (Ubuntu 14.04)
389 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
390 4.4.0-116 (Ubuntu 16.04)
392 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
393 * Recommended ZFS / SPL version: 0.7.6
394 * Tested with ZFS / SPL version: 0.7.6
395 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
396 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
397 stack overflow. For more information, please refer to bugzilla 17630.
398 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
399 the re-exporting nodes to be mounted with "32bitapi" mount option
400 * Note for Ubuntu 14.04.5:
401 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
402 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
404 --------------------------------------------------------------------------------
406 2017-07-05 Intel Corporation
408 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
409 for currently supported client and server kernel versions.
410 * Server known to build on patched kernels:
411 2.6.32-431.29.2.el6 (RHEL6.5)
412 2.6.32-504.30.3.el6 (RHEL6.6)
413 2.6.32-573.26.1.el6 (RHEL6.7)
414 2.6.32-642.15.1.el6 (RHEL6.8)
415 2.6.32-696.3.1.el6 (RHEL6.9)
416 3.10.0-514.21.1.el7 (RHEL7.3)
417 3.0.101-0.47.71 (SLES11 SP3)
418 3.0.101-100 (SLES11 SP4)
419 3.12.74-60.64.40 (SLES12 SP1)
420 4.4.59-92.17 (SLES12 SP2)
421 vanilla linux 4.6.7 (ZFS only)
422 * Client known to build on unpatched kernels:
423 2.6.32-431.29.2.el6 (RHEL6.5)
424 2.6.32-504.30.3.el6 (RHEL6.6)
425 2.6.32-573.26.1.el6 (RHEL6.7)
426 2.6.32-642.15.1.el6 (RHEL6.8)
427 2.6.32-696.3.1.el6 (RHEL6.9)
428 3.10.0-514.21.1.el7 (RHEL7.3)
429 3.0.101-0.47.71 (SLES11 SP3)
430 3.0.101-100 (SLES11 SP4)
431 3.12.74-60.64.40 (SLES12 SP1)
432 4.4.59-92.17 (SLES12 SP2)
434 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
435 * Recommended ZFS / SPL version: 0.6.5.9
436 * Tested with ZFS / SPL version: 0.6.5.9
437 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
438 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
439 stack overflow. For more information, please refer to bugzilla 17630.
440 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
441 the re-exporting nodes to be mounted with "32bitapi" mount option
443 --------------------------------------------------------------------------------
445 12-07-2016 Intel Corporation
447 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
448 for currently supported client and server kernel versions.
449 * Server known to build on patched kernels:
450 2.6.32-431.29.2.el6 (RHEL6.5)
451 2.6.32-504.30.3.el6 (RHEL6.6)
452 2.6.32-573.26.1.el6 (RHEL6.7)
453 2.6.32-642.11.1.el6 (RHEL6.8)
454 3.10.0-514.el7 (RHEL7.3)
455 3.0.101-0.47.71 (SLES11 SP3)
456 3.0.101-80 (SLES11 SP4)
457 3.12.62-60.64.8 (SLES12 SP1)
458 vanilla linux 4.5.7 (ZFS only)
459 * Client known to build on unpatched kernels:
460 2.6.32-431.29.2.el6 (RHEL6.5)
461 2.6.32-504.30.3.el6 (RHEL6.6)
462 2.6.32-573.26.1.el6 (RHEL6.7)
463 2.6.32-642.11.1.el6 (RHEL6.8)
464 3.10.0-514.el7 (RHEL7.3)
465 3.0.101-0.47.71 (SLES11 SP3)
466 3.0.101-80 (SLES11 SP4)
467 3.12.62-60.64.8 (SLES12 SP1)
469 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
470 * Recommended ZFS / SPL version: 0.6.5.8
471 * Tested with ZFS / SPL version: 0.6.5.8
472 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
473 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
474 stack overflow. For more information, please refer to bugzilla 17630.
475 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
476 the re-exporting nodes to be mounted with "32bitapi" mount option
478 --------------------------------------------------------------------------------
480 02-29-2016 Intel Corporation
482 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
483 for currently supported client and server kernel versions.
484 * Server known to build on patched kernels:
485 2.6.32-431.29.2.el6 (RHEL6.5)
486 2.6.32-504.30.3.el6 (RHEL6.6)
487 2.6.32-573.12.1.el6 (RHEL6.7)
488 3.10.0-327.3.1.el7 (RHEL7.2)
489 3.0.101-0.47.71 (SLES11 SP3)
490 3.0.101-68 (SLES11 SP4)
491 vanilla linux 4.2.1 (ZFS only)
492 * Client known to build on unpatched kernels:
493 2.6.32-431.29.2.el6 (RHEL6.5)
494 2.6.32-504.30.3.el6 (RHEL6.6)
495 2.6.32-573.12.1.el6 (RHEL6.7)
496 3.10.0-327.3.1.el7 (RHEL7.2)
497 3.0.101-0.47.71 (SLES11 SP3)
498 3.0.101-68 (SLES11 SP4)
501 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
502 * Recommended ZFS / SPL version: 0.6.4.2
503 * Tested with ZFS / SPL version: 0.6.4.2
504 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
505 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
506 stack overflow. For more information, please refer to bugzilla 17630.
507 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
508 the re-exporting nodes to be mounted with "32bitapi" mount option
510 --------------------------------------------------------------------------------
512 03-10-2015 Intel Corporation
514 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
515 for currently supported client and server kernel versions.
516 * Server known to build on patched kernels:
517 2.6.32-431.29.2.el6 (RHEL6.5)
518 2.6.32-504.8.1.el6 (RHEL6.6)
519 3.0.101-0.46 (SLES11 SP3)
520 * Client known to build on unpatched kernels:
521 2.6.32-431.29.2.el6 (RHEL6.5)
522 2.6.32-504.8.1.el6 (RHEL6.6)
523 3.10.0-123.20.1.el7 (RHEL7)
524 3.0.101-0.46 (SLES11 SP3)
525 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
526 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
527 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
528 stack overflow. For more information, please refer to bugzilla 17630.
529 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
530 the re-exporting nodes to be mounted with "32bitapi" mount option
532 Severity : enhancement
534 Description: control OST-index in IDIF via ROCOMPAT flag.
535 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
536 last_rcvd file. For new formatted OST device, it will be auto set;
537 for the case of upgrading from old OST device, you can enable it
538 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
539 enabled, for new created OST-object, its IDIF-in-LMA will contain
540 the OST-index; for the existing OST-object, the OSD will convert
541 old format IDIF as new format IDIF with OST-index stored in the
542 LMA EA when accessing such OST-object or via OI scrub. Once such
543 flag is enabled, it cannot be reverted back, so the system cannot
544 be downgraded to the orignal incompatible version.
546 --------------------------------------------------------------------------------
548 07-30-2014 Intel Corporation
550 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
551 for currently supported client and server kernel versions.
552 * Server known to build on patched kernels:
553 2.6.32-431.20.3.el6 (RHEL6.5)
554 3.0.101-0.31 (SLES11 SP3)
555 * Client known to build on unpatched kernels:
556 2.6.32-431.20.3.el6 (RHEL6.5)
557 3.0.101-0.31 (SLES11 SP3)
560 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
561 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
562 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
563 stack overflow. For more information, please refer to bugzilla 17630.
564 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
565 the re-exporting nodes to be mounted with "32bitapi" mount option
567 --------------------------------------------------------------------------------
569 10-31-2013 Intel Corporation
571 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
572 * Server builds for kernels:
573 2.6.32-358.18.1.el6 (RHEL6.4)
574 * Client builds for unpatched kernels:
575 2.6.32-358.18.1.el6 (RHEL6.4)
576 3.0.92-0.8 (SLES11 SP3)
578 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
579 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
580 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
581 stack overflow. For more information, please refer to bugzilla 17630.
582 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
583 the re-exporting nodes to be mounted with "32bitapi" mount option
585 --------------------------------------------------------------------------------
587 05-30-2013 Intel Corporation
589 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
590 * Server builds for kernels:
591 2.6.32-358.6.2.el6 (RHEL6.4)
592 * Client builds for unpatched kernels:
593 2.6.32-358.6.2.el6 (RHEL6.4)
594 3.0.74-0.6.6 (SLES11 SP2)
596 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
597 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
598 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
599 stack overflow. For more information, please refer to bugzilla 17630.
600 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
601 the re-exporting nodes to be mounted with "32bitapi" mount option
603 --------------------------------------------------------------------------------
605 9-30-2012 Whamcloud, Inc.
607 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
608 * Server support for kernels:
609 2.6.32-279.5.1.el6 (RHEL6)
610 * Client support for unpatched kernels:
611 2.6.32-279.5.1.el6 (RHEL6)
613 2.6.18-238.19.1.el5 (RHEL5)
614 2.6.16.60-0.69.1 (SLES10)
615 * Recommended e2fsprogs version: 1.41.90.wc4
616 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
617 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
618 stack overflow. For more information, please refer to bugzilla 17630.
619 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
620 the re-exporting nodes to be mounted with "32bitapi" mount option
622 Severity : enhancement
624 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
626 Severity : enhancement
628 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
630 Severity : enhancement
632 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
634 Severity : enhancement
636 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
638 Severity : enhancement
640 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
642 --------------------------------------------------------------------------------
644 04-30-2012 Whamcloud, Inc.
646 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
647 * Server support for kernels:
648 2.6.32-220.4.2.el6 (RHEL6)
649 * Client support for unpatched kernels:
650 2.6.32-220.4.2.el6 (RHEL6)
652 2.6.18-238.19.1.el5 (RHEL5)
653 2.6.16.60-0.69.1 (SLES10)
654 * Recommended e2fsprogs version: 1.41.90.wc4
655 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
656 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
657 stack overflow. For more information, please refer to bugzilla 17630.
658 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
659 the re-exporting nodes to be mounted with "32bitapi" mount option
661 Severity : enhancement
663 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
665 Severity : enhancement
667 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
669 Severity : enhancement
671 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
673 Severity : enhancement
675 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
677 --------------------------------------------------------------------------------
679 09-30-2011 Whamcloud, Inc.
681 * Server support for kernels:
682 2.6.18-238.19.1.el5 (RHEL5)
683 2.6.32-131.6.1.el6 (RHEL6)
684 * Client support for unpatched kernels:
685 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
686 2.6.18-238.19.1.el5 (RHEL5)
687 2.6.32-131.6.1.el6 (RHEL6)
688 2.6.29.4-167.fc11 (FC11)
689 2.6.16.60-0.69.1 (SLES10)
691 * Recommended e2fsprogs version: 1.41.90.wc3
692 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
693 Lustre filesystem with 4K stack may cause a stack overflow. For more
694 information, please refer to bugzilla 17630.
695 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
696 mounted with "32bitapi" mount option
698 Severity : enhancement
699 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
701 Severity : enhancement
702 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
704 Severity : enhancement
705 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
708 Frequency : only with SLES10
710 Description: Use OFED "KMP" provided by Novell
711 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
712 Lustre is now built against this package. That means you need to
713 install the ofed-kmp package from Novell for the patchless client.
714 Note that the ofed-kmp that Novell ships may not exactly match the
715 kernel version but should still be compatible.
717 Severity : enhancement
719 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
720 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
722 --------------------------------------------------------------------------------
724 2010-07-15 Oracle, Inc.
726 * Server support for kernels:
727 2.6.18-164.11.1.el5 (RHEL 5)
728 2.6.18-164.11.1.0.1.el5 (OEL 5)
729 * Client support for unpatched kernels:
730 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
731 2.6.18-164.11.1.el5 (RHEL 5),
732 2.6.18-164.11.1.0.1.el5 (OEL 5)
733 2.6.16.60-0.42.8 (SLES 10),
735 2.6.29.4-167.fc11 (FC11)
736 * Recommended e2fsprogs version: 1.41.10-sun2
738 Severity : enhancement
740 Description: Distribute lfs_migrate script with lustre
741 Details : The lfs_migrate mentioned in the manual is now included in the
742 lustre tree in lustre/scripts and will distributed in lustre RPMs.
744 Severity : enhancement
746 Description: Failover nid list added to import proc file.
748 Severity : enhancement
750 Description: "som_preview" mount option required to use SOM
751 Details : The clients will need to use som_preview mount option
755 Description: Lustre does not do 1MB IOs to HW RAID
756 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
758 Severity : enhancement
760 Description: lfs quota output is not very convenient for awk/sed-parsing
761 Details : Some positions in lfs quota output table could be empty or
762 non-empty which made it hard to parse it with scripts, now
763 a dash is put instead of space where there is not supposed
768 Description: Update to OFED 1.5.1
769 Details : OFED 1.5.1 will be used in environments that don't use
770 vendor supplied OFED.
774 Description: Remove lnet. prefix from lctl params display
775 Details : Also renamed lnet.memused to lnet_memused to
780 Description: Use conf_param -d to remove permanant settings.
784 Description: Add list_param -R
785 Details : lctl list_param now supports recursive mode.
788 Bugzilla : 21829, 22312
789 Description: Fix llobdstat.
790 Details : 1. Now updating the write_bytes counter
791 2. Port script fixes from 1.8.
795 Description: decrease the usage of memory on clients.
796 Details : 1. On clients, recycle dentries and inodes unused.
797 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
802 Description: llobdstat fix and enhancement
803 Details : add a counter to set a limit to how many samples will be returned
804 fix a wildcard in the path to limit to obdfilter stats only
808 Description: lfs check only allowed for root
809 Details : Code cleanups and added a regresion test of non-root use of
814 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
815 Details : Fix hr->hr_index overload.
819 Description: Fix FIEMAP functionality.
823 Description: FC11 patchless client support added
827 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
828 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
832 Description: New acc-sm script and various test-framework enhancments.
836 Description: Previously landed patch was reverted.
840 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
844 Description: Add ldiskfs maxdirsize mount option.
848 Description: Fix mkfs.lustre support for 16TB LUNs.
852 Description: Print mmp_check_interval
853 Details : print mmp_check_interval and make it possible to abort
854 mount operation in case it takes too long.
858 Description: Add cascading_rw to lustre mpi tests.
862 Description: Store SOM data in EA.
863 Details : Also several SOM related cleanups.
867 Description: Add refcount to md_open_data
871 Description: Use refcount on lov_request_set to prevent memory corruption.
875 Description: Statahead: use parent dir's i_mutex for synchronization.
879 Description: loadgen improvements.
883 Description: Enable peer health detection in ksocklnd and ko2iblnd.
887 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
891 Description: Fix race on obd_devcie:md_stats
895 Description: Fix SLES 11 client builds.
896 Details : Resolved uninitalized pointer gcc warnings.
900 Description: Allow non-root access for "lfs check".
901 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
905 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
907 Severity : enhancement
909 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
910 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
914 Description: Sanity test_27p fails.
915 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
916 Sanity tests 27 were cleaned up and improved.
920 Description: Unify req format on client/servers
921 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
922 buffers and swabbers approach..
927 Description: drop join file stuff.
929 Severity : enhancement
931 Description: Add 2.6.27 and SLES11 (patchless client) support
935 Description: Fix in ptlrpc_expire_one_request() to print the signed time
938 Severity : enhancement
940 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
943 Severity : enhancement
945 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
947 Severity : enhancement
949 Description: correctly handle big reply message.
950 Details : send LNet event if reply is bigger than buffer and adjust this buffer
955 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
957 Severity : enhancement
959 Description: Add support for OEL5.
963 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
965 Severity : enhancement
967 Description: "weak-modules" support
968 Details : Implement "weak-modules" support which enables kernel modules
969 to be used with any kernel that implements the same kABI. In
970 order to achieve this modules are now installed in
971 /lib/modules/$(uname -r)/updates/kernel on all distributions.
975 Description: client could not reconnect to OST because of an active request.
976 Details : The client now retries to reconnect to the same server, if a
977 connect request failed with EBUSY or -EAGAIN. Test case 35b
978 conf-sanity.sh has been created.
982 Description: Don't allow make backward step on assignin osc next id.
983 Details : race between allocation next id and ll_sync thread can be cause
984 of set wrong osc next id and can be kill valid ost objects.
988 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
992 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
996 Description: Error handling in osc_statfs_interpret() has been improved.
997 Details : Check in osc_statfs_interpret() for EBADR.
1001 Description: Do not send statfs() requests to OSTs disabled by administrator.
1002 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
1004 Severity : enhancement
1006 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
1008 Severity : enhancement
1010 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
1012 Severity : enhancement
1014 Description: Enhancement for lfs(1) command to use numeric uid/gid.
1018 Description: Improved cleanup procedure for conf-sanity test_46a.
1022 Description: Do not start bulk transfer, if client tried to reconnect.
1023 Details : Do not start bulk transfer, if client tried to reconnect,
1024 but failed due to active rpcs. Check for exp_abort_active_req
1025 has been implemented.
1029 Description: Prevent inconsistences between linux and lustre mount structures.
1030 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
1031 Make the sleep interruptible.
1033 Severity : enhancement
1035 Description: provide server to client comms path
1036 Details : server to client communications path via new LDLM_SET_INFO rpc
1040 Description: truncate starts GFP_FS allocation under transaction causing deadlock
1041 Details : ldiskfs_truncate calls grab_cache_page which may start page
1042 allocation under an open transaction. This may lead to
1043 calling prune_icache with consequent lustre reentrance.
1047 Description: send correctly lsm on open replay
1048 Details : MDS is trust to LSM size on replay open, but client can set wrong size
1053 Description: enable client interface failover
1054 Details : When a child reconnects from another NID, properly update export
1055 nid hash position and ldlm reverse import.
1057 Severity : enhancement
1059 Description: Communicate OST degraded/readonly state via statfs to MDS
1060 Details : Flags in the statfs returned from OSTs indicate whether the
1061 OST is in a degraded RAID state, or if the filesystem has
1062 turned read-only after a filesystem error is detected.
1067 Description: don't panic if EPROTO was hit when reading symlink
1068 Details : correctly handling request reference in error cases.
1070 Severity : enhancement
1072 Description: Add LustreNetLink, a kernel-userspace communcation path.
1074 Severity : enhancement
1076 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1081 Description: access to llog context before init.
1082 Details : move handling CATALOGS file at osc layer and forbid access to llog
1083 context before init.
1086 Frequency : rare, only when client is writing to OST on local node
1088 Description: Avoid deadlock for local client writes
1089 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1090 memory freeing context. This allows OST threads to set the
1091 PF_MEMALLOC flag on task structures in order to allocate memory
1092 from reserved pools and complete IO.
1093 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1094 so that the OST threads generate memory pressure and allow
1095 inactive pages to be reclaimed.
1097 Severity : enhancement
1099 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1104 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1105 Details : move ldlm namespace creation in setup phase to avoid grab
1106 _lprocfs_lock with cli_sem held.
1110 Description: Temporarily disable grant shrink.
1111 Details : Disable the feature for debugging.
1114 Frequency : only when formatting test filesystems
1116 Description: Unable to run several mkfs.lustre on loop devices at the same time
1117 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1118 formatting. The solution is to proper handle the error.
1120 Severity : enhancement
1122 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1124 Severity : enhancement
1126 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1131 Description: resolve race between obd_disconnect and class_disconnect_exports
1132 Details : if obd_disconnect will be called to already disconnected export he
1133 forget release one reference and osc module can't unloaded.
1135 Severity : enhancement
1137 Description: Allow tuning service thread via /proc
1138 Details : For each service a new
1139 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1140 created that can be used to set min/max thread counts, and get the
1141 current number of running threads.
1146 Descriptoin: don't return error if have partially created objects for file.
1147 Details : lov_update_create_set uses set->set_success as index for created
1148 objects, so if some requests will be failed, they will have hole
1149 at end of array and qos_shrink_lsm for allocate correct lsm.
1151 Severity : enhancement
1153 Description: Update OFED support to 1.4.1
1158 Description: Reduce small size read RPC
1159 Details : Set read-ahead limit for every file and only do read-ahead when
1160 available read-ahead pages are bigger than 1M to avoid small size
1163 Severity : enhancement
1165 Description: move AT tunable parameters for more consistent usage
1166 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1168 Severity : enhancement
1170 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1171 Details : allow skip disconnected ost for send statfs request and hide error
1177 Description: fix racy locking of mballoc block bitmaps causing BUG
1178 Details : The locking of the mballoc buddy bitmap and the in-memory
1179 block bitmap was using two different spin locks in some
1180 cases. This made it possible to incorrectly access the
1181 mballoc bitmap while another process was modifying it,
1182 causing a sanity assertion to fail. While no on-disk corruption
1183 was reported, there was some risk of this happening.
1186 Frequency : rare, on llog test 6
1188 Descriptoin: don't allow connect to already connected import
1189 Details : allowing connect to already connected import is hide connecting problem.
1194 Descriptoin: allow kill process which wait statahead result
1195 Details : for some reasons 'ls' can stick in waiting result from statahead,
1196 in this case need way for kill this process.
1198 Severity : enhancement
1200 Description: Add state history info file, enhance import info file
1201 Details : Track import connection state changes in a new osc/mdc proc file;
1202 add overview-type data to the osc/mdc import proc file.
1204 Severity : enhancement
1206 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1208 Severity : enhancement
1210 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1215 Descriptoin: OST grant too much space to client even there are not enough space.
1216 Details : Client will shrink its grant cache to OST if there are no write
1217 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1218 this grant cache if there are already not enough avaible space
1219 (left_space < total_clients * 32M).
1222 Frequency : start MDS on uncleanly shutdowned MDS device
1224 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1225 Details : stay in waiting mds<>ost recovery finished produce random bugs
1226 due race between two ll_sync thread for one lov target. send
1227 ACTIVATE event only if connect realy finished and import have
1231 Frequency : rare, connect and disconnect target at same time
1233 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1234 Details : don't call obd_disconnect under lov_lock. this long time
1235 operation and can block ptlrpcd which answer to connect request.
1240 Descriptoin: don't lose wakeup for imp_recovery_waitq
1241 Details : recover_import_no_retry or invalidate_import and import_close can
1242 both sleep on imp_recovery_waitq, but we was send only one wakeup
1246 Frequency : always with long access acl
1248 Descriptoin: mds can't pack reply with long acl.
1249 Details : mds don't control size of acl but they limited by reint/getattr
1252 Severity : enhancement
1254 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1256 Severity : enhancement
1258 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1261 Frequency : start MDS on uncleanly shutdowned MDS device
1263 Descriptoin: aborting recovery hang on MDS
1264 Details : don't throttle destroy RPCs for the MDT.
1267 Frequency : on remount
1269 Description: external journal device not working after the remount
1270 Details : clear dev_rdonly flag for external journal devices in
1276 Description: shutdown vs evict race
1277 Details : client_disconnect_export vs connect request race.
1278 if client will evicted at this time - we start invalidate
1279 thread without referece to import and import can be freed
1283 Frequency : rare, need acl's on inode.
1285 Description: client can't handle ost additional correctly
1286 Details : if ost was added after client connected to mds client can have
1287 hit lnet_try_match_md ... to big messages to wide striped files.
1288 in this case need teach client to handle config events about add
1289 lov target and update client max ea size at that event.
1291 Severity : enhancement
1293 Description: Changelogs
1294 Details : Changelogs are a lightweight mechanism to track filesystem
1295 metadata and namespace changes. The changelog is recorded
1296 permanently on the MDTs, and is periodically "consumed" / purged
1297 when records are no longer needed.
1299 Severity : enhancement
1301 Description: compact fld format with extents
1302 Details : Store range of seq rather than every seq in FLD. Seq
1303 controller update FLD rather than clients. In Case of CMD, mdt0
1304 has FLD, all other metadata server act as non persistent proxy
1305 for FLD queries and cache fld entries in fld cache.
1310 Description: don't skip ost target if they assigned to file
1311 Details : Drop slow OSCs if we can, but not for requested start idx.
1312 This means "if OSC is slow and it is not the requested
1313 start OST, then it can be skipped, otherwise skip it only
1314 if it is inactive/recovering/out-of-space.
1318 Description: more cleanup in mds_lov
1319 Details : not send LOV EA under replay, we can't know about they size at this
1320 time. Don't allow client connect to mds before any ost connected,
1321 for avoid problems with LOV EA size and returning EIO to client.
1323 Severity : enhancement
1325 Description: Interoperability at server side (Disk interoperability)
1327 Severity : enhancement
1329 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1331 Severity : enhancement
1333 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1335 Severity : enhancement
1337 Description: New client IO stack (CLIO).
1339 Severity : enhancement
1341 Description: Commit on sharing. Eliminate inter-client dependencies between
1342 uncommitted transactions by doing transaction commits.
1343 Thereby clients may recovery independently.
1346 Frequency : Create a symlink file with a very long name
1348 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1349 Details : If there is no extra space in the request for early cancels,
1350 ldlm_req_handles_avail() returns 0 instead of a negative value.
1352 Severity : enhancement
1354 Description: Add /proc entry for import status
1355 Details : The mdc, osc, and mgc import directories now have
1356 an import directory that contains useful import data for debugging
1357 connection problems.
1359 Severity : enhancement
1361 Description: Re-disable certain /proc logging
1362 Details : Enable and disable client's offset_stats, extents_stats and
1363 extents_stats_per_process stats logging on the fly.
1366 Frequency : Only on FC kernels 2.6.22+
1368 Description: oops in statahead
1369 Details : Do not drop reference count for the dentry from VFS when lookup,
1370 VFS will do that by itself.
1372 Severity : enhancement
1374 Description: Generic /proc file permissions
1375 Details : Set /Proc file permissions in a more generic way to enable non-
1376 root users operate on some /proc files.
1380 Description: Hitting mdc_commit_close() ASSERTION
1381 Details : Properly handle request reference release in
1382 ll_release_openhandle().
1386 Description: quota recovery deadlock during mds failover
1387 Details : This patch includes att18982, att18236, att18237 in bz14840.
1389 1. fix osts hang when mds does failover with quotaon
1390 2. prevent watchdog storm when osts threads wait for the
1395 Frequency : only patchless client
1396 Description: add workaround for race between add/remove dentry from hash
1398 Severity : enhancement
1400 Description: Allow OST glimpses to return PW locks
1404 Description: LBUG when llog conf file is full
1405 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1409 Description: Prevent import from entering FULL state when server in recovery
1413 Description: service mount cannot take device name with ":"
1414 Details : Only when device name contains ":/" will mount treat it as
1420 Description: replace ptlrpcd with the statahead thread to interpret the async
1421 statahead RPC callback
1425 Frequency : on recovery
1426 Description: I/O failures after umount during fail back
1427 Details : if client reconnected to restarted server we need join to recovery
1428 instead of find server handler is changed and process self
1429 eviction with cancel all locks.
1431 Severity : enhancement
1433 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1435 Severity : enhancement
1437 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1439 Severity : enhancement
1441 Description: Upcall on Lustre log has been dumped
1442 Details : Allow for a user mode script to be called once a Lustre log has
1443 been dumped. It passes the filename of the dumped log to the
1444 script, the location of the script can be specified via
1445 /proc/sys/lnet/debug_log_upcall.
1450 Description: avoid idr_remove called for id which is not allocated.
1451 Details : Move assigment s_dev for clustered nfs to end of initialization,
1452 to avoid problem with error handling.
1457 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1458 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1459 into CONN_USED_HASH and this prodice warning when put connection
1460 again in unused hash.
1463 Severity : enhancement
1465 Description: File striping can now be set to use an arbitrary pool of OSTs.
1467 Severity : enhancement
1469 Description: Export bytes_read/bytes_write count on OSC/OST.
1473 Description: Early reply size mismatch, MGC loses connection
1474 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1475 the connect flags are properly negotiated.
1480 Description: quotas are not honored with O_DIRECT
1481 Details : all writes with the flag O_DIRECT will use grants which leads to
1482 this problem. Now using OBD_BRW_SYNC to guard this.
1486 Description: add quota statistics
1487 Details : 1. sort out quota proc entries and proc code.
1488 2. add quota statistics
1490 Severity : enhancement
1492 Description: enable quota support for HEAD.
1496 Description: Properly propagate oinfo flags from lov to osc for statfs
1497 Details : restore missing copy oi_flags to lov requests.
1499 Severity : enhancement
1501 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1502 : l_getgroups(8), lst(8), routerstat(8)
1503 Details : included man pages for llobdstat(8), llstat(8),
1504 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1506 Severity : enhancement
1508 Description: configure's --enable-quota should check the
1509 : kernel .config for CONFIG_QUOTA
1510 Details : configure is terminated if --enable-quota is passed but
1511 : no quota support is in kernel
1515 Description: Remove portals compatibility
1516 Details : Remove portals compatibility, not interoperable with releases
1521 Description: Resolve device initialization race
1522 Details : Prevent proc handler from accessing devices added to the
1523 obd_devs array but yet be intialized.
1525 Severity : enhancement
1527 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1529 Severity : enhancement
1531 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1536 Description: Using wrong pointer in osc_brw_prep_request
1537 Details : Access to array[-1] can produce panic if kernel compiled with
1538 CONFIG_PAGE_ALLOC enabled
1542 Description: Client runs out of low memory
1543 Details : Consider only lowmem when counting initial number of llap pages
1547 Description: *optional* service tags registration
1548 Details : if the "service tags" package is installed on a Lustre node
1549 When the filesystem is mounted, a local-node service tag will
1550 be created. See http://inventory.sun.com/ for more information
1551 about the Service Tags asset management system.
1555 Description: Kernel BUG tries to release flock
1556 Details : Lustre does not destroy flock lock before last reference goes
1557 away. So always drop flock locks when client is evicted and
1558 perform unlock regardless of successfulness of speaking to MDS.
1562 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1566 Description: sanity test 65a fails if stripecount of -1 is set
1567 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1572 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1573 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1576 Severity : enhancement
1578 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1582 Description: do not process already freed flock
1583 Details : flock can possibly be freed by another thread before it reaches
1584 to ldlm_flock_completion_ast.
1588 Description: LBUG during stress test
1589 Details : Need properly lock accesses the flock deadlock detection list.
1593 Description: oops in page fault handler
1594 Details : kernel page fault handler can return two special 'pages' in error
1595 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1599 Description: timeout with invalidate import.
1600 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1601 should be handled by ptlrpcd. This produce long age waiting and
1602 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1604 Severity : enhancement
1606 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1610 Description: enable MGS and MDT services start separately
1611 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1612 a co-located MGS without starting the MGS, which is a complement
1613 to 'nosvc' mount option.
1617 Frequency : after recovery
1618 Description: precreate to many object's after del orphan.
1619 Details : del orphan st in oscc last_id == next_id and this triger growing
1620 count of precreated objects. Set flag LOW to skip increase count
1621 of precreated objects.
1625 Frequency : rare, on clear nid stats
1626 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1627 Details : when clean nid stats sometimes try destroy live entry,
1628 and this produce panic in free.
1632 Description: Stack overflow during MDS log replay
1633 ease stack pressure by using a thread dealing llog_process.
1637 Description: wait until IO finished before start new when do lock cancel.
1638 Details : VM protocol want old IO finished before start new, in this case
1639 need wait until PG_writeback is cleared until check dirty flag and
1640 call writepages in lock cancel callback.
1642 Severity : enhancement
1644 Description: using special macro for print time and cleanup in includes.
1648 Description: mds_mfd_close() ASSERTION(rc == 0)
1649 Details : In mds_mfd_close(), we need protect inode's writecount change
1650 within its orphan write semaphore to prevent possible races.
1654 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1658 Frequency : rare, on shutdown ost
1659 Description: don't hit live lock with umount ost.
1660 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1661 use shrink_dcache_sb instead.
1665 Description: don't panic with use echo client
1666 Details : echo client pass NULL as client nid pointer and this produce null
1667 pointer dereference.
1671 Description: fix build on ppc32
1672 Details : compile code with -m64 flag produce wrong object file for ppc32.
1676 Description: add message levels for liblustreapi
1680 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1681 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1682 will result in an -EAGAIN and in the exit call path we call
1683 alloc_rr() which will with increasing aggressiveness attempt to
1684 aquire precreated objects on the minimum number of required OSCs.
1688 Description: Use old size assignment to avoid deadlock
1689 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1690 the scheduling while holding a spinlock. We do not need locking
1691 for size in ll_update_inode() because size is only updated from
1692 the MDS for directories or files without objects, so there is no
1693 other place to do the update, and concurrent access to such inodes
1694 are protected by the inode lock.
1698 Description: resolve "_IOWR redefined" build error on SLES10
1702 Description: dump the memory debugging after all modules are unloaded to
1703 suppress false negative in conf_sanity test 39
1705 Severity : enhancement
1707 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1712 Description: fix SLES kernel versioning
1713 Details : the kernel version for our SLES 10 kernel did not include a "-"
1714 before the "smp" at the end. while this was not a problem in
1715 general, it did mean that software trying to use the kernel
1716 version to try to detect a vendor specific kernel would fail.
1717 this was most evident by the OFED build scripts.
1721 Description: Don't update lov_desc members until making sure they are valid
1722 Details : When updating lov_desc members via proc fs, need fix their
1723 validities before doing the real update.
1727 Description: don't put request into delay list while invalidate in flight.
1728 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1729 import in flight. this produce timeout for invalidate and sometimes
1730 can cause stale data.
1734 Frequency : on ppc only
1735 Description: not convert ost objects for directory because it's not exist.
1736 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1740 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1742 Severity : enhancement
1744 Description: Update to support 2.6.22.14 vanilla kernel.
1748 Frequency : rare, on recovery
1749 Description: read procfs can produce deadlock in some situation
1750 Details : Holding lprocfs lock which send rpc can produce block for destroy
1751 obd objects and this also block reconnect with -EALREADY. This
1752 isn't fix all lprocfs bugs - but make it rare.
1754 Severity : enhancement
1756 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1759 Frequency : frequent on X2 node
1761 Description: mdc_set_open_replay_data LBUG
1762 Details : Set replay data for requests that are eligible for replay.
1766 Description: lustre_mgs: operation 101 on unconnected MGS
1767 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1768 MGC, and late on MGC cannot successfully connect to MGS and a lot
1769 of the error messages complaining that MGS is not connected.
1773 Frequency : on network error
1774 Description: panic with double free request if network error
1775 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1776 but it's true only for synchronus enqueue, for async enqueue
1777 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1780 Severity : enhancement
1782 Description: client-side metadata stat-ahead during readdir(directory readahead)
1783 Details : perform client-side metadata stat-ahead when the client detects
1784 readdir and sequential stat of dir entries therein
1787 Frequency : on start mds
1789 Description: Implement get_info(last_id) in obdfilter.
1792 Frequency : occasional
1794 Description: Correctly check stale fid, not start epoch if ost not support SOM
1795 Details : open with flag O_CREATE need set old fid in op_fid3 because
1796 op_fid2 was overwritten with new generated fid, but mds can answer
1797 with one of these two fids and both is not stale. Setattr
1798 incorrectly started an epoch and assume will be called
1799 done_writing, but without SOM done_writing ever being called.
1802 Frequency : rare, depends on device drivers and load
1804 Description: MDS or OSS nodes crash due to stack overflow
1805 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1806 In some cases, in conjunction with device drivers that use a lot
1807 of stack the MDS (or possibly OSS) service threads could overflow
1808 the stack. One change which was identified to consume additional
1809 stack has been reworked to avoid the extra stack usage.
1812 Frequency : occasional
1814 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1815 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1816 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1817 oscc_has_objects and see OSC has no objests and call
1818 oscc_internal_create to resend create request.
1820 Severity : enhancement
1822 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1824 Severity : enhancement
1826 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1829 Frequency : very rare
1831 Description: Fix replay if there is an un-replied request and open
1832 Details : In some cases, older replay request will revert the
1833 mcd->mcd_last_xid on MDS which is used to record the client's
1834 latest sent request.
1836 Severity : enhancement
1838 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1840 Severity : enhancement
1842 Description: Add rhel5 support to HEAD.
1844 Serverity : enhancement
1846 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1851 Description: Don't allow skipping OSTs if index has been specified.
1852 Details : Don't allow skipping OSTs if index has been specified, make
1853 locking in internal create lots better.
1857 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1858 Details : access to bitfield in structure is always rounded to long
1859 and this produce problem with not atomic change any bit.
1863 Description: Lustre make rpms failed.
1864 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1865 builds Lustre rpms from tarball.
1868 Frequency : rare on shutdown ost
1870 Description: If llog cancel was not send before clean_exports phase, this can
1871 produce deadlock in llog code.
1872 Details : If llog thread has last reference to obd and call class_import_put
1873 this produce deadlock because llog_cleanup_commit_master wait when
1874 last llog_commit_thread exited, but this never success because was
1875 called from llog_commit_thread.
1879 Description: allow userland application know is lost one of stripes.
1880 Details : fill lvb_blocks with error code on ost and return it to
1881 application if error flag found.
1885 Description: NULL lov_tgts causing MDS oops
1886 Details : more safe checks for NULL lov_tgts for avoid oops.
1888 Severity : enhancement
1890 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1894 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1895 Details : If an OST is down the MDS will hang indefinitely in
1896 obd_statfs_rqset() waiting for the statfs data. While for
1897 MDS QOS usage of statfs, it should not stuck in waiting.
1899 Severity : enhancement
1901 Description: remote_acl support
1902 Details : Support ACL-based permission check for remote user.
1903 Support setfacl/getfacl for remote user with the utils
1904 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1905 the system "{s,g}etfacl" utils.
1907 Severity : enhancement
1909 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1911 Severity : enhancement
1913 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1917 Description: make mgs_setparam() handle fsname containing dash
1918 Details : fsname containing a dash does not work with lctl conf_param
1920 Severity : enhancement
1922 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1924 Severity : enhancement
1926 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1930 Frequency : rare, at shutdown
1931 Description: access already free / zero obd_namespace.
1932 Details : if client_disconnect_export was called without force flag set,
1933 and exist connect request in flight, this can produce access to
1934 NULL pointer (or already free pointer) when connect_interpret
1935 store ocd flags in obd_namespace.
1939 Frequency : only at startup
1940 Description: not alloc memory with spinlock held.
1941 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1942 if any spinlock held.
1944 Severity : enhancement
1946 Description: make lustre randomly fail allocating memory
1947 Details : Make lustre randomly failed allocating memory for testing purpose.
1949 Severity : enhancement
1951 Description: lost problems with lov objid file
1952 Details : Fixes some scability and access to not inited memory problems
1953 in work with lov objdid file.
1958 Description: lfs find does not continue on file error
1959 Details : Continue other files processing when a file/dir is absent.
1963 Description: Inconsistent usage of lustre_pack_reply()
1964 Details : Standardize the usage of lustre_pack_reply() such that it
1965 always generate a CERROR on failure.
1968 Frequency : occasional
1970 Description: MDS hang or stay in waiting lock
1971 Details : If client receive lock with CBPENDING flag ldlm need send lock
1972 cancel as separate rpc, to avoid situation when cancel request
1973 can't processed due all i/o threads stay in wait lock.
1977 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1981 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1985 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1989 Description: Update to RHEL4 latest kernel.
1991 Severity : enhancement
1993 Description: Build SLES10 patchless client fails
1994 Details : The configure was broken by run ./configure with
1995 --with-linux-obj=.... argument for patchless client. When the
1996 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1997 can't search header adequately. Use absolute path such as
1998 -I($LINUX)/include instead.
2002 Description: interrupt oig_wait produce painc on resend.
2003 Details : brw_redo_request can be used for resend requests from ptlrpcd and
2004 private set, and this produce situation when rq_ptlrpcd_data not
2005 copyed to new allocated request and triggered LBUG on assert
2006 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
2007 ptlrpcd set if request is changed and can be safety changed to use
2012 Description: LASSERT_{REQ,REP}SWAB macros are buggy
2013 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
2014 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
2015 these macros and replace them with their difinitions instead.
2019 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
2020 Details : Update which_patch & target file for SLES10 latest kernel.
2023 Description: Remove client patches from SLES 10 kernel.
2024 Details : This causes SLES 10 clients to behave as patchless clients
2025 even on a Lustre-patched (server) kernel.
2027 Severity : enhancement
2029 Description: self-adjustable client's lru lists
2030 Details : use adaptive algorithm for managing client cached locks lru
2031 lists according to current server load, other client's work
2032 pattern, memory activities, etc. Both, server and client
2033 side namespaces provide number of proc tunables for controlling
2036 Severity : enhancement
2038 Description: light-weight GSS support
2039 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
2040 and reduce performance overhead.
2042 Severity : enhancement
2044 Description: Linux keyring support
2045 Details : Support using service of Linux keyring for Lustre GSS internal
2046 context refresh/cache mechanism.
2050 Description: Fix errors in lfs documentation
2051 Details : Fixes man pages
2055 Description: don't use GFP_* in generic Lustre code.
2056 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2060 Description: obdclass is limited by single OBD_ALLOC(idarray)
2061 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2065 Description: warnings with build patchless client with vanila 2.6.19 and up
2066 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2067 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2071 Frequency : Only for SLES
2072 Description: sanity_quota fail test_1
2073 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2074 file, which makes TSTID[2] inunique.
2078 Frequency : Only for SLES9
2079 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2080 Details : Update target/ChangeLog/which_patch .
2084 Frequency : Only for test_10 in sanity-quota.sh
2085 Description: a bug in quota test code
2086 Details : checking a test flag in a wrong place
2090 Frequency : Only for test_13 in sanity-quota.sh
2091 Description: a bug in quota test script
2092 Details : shouldn't check overfull allocated quota
2096 Description: jbd statistics
2097 Details : Port older jbd statistics patch for sles10
2099 Severity : enhancement
2101 Description: Add printing inode info into message about error in writepage.
2106 Description: reply_lock_interpret crash due to race with it and lock cancel.
2107 Details : Do not replay locks that are being cancelled. Do not reference
2108 locks by their address during replay, just by their handle.
2112 Frequency : When flocks are used.
2113 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2114 Details : Flock locks might destroy just granted lock if it could be merged
2115 with another existing flock, this is done in completion handler,
2116 so teach ldlm_cli_enquque_fini that this is a valid case for
2122 Description: Oops in read and write path when failing to allocate lock.
2123 Details : Check if lock allocation failed and return error back.
2127 Description: lstripe command fails for valid OST index
2128 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2129 instead of lov->desc.ld_active_tgt_count.
2133 Description: sanity.sh failed test 103
2134 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2139 Description: df doesn't work properly if diskfs blocksize != 4K
2140 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2144 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2145 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2147 Severity : enhancement
2149 Description: Adaptive timeouts
2150 Details : RPC timeouts adapt to changing server load and network
2151 conditions to reduce resend attempts and improve recovery time.
2155 Description: llapi_file_create() does not allow some changes
2156 Details : add llapi_file_open() that allows specifying the mode and
2157 open flags, and also returns an open file handle.
2161 Description: merge and cleanup kernel patches.
2162 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2166 Description: Add journal checksum support.(Kernel part)
2167 Details : The journal checksum feature adds two new flags i.e
2168 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2169 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2170 indicates that the commit block contains the checksum for
2171 the blocks described by the descriptor blocks. Now commit
2172 record can be sent to disk without waiting for descriptor
2173 blocks to be written to disk. This behavior is controlled
2174 using JBD2_FEATURE_ASYNC_COMMIT flag.
2178 Description: OSS needs mutliple precreate threads
2179 Details : Add ability to start more than one create thread per OSS.
2183 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2184 Details : execute lfs setstripe on client
2188 Description: mds_obd_create error creating tmp object
2189 Details : When the user sets quota on root, llog will be affected and can't
2190 create files and write files.
2193 Frequency : Always on ia64 patchless client, and possibly others.
2195 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2196 Details : This allows the patchless client to be loaded on architectures
2197 without this export.
2201 Description: RedHat Update kernel for RHEL5
2202 Details : Add the kernel config file for RHEL5.
2206 Description: RedHat Update kernel for RHEL5
2207 Details : Modify the kernel config file more closer RHEL5.
2211 Description: Build failure against Centos5 (RHEL5)
2212 Details : Define PAGE_SIZE when it isn't present.
2216 Description: test_7 on sanity-quota.sh can't be used on separate machines
2217 Details : add facet to handle it
2221 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2222 Details : with NFS, the anon dentry's parent was set to itself in
2223 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2224 corresponding dentry other than use rec->ur_name.
2226 Severity : enhancement
2228 Description: lfs setstripe enhancement
2229 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2233 Description: enable data checksumming by default
2234 Details : enable checksum by default, allow --disable-checksum
2235 configure option and "-o nochecksum" mount option. Checksums
2236 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2240 Description: lustre support for RHEL5
2241 Details : Add support for RHEL5.
2245 Description: Client eviction due to failover config
2246 Details : after a connection loss, the lustre client should attempt to
2247 reconnect to the last active server first before trying the
2248 other potential connections.
2252 Description: when mds and osts use different quota unit(32bit and 64bit),
2253 quota will be released repeatly.
2254 Details : void sending multiple quota reqs to mds, which will keep the
2255 status between the reqs.
2259 Description: osts not allocated evenly to files
2260 Details : change the condition to increase offset_idx
2263 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2265 Description: Data corruption for OSTs that are formatted larger than 2TB
2267 Details : When generating the bio request for lustre file writes the
2268 sector number would overflow a temporary variable before being
2269 used for the IO. The data reads correctly from Lustre (which
2270 will overflow in a similar manner) but other file data or
2271 filesystem metadata may be corrupted in some cases.
2275 Description: Tune the kernel for good SCSI performance.
2276 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2277 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2282 Description: rewrite ext2-derived code in obdclass/uuid.c
2283 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2284 from scratch preserving functionality.
2288 Description: "lfs find -obd UUID" prints directories
2289 Details : "lfs find -obd UUID" will return all directory names instead
2290 of just file names. It is incorrect because the directories
2291 do not reside on the OSTs.
2295 Description: use i_size_read and i_size_write in 2.6 port
2296 Details : replace inode->i_size access with i_size_read/write()
2299 Frequency : when removing large files
2301 Description: scheduling issue during removal of large Lustre files
2302 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2303 It causes scheduling issues when removing large files (17TB in the
2307 Frequency : only with liblustre clients on XT3
2309 Description: evictions taking too long
2310 Details : allow llrd to evict clients directly on OSTs
2313 Frequency : only on ppc
2315 Description: /proc/fs/lustre/devices broken on ppc
2316 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2317 We should make sure the type of 'index' is loff_t and then cast
2318 explicitly as needed below. Do not assign an explicitly cast
2322 Frequency : only for rhel5
2324 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2325 Details : Modify the target file & which_kernel.
2327 Severity : enhancement
2329 Description: omit set fsid for export NFS
2330 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2331 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2335 Frequency : Always, for kernels after 2.6.16
2336 Description: Fix warning idr_remove called for id=.. which is not allocated.
2337 Details : Last kernels save old s_dev before kill super and not allow
2338 to restore from callback - restore it before call kill_anon_super.
2342 Description: buffer overruns could theoretically occur
2343 Details : llapi_semantic_traverse() modifies the "path" argument by
2344 appending values to the end of the origin string, and a
2345 overrun may occur. Adding buffer overrun check in liblustreapi.
2349 Description: Fix error on 'ls .' at the top of the Lustre mount.
2350 Details : Don't revalidate dentry if it is a root dentry.
2354 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2355 Details : Modify vm-tunables-rhel4.patch.
2359 Description: Kernel config for 2.6.18-vanilla.
2360 Details : Modify targets/2.6-vanilla.target.in.
2361 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2362 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2363 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2364 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2368 Description: improve handling recoverable errors
2369 Details : if request processig with error which can be recoverable on server
2370 request should be resend, otherwise page released from cache and
2375 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2376 Details : Modify target file & which_patch.
2377 A flaw was found in the IA32 system call emulation provided
2378 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2379 value could be stored in the %RAX register, which could trigger an
2380 out-of-bounds system call table access. An untrusted local user
2381 could exploit this flaw to run code in the kernel
2382 (ie a root privilege escalation). (CVE-2007-4573).
2386 Description: O_DIRECT bypasses client statistics.
2387 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2388 read_ahead_stats were updated. copied the stats section from
2389 osc_send_oap_rpc() into async_internal().
2393 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2397 Description: change order of libsysio includes
2398 Details : '#include sysio.h' should always come before '#include xtio.h'
2402 Description: handle "serious error: objid * already exists" more gracefully
2403 Details : If LAST_ID value on disk is smaller than the objects existing in
2404 the O/0/d* directories, it indicates disk corruption and causes an
2405 LBUG(). If the object is 0-length, then we should use the existing
2406 object. This will help to avoid a full fsck in most cases.
2408 Severity : enhancement
2410 Description: adapt the lustre_config script to support the upgrade case
2411 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2412 server targets to 1.6 in parallel.
2416 Description: To avoid grant space > avaible space when the disk is almost
2417 full. Without this patch you might see the error "grant XXXX >
2418 available" or some LBUG about grant, when the disk is almost
2420 Details : In filter_check_grant, for non_grant cache write, we should
2421 check the left space by if (*left > ungranted + bytes), instead
2422 of (*left > ungranted), because only we are sure the left space
2423 is enough for another "bytes", then the ungrant space should be
2424 increase. In client, we should update cl_avail_grant only there
2425 is OBD_MD_FLGRANT in the reply.
2429 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2430 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2431 described in CVE-2007-4573. This problem could allow untrusted
2432 local users to gain root access.
2435 Frequency : when using O_DIRECT and quotas
2437 Description: Incorrect file ownership on O_DIRECT output files
2438 Details : block usage reported by 'lfs quota' does not take into account
2439 files that have been written with O_DIRECT.
2444 Description: touch file failed when fs is not full
2445 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2446 otherwise we can get ENOSP while fs is not full.
2450 Description: parallel lock callbacks
2451 Details : Instead of sending blocking and completion callbacks as separated
2452 requests, adding them to a set and sending in parallel.
2455 Frequency : only for Cray XT3
2456 Bugzilla : 12829/13455
2457 Description: Changing primary group doesn't change the group lustre assigns to
2459 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2460 overridden with the primary group provided by the group upcall,
2461 whereas the supplied fsgid can be trusted if it is in the list of
2462 supplementary groups returned by the group upcall.
2464 Severity : enhancement
2466 Description: Allow masking D_WARNING, D_ERROR messages from console
2467 Details : Console messages can now be disabled via lnet.printk.
2471 Description: User code with malformed file open parameter crashes client node
2472 Details : Before packing join_file req, all the related reference should be
2473 checked carefully in case some malformed flags cause fake
2474 join_file req on client.
2478 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2479 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2480 mode is equal to the granted mode. However, it should also check
2481 that LDLM_FL_CP_REQD is not set.
2485 Description: Heavy nfs access might result in deadlocks
2486 Details : After ELC code landed, it is now improper to enqueue any mds
2487 locks under och_sem, because enqueue might want to decide to
2488 cancel open locks for same inode we are holding och_sem for.
2492 Description: Client eviction while running blogbench
2493 Details : A lot of unlink operations with concurrent I/O can lead to a
2494 deadlock causing evictions. To address the problem, the number of
2495 oustanding OST_DESTROY requests is now throttled to
2496 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2497 callbacks are processed in priority.
2501 Description: enable ACLs on MDS by default
2502 Details : ACLs must be enabled on MDS by default.
2505 Frequency : PPC/PPC64 only
2507 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2508 Details : fix duplicated definitions between asm-ppc64/types.h and
2509 lustre_types.h on PPC.
2512 Frequency : PPC/PPC64 only
2514 Description: asm-ppc/segment.h does not exist
2515 Details : fix compile issue on PPC.
2519 Description: better handle error messages in extents code
2522 Frequency : RHEL4 only
2524 Description: mkfs is very slow on IA64/RHEL4
2525 Details : A performance regression has been discovered in the MPT Fusion
2526 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2527 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2528 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2530 Severity : enhancement
2532 Description: SNMP support enhancement
2533 Details : Adding total number of sampled request for an MDS node in snmp
2536 Severity : enhancement
2538 Description: Optimize ldlm waiting list processing for PR extent locks
2539 Details : When processing waiting list for read extent lock and meeting read
2540 lock that is same or wider to it that is not contended, skip
2541 processing rest of the list and immediatelly return current
2542 status of conflictness, since we are guaranteed there are no
2543 conflicting locks in the rest of the list.
2547 Description: Time out and refuse to reconnect
2548 Details : When the failover node is the primary node, it is possible
2549 to have two identical connections in imp_conn_list. We must
2550 compare not conn's pointers but NIDs, otherwise we can defeat
2551 connection throttling.
2555 Description: port llog fixes from b1_6 into HEAD
2556 Details : Port llog reference couting and some llog cleanups from b1_6
2557 (bug 10800) into HEAD, for protect from panic and access to already
2558 free llog structures.
2562 Description: Detect stride IO mode in read-ahead
2563 Details : When a client does stride read, read-ahead should detect that and
2564 read-ahead pages according to the detected stride pattern.
2568 Description: data checksumming impacts single node performance
2569 Details : add support for several checksum algorithm. Currently, only CRC32
2570 and Adler-32 are supported. The checksum type can be changed on
2571 the fly via /proc/fs/lustre/osc/*/checksum_type.
2575 Description: use adler32 for page checksums
2576 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2581 Description: build for x2 fails
2582 Details : fix compile issue on Cray systems.
2586 Description: Properly match for duplicate locks
2587 Details : Due to different lock order from skiplists code, we need to
2588 traverse entire list for now
2591 Frequency : only on PPC/SLES10
2593 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2594 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2595 be defined. Add a hack in mkfs_lustre.c to work around this compile
2600 Description: LASSERT on MDS when client holding flock lock dies
2601 Details : ldlm pool logic depends on number of granted locks equal to
2602 number of released locks which is not true for flock locks, so
2603 just exclude such locks from consideration.
2607 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2608 Details : Use fsfilt_sync() for both the whole filesystem sync and
2609 individual file sync to eliminate dangerous inode locking
2610 with I_LOCK that can lead to a deadlock.
2614 Description: performance in 1.6.3
2615 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2616 sure that 1MB requests are not fragmented by the block layer.
2618 Severity : enhancement
2620 Description: organize the server-side client stats on per-nid basis
2621 Details : Change the structure of stats under obdfilter and mds to
2629 The "uuid"s file would list the uuids of _active_ exports.
2630 And the clear entry is to clear all stats and stale nids.
2632 Severity : enhancement
2634 Description: eliminate client locks in face of contention
2635 Details : file contention detection and lockless i/o implementation
2636 for contended files.
2640 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2643 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2645 Description: MDS may lose file striping (and hence file data) in some cases
2646 Details : If there are additional extended attributes stored on the MDS,
2647 in particular ACLs, SELinux, or user attributes (if user_xattr
2648 is specified for the client mount options) then there is a risk
2649 of attribute loss. Additionally, the Lustre file striping
2650 needs to be larger than default (e.g. striped over all OSTs),
2651 and an additional attribute must be stored initially in the
2652 inode and then increase in size enough to be moved to the
2653 external attribute block (e.g. ACL growing in size) for file
2658 Description: skiplist implementation simplification
2659 Details : skiplists are used to group compatible locks on granted list
2660 that was implemented as tracking first and last lock of each
2661 lock group the patch changes that to using doubly linked lists
2665 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2666 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2667 because LASSERT on both the data supplied by a client, and the
2668 data on disk is dangerous and incorrect.
2670 Severity : enhancement
2672 Description: Slow truncate/writes to huge files at high offsets.
2673 Details : Directly associate cached pages to lock that protect those pages,
2674 this allows us to quickly find what pages to write and remove
2675 once lock callback is received.
2679 Description: more ldlm soft lockups
2680 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2681 starve other threads from the resource lock for a long time in
2682 case of long waiting queue, so change the debug level from
2683 D_OTHER to the less frequently used D_INFO.
2685 Severity : enhancement
2687 Description: add -gid, -group, -uid, -user options to lfs find
2691 Description: Hung threads in invalidate_inode_pages2_range
2692 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2693 once one is completed. As a result, some RPCs are stuck in the
2694 queue and are never sent.
2698 Description: filter threads hungs on waiting journal commit
2699 Details : Cleanup filter group llog code, then only filter group llog will
2700 be only created in the MDS/OST syncing process.
2704 Description: Procfs and llog threads access destoryed import sometimes.
2705 Details : Sync the import destoryed process with procfs and llog threads by
2706 the import refcount and semaphore.
2708 Severity : enhancement
2710 Description: openlock cache of b1_6 port to HEAD
2715 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2716 Details : If insertion of an extent fails, then discard the inode
2717 preallocation and free data blocks else it can lead to duplicate
2722 Description: don't always update ctime in ext3_xattr_set_handle()
2723 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2724 In some cases the ctime should not be updated, for example for
2725 2.0->1.8 compatibility it is necessary to delete an xattr and it
2726 should not update the ctime.
2730 Bugzilla : 15713/16362
2731 Description: Assertion in iopen_connect_dentry in 1.6.3
2732 Details : looking up an inode via iopen with the wrong generation number can
2733 populate the dcache with a disconneced dentry while the inode
2734 number is in the process of being reallocated. This causes an
2735 assertion failure in iopen since the inode's dentry list contains
2736 both a connected and disconnected dentry.
2740 Description: assertion failure in ldlm_handle2lock()
2741 Details : fix a race between class_handle_unhash() and class_handle2object()
2742 introduced in lustre 1.6.5 by bug 13622.
2747 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2748 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2753 Description: oops in sd_iostats_seq_show()
2754 Details : unloading/reloading the scsi low level driver triggers a kernel
2755 bug when trying to access the sd iostat file.
2760 Description: Kernel panics during QLogic driver reload
2761 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2762 patch, causing memory corruption.
2767 Description: journal_dev option does not work in b1_6
2768 Details : pass mount option during pre-mount.
2770 Severity : enhancement
2772 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2773 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2774 extent information of a file. It can be used to map logical blocks
2775 in a file to physical blocks in the block device.
2779 Description: LDLM soft lockups - improvement
2780 Details : It is be possible to send the lock handle along with each read
2781 or write request because the client is already doing a lock match
2782 itself so there isn't any reason the OST should have to re-do that
2786 Frequency : only on Cray X2
2788 Description: X2 build failures
2789 Details : fix build failures on Cray X2.
2793 Description: xid & resent requests
2794 Details : Initialize RPC XID from clock at startup (randomly if clock is
2797 Severity : enhancement
2799 Description: Add lustre_start utility to start or stop multiple Lustre servers
2804 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2805 Details : In case of memory pressure, list_del() can be called twice on
2806 req->rq_history_list, causing a kernel oops.
2810 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2811 Details : fix stack overflow in the distributed lock manager by defering
2812 export eviction after a failed AST to the elt thread instead of
2813 handling it in the dlm interpret routine.
2817 Description: Convert some comments to new format.
2818 Details : Update documenting comments to match doxygen conventions.
2822 Description: Grammar fixes.
2823 Details : A couple of trivial spelling fixes.
2827 Description: OSD_COUNTERS-mandatory
2828 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2829 to simplify the code.
2833 Description: simplify cmm_device freeing logic.
2834 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2838 Description: Add lockdep support to dt_object_operations locking interface.
2839 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2840 parameter indicating lock ordering. Update mdd code to use new
2845 Description: Introduce failloc constants for lockless IO tests.
2846 Details : Add two new failloc constants to test lockless IO. Only one of
2847 them in implemented---another is checked in yet to be landed
2852 Description: Add lockdep support for inode mutex.
2853 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2854 in the situations where Lustre has to lock more than one inode
2859 Description: Add optional invariants checking support.
2860 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2861 configure switch. This macro is to be used for consistency and
2862 sanity checks that are too expensive to be left in `production'
2867 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2868 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2873 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2874 Details : Use spin_lock_nested() in (the only) situation where more than
2875 one ldlm_lock is locked simultaneously. Also, fix possible
2876 dead-lock in ldlm_lock_change_resource() by enforcing particular
2881 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2882 Details : Instead of specifying each ldlm_lock call-back through separate
2883 parameter, wrap them into struct ldlm_callback_suite.
2887 Description: Kill join_lru obd method and its callers.
2888 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2889 in memory---a requirement implemented through ->o_join_lru() obd
2890 method in HEAD. Remove this method and its users.
2894 Description: Add asynchronous ldlm ENQUEUE completion handler.
2895 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2896 case that stock ldlm_completion_ast() cannot handle as it waits
2897 until lock is granted. Introduce new ldlm_completion_ast_async()
2898 for this. Also comment ldlm_completion_ast().
2902 Description: ldlm_error <-> errno conversion.
2903 Details : Add functions to map (rather arbitrary) between LDLM error codes
2904 and standard errno values. CLIO needs this to prevent LDLM specific
2905 constants from escaping ldlm and osc.
2909 Description: Kill unused ldlm_handle2lock_ns() function.
2910 Details : Kill unused ldlm_handle2lock_ns() function.
2914 Description: Add lu_ref support to ldlm_lock
2915 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2916 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2917 and ldlm_resource. LDLM interface has to be changed, because code that
2918 releases a reference on a lock, has to "know" what reference this is.
2919 In the most frequent case
2921 lock = ldlm_handle2lock(handle);
2923 LDLM_LOCK_PUT(lock);
2925 no changes are required. When any other reference (received _not_ from
2926 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2927 instead of LDLM_LOCK_PUT().
2929 Arguably, changes are pervasive, and interface requires some discipline
2930 for proper use. On the other hand, it was very instrumental in finding
2931 a few leaked lock references.
2935 Description: Add ldlm_lock_addref_try().
2936 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2937 attempts to addref a lock that might be being canceled
2942 Description: Add ldlm_weigh_callback().
2943 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2944 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2945 hard-coded `number of pages' logic.
2949 Description: Add lockdep annotations to llog code.
2950 Details : Use appropriately tagged _nested() locking calls in the places
2951 where llog takes more than one ->lgh_lock lock.
2955 Description: Add loi_kms_set().
2956 Details : Wrap kms updates into a helper function.
2960 Description: Constify instances of struct lsm_operations.
2961 Details : Constify instances of struct lsm_operations.
2965 Description: lu_conf support.
2966 Details : On a server, a file system object is uniquely identified
2967 by a fid, which is sufficient to locate and load all object
2968 state (inode). On a client, on the other hand, more data are
2969 necessary instantiate an object. Change lu_object_find() and
2970 friends to take additional `lu_conf' argument describing object.
2971 Typically this includes layout information.
2975 Description: lu_context fixes.
2976 Details : Introduce new lu_context functions that are needed on the client
2977 side, where some system threads (ptlrpcd) are shared by multiple
2978 modules, and so cannot be stopped during module shutdown.
2982 Description: Add start and stop methods to lu_device_type_operations.
2983 Details : Introduce two new methods in lu_device_type_operations, that are
2984 invoked when first instance of a given type is created and last one
2985 is destroyed respectively. This is need by CLIO.
2989 Description: Add lu_ref support to struct lu_device.
2990 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2991 track leaked references.
2995 Description: Introduce lu_kmem_descr.
2996 Details : lu_kmem_descr and its companion interface allow to create
2997 and destroy a number of kmem caches at once.
3001 Description: Fix lu_object finalization race.
3002 Details : Fix a race between lu_object_find() finding an object and its
3003 concurrent finalization. This race is (most likely) not possible
3004 on the server, but might happen on the client.
3008 Description: Introduce lu_ref interface.
3009 Details : lu_ref is a debugging module allowing to track references to
3010 a given object. It is quite cpu expensive, and has to be
3011 explicitly enabled with --enable-lu_ref. See usage description
3016 Description: Factor lu_site procfs stats into a separate function.
3017 Details : Separate lu_site stats printing code into a separate function
3018 that can be reused on a client.
3022 Description: Constify instances of struct {lu,dt,md}_device_operations.
3023 Details : Constify instances of struct {lu,dt,md}_device_operations.
3027 Description: Introduce struct md_site and move meta-data specific parts of
3028 struct lu_site here.
3029 Details : Move md-specific fields out of struct lu_site into special struct
3030 md_site, so that lu_site can be used on a client.
3034 Description: Kill mdd_lov_destroy().
3035 Details : Remove unused mdd code.
3039 Description: Add st_block checking to multistat.c.
3040 Details : Add st_block checking to multistat.c.
3044 Description: Add lu_ref support to struct obd_device.
3045 Details : Add lu_ref tracking to obd_device.
3049 Description: Kill obd_set_fail_loc().
3050 Details : Remove unused code.
3054 Description: Add special type for ptlrpc_request interpret functions.
3055 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
3056 there. Actual usage will be in CLIO.
3060 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
3061 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
3062 doesn't work with lockdep.
3066 Description: Add rwv.c test program.
3067 Details : New testing program exercising readv(2) and writev(2) (Qian).
3071 Description: Add sendfile.c test program.
3072 Details : New testing program exercising sendfile(2) (Jay).
3076 Description: Ratelimit a message that can be very frequent.
3077 Details : Ratelimit a memory allocation failure message that can
3082 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3083 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3087 Description: Liblustre build fixes.
3088 Details : Liblustre build fixes.
3092 Description: libcfs: add cfs_{need,cond}_resched() interface.
3093 Details : libcfs: add cfs_{need,cond}_resched() definition and
3094 implementations for Linux, NT, and liblustre.
3096 Severity : enhancement
3098 Description: More exported tunables for mballoc
3099 Details : Add support for tunable preallocation window and new tunables for
3100 large/small requests
3104 Description: Detect corruption of block bitmap and checking for preallocations
3105 Details : Checks validity of on-disk block bitmap. Also it does better
3106 checking of number of applied preallocations. When corruption is
3107 found, it turns filesystem readonly to prevent further corruptions.
3111 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3112 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3113 window to 0. If it does hit cache miss, and read-pattern is still
3114 stride-io mode, does not reset the stride window, but also does not
3115 increase the stride window length in this case.
3119 Frequency : only for big-endian servers
3120 Description: Check if system is big-endian while mounting fs with extents feature
3121 Details : Mounting a filesystem with extents feature will fail on big-endian
3122 systems since ext3-based ldiskfs is not supported on big-endian
3123 systems. This can be over-riden with "bigendian_extents" mount option.
3125 Severity : enhancement
3127 Description: The root squash functionality
3128 Details : A security feature, which is to prevent users from being able
3129 to mount lustre on their desktop, run as root, and delete
3130 all of the files in the filesystem. The goal is accomplished by
3131 remapping user id (UID) and group id (GID) of the root user to
3132 a UID and GID specified by the system administartor via Lustre
3133 configuration management server (MGS). The functionality also
3134 allows to specify sets of clients for which the remapping does
3139 Description: Excessive recovery window
3140 Details : With AT enabled, the recovery window can be excessively long (6000+
3141 seconds). To address this problem, we no longer use
3142 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3143 timeout no longer depends on the service time, it is set to
3144 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3145 time via pb_service_time.
3149 Description: Watchdog triggered on MDS failover
3150 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3151 the OSTs know that the MDS "UUID" can be reused for the same export
3152 from a different NID, so we do not need to wait for the export to be
3156 Frequency : rare, only if using MMP with Linux RAID
3158 Description: MMP doesn't work with Linux RAID
3159 Details : While using HA for Lustre servers with Linux RAID, it is possible
3160 that MMP will not detect multiple mounts. To make this work we
3161 need to unplug the device queue in RAID when the MMP block is being
3162 written. Also while reading the MMP block, we should read it from
3163 disk and not the cached one.
3165 Severity : enhancement
3167 Description: open file using fid
3168 Details : A file can be opened using just its fid, like
3169 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3172 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3175 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3176 Details : add proper locking for creation of jbd_4k slab cache
3180 Description: MMP check in ext3_remount() fails without displaying any error
3181 Details : When multiple mount protection fails during remount, proper error
3184 Severity : enhancement
3186 Description: Allow stripe size to be up to 4G-64k
3187 Details : Fix math logic to allow large stripe sizes.
3191 Description: add check for >8TB ldiskfs filesystems
3192 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3193 Don't allow >8TB ldiskfs filesystems to be mounted without
3194 force_over_8tb mount option
3199 Descriptoin: OSS complains "too many blocks" and crashes
3200 Details : In filter_commitrw_write():
3201 1. lquota_chkquota();
3202 2. obdo_from_inode();
3203 3. lquota_pending_commit();
3204 "2." has changed uid and gid in this case, then caused
3205 lquota_chkquota() and lquota_pending_commit() to use
3206 different uid and gid. That is the root of the bug.
3210 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3211 Details : Detect and print out the 'stripe count = -1' case properly.
3215 Description: Out or order replies might be lost on replay
3216 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3217 request with tid smaller than one currently being added, add it
3218 to the start, not end of the list.
3220 --------------------------------------------------------------------------------
3222 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3224 * Support for kernels:
3225 2.6.5-7.283 (SLES 9),
3226 2.6.9-55.EL (RHEL 4),
3227 2.6.16.46-0.14 (SLES 10),
3228 2.6.18.8 vanilla (kernel.org)
3229 * Client support for unpatched kernels:
3230 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3231 2.6.16 - 2.6.22 vanilla (kernel.org)
3232 * Due to recently discovered recovery problems, we do not recommend
3233 using patchless RHEL 4 clients with this or any earlier release.
3234 * Recommended e2fsprogs version: 1.39.cfs8
3235 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3236 * Starting with this release, the ldiskfs backing filesystem required
3237 by Lustre is now in its own package, lustre-ldiskfs. This package
3238 should be installed. It is versioned separately from Lustre and
3239 may be released separately in future.
3243 Description: block reactivating mgc import until all deactivates complete
3244 Details : Fix race when failing back MDT/MGS to itself (testing)
3247 Frequency : at statup only
3249 Description: mds_lov_synchronize race leads to various problems
3250 Details : simultaneous MDT->OST connections at startup can cause the
3251 sync to abort, leaving the OSC in a bad state.
3253 Severity : enhancement
3255 Description: add optional extra BUILD_VERSION info
3256 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3257 to override the lustre version.
3260 Frequency : 2.6.18 servers only
3262 Description: ll_kern_mount() doesn't release the module reference
3263 Details : The ldiskfs module reference count never drops down to 0
3264 because ll_kern_mount() doesn't release the module reference.
3269 Description: server LBUG when using old ost_num_threads parameter
3270 Details : Accept the old ost_num_threads parameter but warn that it
3271 is deprecated, and fix an off-by-one error that caused an LBUG.
3276 Description: Transient SCSI error results in persistent IO issue
3277 Details : iobuf->dr_error is not reinitialized to 0 between two
3281 Frequency : sometimes when underlying device returns I/O errors
3283 Description: OSTs not going read-only during write failures
3284 Details : OSTs are not remounted read-only when the journal commit threads
3285 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3286 instead of the ext3 wrappers.
3290 Description: poor connect scaling with increasing client count
3291 Details : Don't run filter_grant_sanity_check for more than 100 exports
3292 to improve scaling for large numbers of clients.
3295 Frequency : SLES10 only
3297 Description: sanity-quota.sh quotacheck failed: rc = -22
3298 Details : Quotas cannot be enabled on SLES10.
3301 Frequency : liblustre clients only
3303 Description: getdirentries does not give error when run on compute nodes
3304 Details : getdirentries does not fail when the size specified as an argument
3305 is too small to contain at least one entry
3307 Severity : enhancement
3309 Description: Add LNET router traceability for debug purposes
3310 Details : If a checksum failure occurs with a router as part of the
3311 IO path, the NID of the last router that forwarded the bulk data
3312 is printed so it can be identified.
3317 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3318 Details : Due to a race condition, liblustre clients were occasionally
3319 evicted incorrectly.
3321 Severity : enhancement
3323 Description: lfs setstripe use optional parameters instead of postional
3326 Severity : enhancement
3328 Description: Nanosecond timestamp support for ldiskfs
3329 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3330 resolution timestamps. There is not yet support for this at
3331 the Lustre filesystem level.
3334 Frequency : during server recovery
3336 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3337 Details : request with rq_no_resend flag not awake l_wait_event if they get
3341 Frequency : nfs export on patchless client
3343 Description: connectathon hang when test nfs export over patchless client
3344 Details : Disconnected dentry cannot be found with lookup, so we do not need
3345 to unhash it or make it invalid
3348 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3349 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3350 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3355 Description: open req refcounting wrong on reconnect
3356 Details : If reconnect happened between getting open reply from server and
3357 call to mdc_set_replay_data in ll_file_open, we will schedule
3358 replay for unreferenced request that we are about to free.
3359 Subsequent close will crash in variety of ways.
3360 Check that request is still eligible for replay in
3361 mdc_set_replay_data().
3366 Description: disable writes to filesystem when reading health_check file
3367 Details : the default for reading the health_check proc file has changed
3368 to NOT do a journal transaction and write to disk, because this
3369 can cause reads of the /proc file to hang and block HA state
3370 checking on a healthy but otherwise heavily loaded system. It
3371 is possible to return to the previous behaviour during configure
3372 with --enable-health-write.
3374 Severity : enhancement
3376 Description: 64-bit inode version
3377 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3378 to the inode. This will be required for version-based recovery.
3383 Description: MDS fails to start if a duplicate client export is detected
3384 Details : in some rare cases it was possible for a client to connect to
3385 an MDS multiple times. Upon recovery the MDS would detect this
3386 and fail during startup. Handle this more gracefully.
3388 Severity : enhancement
3390 Description: Add -o localflock option to simulate old noflock
3392 Details : This will achieve local-only flock/fcntl locks
3398 Description: log_commit_thread vs filter_destroy race leads to crash
3399 Details : Take import reference before releasing llog record semaphore
3404 Description: Wrong request locking in request set processing
3405 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3406 imp_delayed_list, in this place should be used imp_lock.
3409 Frequency : when reconnection
3411 Description: Grant Leak when osc reconnect to OST
3412 Details : When osc reconnect ost, OST(filter) should check whether it
3413 should grant more space to client by comparing fed_grant and
3414 cl_avail_grant, and return the granted space to client instead
3415 of "new granted" space, because client will call osc_init_grant
3416 to update the client grant space info.
3419 Frequency : when client reconnect to OST
3421 Description: Grant Leak when osc do resend and replay bulk write
3422 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3423 bulk write request, because the grant info while be sync between
3424 OSC and OST when reconnect, and we should ignore the grant info
3425 these of resend/replay write req.
3430 Description: Grant space more than available left space sometimes.
3431 Details : When then OST is about to be full, if two bulk writing from
3432 different clients came to OST. Accord the avaliable space of the
3433 OST, the first req should be permitted, and the second one
3434 should be denied by ENOSPC. But if the seconde arrived before
3435 the first one is committed. The OST might wrongly permit second
3436 writing, which will cause grant space > available space.
3439 Frequency : when client is evicted
3441 Description: Grant might be wrongly erased when osc is evicted by OST
3442 Details : when the import is evicted by server, it will fork another
3443 thread ptlrpc_invalidate_import_thread to invalidate the
3444 import, where the grant will be set to 0. While the original
3445 thread will update the grant it got when connecting. So if
3446 the former happened latter, the grant will be wrongly errased
3447 because of this race.
3452 Description: Checking Stale with correct fid
3453 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3454 to check whether it is stale, because sometimes, we want the
3455 enqueue happened anyway, and op_data.fid2 will not be initialized.
3457 Severity : enhancement
3459 Description: update patchless client
3460 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3463 Frequency : only with 2.4 kernel
3465 Description: random memory corruption
3466 Details : size of struct ll_inode_info is to big for union inode.u and this
3467 can be cause of random memory corruption.
3472 Description: Memory leak in recovery
3473 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3474 It should also check obd_fail before fsfilt_start, otherwise if
3475 fsfilt_start return -EROFS,(failover mds during mds recovery).
3476 then the req will return with repmsg->transno = 0 and rc = EROFS.
3477 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3478 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3479 be freed no matter whether fsfilt_commit success or not.
3482 Frequency : only with huge count clients
3484 Description: Prevents from taking the superblock lock in llap_from_page for
3486 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3487 for llap_from_page call in ll_removepage prevents from taking the
3488 superblock lock for a soon died page.
3493 Description: Not check open intent error before release open handle
3494 Details : in some rare cases, the open intent error is not checked before
3495 release open handle, which may cause
3496 ASSERTION(open_req->rq_transno != 0), because it tries to release
3497 the failed open handle.
3502 Description: Set cat log bitmap only after create log success.
3503 Details : in some rare cases, the cat log bitmap is set too early. and it
3504 should be set only after create log success.
3508 Description: Accessing a block bevice can re-enable I/O when Lustre is
3509 tearing down a device.
3510 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3515 Description: service threads may hog cpus when there are a lot of requests
3516 Details : Insert cond_resched to give other threads a chance to use some CPU
3521 Description: the cat log was not initialized in recovery
3522 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3523 unlink log on mds will not be initialized until mds post
3524 recovery. And also in mds post recovery, the unlink log will
3525 initialization will be done asynchronausly, so there will be race
3526 between add unlink log and unlink log initialization.
3530 Description: brw_stats were being printed incorrectly
3531 Details : brw_stats were being printed as log2 but all of them were not
3532 recorded as log2. Also remove some code duplication arising from
3533 filter_tally_{read,write}.
3537 Frequency : rare, only in recovery.
3538 Description: ASSERTION(req->rq_type != LI_POISON) failed
3539 Details : imp_lock should be held while iterating over imp_sending_list for
3540 prevent destroy request after get timeout in ptlrpc_queue_wait.
3544 Description: replay-single.sh test 52 fails
3545 Details : A lock's skiplist need to be cleanup when it being unlinked
3546 from its resource list.
3550 Description: Short directio read returns full requested size rather than
3552 Details : Direct I/O operations should return actual amount of bytes
3553 transferred rather than requested size.
3555 Severity : enhancement
3557 Description: metadata RPC reduction (e.g. for rm performance)
3558 Details : decrease the amount of synchronous RPC between clients and servers
3559 by canceling conflicing lock before the operation on the client
3560 and packing thier handles into the main operation RPC to server.
3562 Severity : enhancement
3564 Description: Async OSC create to avoid the blocking unnecessarily.
3565 Details : If an OST has no remaining object, system will block on the
3566 creation when it needs to create a new object on this OST. Now,
3567 ways use pre-created objects when available, instead of blocking on
3568 an empty osc while others are not empty. If we must block, we
3569 block for the shortest possible period of time.
3573 Description: improve handling recoverable errors
3574 Details : if request processig with error which can be recoverable on server
3575 request should be resend, otherwise page released from cache and
3578 Severity : enhancement
3580 Description: refine locking for avoid write wrong info into lov_objid file
3581 Details : fix possible races with add new target and write/update data in
3584 --------------------------------------------------------------------------------
3586 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3591 Frequency : on some architectures
3593 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3594 Details : Uninitialized flags sometimes cause configuration commands to
3598 Frequency : patchless clients only
3600 Description: missing __iget() symbol export
3601 Details : The __iget() symbol export is missing. To avoid the need for
3602 this on patchless clients the deathrow inode reaper is turned
3603 off, and we depend on the VM to clean up old inodes. This
3604 dependency was during via the fix for bug 12181.
3606 --------------------------------------------------------------------------------
3608 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3610 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3611 INTEROPERATE with older versions automatically. In many cases a
3612 special upgrade step is needed. Please read the
3613 user documentation before upgrading any part of a live system.
3614 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3615 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3616 * WARNING: Lustre configuration and startup changes are required with
3617 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3621 Severity : enhancement
3623 Description: MountConf
3624 Details : Lustre configuration is now managed via mkfs and mount
3625 commands instead of lmc and lconf. New obd types (MGS, MGC)
3626 are added for dynamic configuration management. See
3627 https://mail.clusterfs.com/wikis/lustre/MountConf for
3630 Severity : enhancement
3632 Description: dynamic OST addition
3633 Details : OSTs can now be added to a live filesystem
3635 Severity : enhancement
3637 Description: startup order invariance
3638 Details : MDTs and OSTs can be started in any order. Clients only
3639 require the MDT to complete startup.
3641 Severity : enhancement
3643 Description: parallel, asynchronous orphan cleanup
3644 Details : orphan cleanup is now performed in separate threads for each
3645 OST, allowing parallel non-blocking operation.
3647 Severity : enhancement
3649 Description: optimized stripe assignment
3650 Details : stripe assignments are now made based on ost space available,
3651 ost previous usage, and OSS previous usage, in order to try
3652 to optimize storage space and networking resources.
3654 Severity : enhancement
3656 Description: Permanently set tunables
3657 Details : All writable /proc/fs/lustre tunables can now be permanently
3658 set on a per-server basis, at mkfs time or on a live system.
3660 Severity : enhancement
3662 Description: Lustre message v2
3663 Details : Add lustre message format v2.
3665 Severity : enhancement
3667 Description: client OST exclusion list
3668 Details : Clients can be started with a list of OSTs that should be
3669 declared "inactive" for known non-responsive OSTs.
3673 Description: SPEC SFS validation failure on NFS v2 over lustre.
3674 Details : Changes the blocksize for regular files to be 2x RPC size,
3675 and not depend on stripe size.
3677 Severity : enhancement
3679 Description: Multiple MD RPCs in flight.
3680 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3681 To avoid overly-overloading MDS, introduce a limit on number of
3682 MDS RPCs in flight for a single client and add /proc controls
3683 to adjust this limit.
3685 Severity : enhancement
3687 Description: client read/write statistics
3688 Details : Add client read/write call usage stats for performance
3689 analysis of user processes.
3690 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3691 file access. extents_stats shows chunk size distribution.
3692 extents_stats_per_process show chunk size distribution per
3695 Severity : enhancement
3697 Description: mds statistics
3698 Details : Add detailed mds operations statistics in
3699 /proc/fs/lustre/mds/*/stats.
3703 Description: Failure of copying files with lustre special EAs.
3704 Details : Client side always return success for setxattr call for lustre
3705 special xattr (currently only "trusted.lov").
3710 Description: Refcount LNET uuids
3711 Details : The global LNET uuid list grew linearly with every startup;
3712 refcount repeated list entries instead of always adding to
3715 Severity : enhancement
3717 Description: Dynamic service threads
3718 Details : Within a small range, start extra service threads
3719 automatically when the request queue builds up.
3722 Frequency : mixed-endian client/server environments
3724 Description: mixed-endian crashes
3725 Details : The new msg_v2 system had some failures in mixed-endian
3728 Severity : enhancement
3730 Description: Easy OST removal
3731 Details : OSTs can be permanently deactivated with e.g. 'lctl
3732 conf_param lustre-OST0001.osc.active=0'
3734 Severity : enhancement
3736 Description: MGS proc entries
3737 Details : Added basic proc entries for the MGS showing what filesystems
3740 Severity : enhancement
3742 Description: provide MGS failover
3743 Details : Added config lock reacquisition after MGS server failover.
3745 Severity : enhancement
3747 Description: add Linux 2.4 support
3748 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3752 Description: a large application tries to do I/O to the same resource and dies
3753 in the middle of it.
3754 Details : Check the req->rq_arrival time after the call to
3755 ost_brw_lock_get(), but before we do anything about
3756 processing it & sending the BULK transfer request. This
3757 should help move old stale pending locks off the queue as
3758 quickly as obd_timeout.
3761 Frequency : when an incorrect nid is specified during startup
3763 Description: ptlrpc connect to non-existant node causes kernel crash
3764 Details : LNET can't be re-entered from an event callback, which
3765 happened when we expire a message after the export has been
3766 cleaned up. Instead, hand the zombie cleanup off to another
3769 Severity : enhancement
3771 Description: plain/inodebits lock performance improvement
3772 Details : Grouping plain/inodebits in granted list by their request modes
3773 and bits policy, thus improving the performance of search through
3777 Frequency : only if OST filesystem is corrupted
3779 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3780 Details : for a short time RPCs with bulk IO are in the replay list,
3781 but replay of bulk IOs is unimplemented. If the OST filesystem
3782 is corrupted due to disk cache incoherency and then replay is
3783 started it is possible to trip an assertion. Avoid putting
3784 committed RPCs into the replay list at all to avoid this issue.
3787 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3789 Description: System hang on startup
3790 Details : This bug allowed the liblustre (e.g. catamount) client to
3791 return to the app before handling all startup RPCs. This
3792 could leave the node unresponsive to lustre network traffic
3793 and manifested as a server ptllnd timeout.
3795 Severity : enhancement
3797 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3798 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3799 boolean to control whether to print peer debug info when a
3800 client's RPC times out.
3803 Frequency : only for kernels with patches from Lustre below 1.4.3
3805 Description: Remove old rdonly API
3806 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3809 Frequency : only for devices with external journals
3811 Description: Set external device read-only also
3812 Details : During a commanded failover stop, we set the disk device
3813 read-only while the server shuts down. We now also set any
3814 external journal device read-only at the same time.
3817 Frequency : when upgrading from 1.4 while trying to change parameters
3819 Description: The wrong (new) MDC name was used when setting parameters for
3820 upgraded MDT's. Also allows changing of OSC (and MDC)
3821 parameters if --writeconf is specified at tunefs upgrade time.
3824 Frequency : when setting specific ost indicies
3826 Description: QOS code breaks on skipped indicies
3827 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3828 created with --index need not be sequential.
3830 Severity : enhancement
3832 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3833 Details : The uninit_groups feature works in conjunction with the kernel
3834 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3835 up the pass1 processing of e2fsck. This is a read-only feature
3836 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3837 mount filesystems that have had this feature enabled.
3839 Severity : enhancement
3841 Description: Improve multi-block allocation algorithm to avoid fragmentation
3842 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3843 allocation locality and avoid filesystem fragmentation.
3845 ------------------------------------------------------------------------------
3847 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3849 * Support for kernels:
3850 2.4.21-47.0.1.EL (RHEL 3)
3851 2.6.5-7.283 (SLES 9)
3852 2.6.9-42.0.10.EL (RHEL 4)
3853 2.6.12.6 vanilla (kernel.org)
3854 2.6.16.27-0.9 (SLES 10)
3855 * Recommended e2fsprogs version: 1.39.cfs5
3857 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3861 Frequency : occasional, depends on client load and configuration
3862 Bugzilla : 12181, 12203
3863 Description: data loss for recently-modified files
3865 Details : In some cases it is possible that recently written or created
3866 files may not be written to disk in a timely manner (this should
3867 normally be within 30s unless client IO load is very high).
3868 The problem appears as zero-length files or files that are a
3869 multiple of 1MB in size after a client crash or client eviction
3870 that are missing data at the end of the file.
3872 This problem is more likely to be hit on clients where files are
3873 repeatedly created and unlinked in the same directory, clients
3874 have a large amount of RAM, have many CPUs, the filesystem has
3875 many OSTs, the clients are rebooted frequently, and/or the files
3876 are not accessed by other nodes after being written.
3878 The presence of the problem can be detected by looking at
3879 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3880 smaller than the second (nr_unused) then dirty files will not
3881 be flushed automatically to disk. "sync; sleep 10" should be
3882 run several times on the node before unmounting it to update
3883 Lustre (this is also safe to run on nodes without this problem).
3885 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3886 that can cause this same problem, so customers using that kernel
3887 also need to update the kernel in addition to Lustre. In order
3888 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3890 It is normal that files written just before a client crash (less
3891 than 30s) may not yet have been flushed to disk, even for local
3895 Frequency : frequent on thin XT3 nodes
3897 Description: UUID collision on thin XT3 Linux nodes
3898 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3899 randomly, since we relied on an insufficiently-seeded PRNG.
3904 Description: OSS hangs after "All ost request buffers busy"
3905 Details : A deadlock between quota and journal operations caused OSS
3906 hangs after printing "All ost request buffers busy."
3909 Frequency : always on liblustre builds
3911 Description: Cleanup compiler warnings on liblustre
3914 Frequency : always on liblustre builds on XT3
3916 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3920 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3921 > 32000 subdirectories
3922 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3923 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3924 subdirectory count crosses 32000. This will aid e2fsck to
3925 correctly handle more than 32000 subdirectories.
3928 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3930 Description: System hang on startup
3931 Details : This bug allowed the liblustre (e.g. catamount) client to
3932 return to the app before handling all startup RPCs. This
3933 could leave the node unresponsive to lustre network traffic
3934 and manifested as a server ptllnd timeout.
3936 Severity : enhancement
3938 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3939 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3940 boolean to control whether to print peer debug info when a
3941 client's RPC times out.
3946 Description: make O_SYNC working on 2.6 kernels
3947 Details : 2.6 kernels use different method for mark pages for write,
3948 so need add a code to lustre for O_SYNC work.
3953 Description: Failure to close file and release space on NFS
3954 Details : Put inode details into lock acquired in ll_intent_file_open.
3955 Use mdc_intent_lock in ll_intent_open to properly
3956 detect all kind of errors unhandled by mdc_enqueue.
3961 Description: proc file read during shutdown sometimes raced obd removal,
3963 Details : Add lock to prevent obd access after proc file removal.
3966 Frequency : Only for files larger than 4GB on 32-bit clients.
3968 Description: improperly doing page alignment of locks
3969 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3970 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3975 Description: RPCs being resent when they shouldn't be
3976 Details : Some RPCs that should not be resent are being resent. This
3977 can cause inconsistencies in the RPC state machine. Do not
3978 resend such requests.
3981 Frequency : rare, only with NFS export
3983 Description: Crash on NFS re-export node
3984 Details : under very unusual load conditions an assertion is hit in
3985 ll_intent_file_open()
3988 Frequency : only if OST filesystem is corrupted
3990 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3991 Details : for a short time RPCs with bulk IO are in the replay list,
3992 but replay of bulk IOs is unimplemented. If the OST filesystem
3993 is corrupted due to disk cache incoherency and then replay is
3994 started it is possible to trip an assertion. Avoid putting
3995 committed RPCs into the replay list at all to avoid this issue.
4000 Description: large O_DIRECT requests fail under memory pressure/fragmentation
4001 Details : Large single O_DIRECT read and write calls can fail to allocate
4002 a sufficiently large buffer to process the request. In case of
4003 allocation failure the allocation is retried with a smaller
4004 buffer and broken into smaller requests.
4006 Severity : enhancement
4008 Description: Add -o localflock option to simulate old noflock behaviour.
4009 Details : This will achieve local-only flock/fcntl locks coherentness.
4014 Description: versioning check is incomplete
4015 Details : Checking the version difference of client vs. server, report
4016 error if the gap is too big.
4021 Description: add support PG_writeback bit
4022 Details : add support for PG_writeback bit for Lustre, for more carefull
4023 work with page cache in 2.6 kernel. This also fix some deadlocks
4024 and remove hack for work O_SYNC with 2.6 kernel.
4026 Severity : enhancement
4028 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4029 Details : The uninit_groups feature works in conjunction with the kernel
4030 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4031 up the pass1 processing of e2fsck. This is a read-only feature
4032 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4033 mount filesystems that have had this feature enabled.
4035 Severity : enhancement
4037 Description: Improve multi-block allocation algorithm to avoid fragmentation
4038 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4039 allocation locality and avoid filesystem fragmentation.
4041 ------------------------------------------------------------------------------
4043 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
4045 * Support for kernels:
4046 2.6.9-42.0.3.EL (RHEL 4)
4047 2.6.5-7.276 (SLES 9)
4048 2.4.21-47.0.1.EL (RHEL 3)
4049 2.6.12.6 vanilla (kernel.org)
4050 2.6.16.21-0.8 (SLES10)
4051 * Recommended e2fsprogs version: 1.39.cfs2-0
4053 * The backwards-compatible /proc/sys/portals symlink has been removed
4054 in this release. Before upgrading, please ensure that you change
4055 any configuration scripts or /etc/sysctl.conf files that access
4056 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4057 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4058 in advance of the upgrade on any system running Lustre 1.4.6 or
4059 newer, since /proc/sys/lnet was added in that version.
4060 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4064 Frequency : only when quota is used
4066 Description: avoid scanning export list for quota master
4067 Details : Change the algorithms to avoid scanning export list in order
4068 to improve the efficiency.
4071 Frequency : MDS failover only, very rarely
4073 Description: "went back in time" messages on mds failover
4074 Details : The greatest transno may be lost when the current operation
4075 finishes with an error (transno==0) and the client's last_rcvd
4076 record is over-written. Save the greatest transno in the
4077 mds_last_transno for this case.
4080 Frequency : always for specific kernels and striping counts
4082 Description: client may get "Matching packet too big" without ACL support
4083 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4084 when trying to access files in certain configurations. The
4085 clients should in fact be denied when mounting because they do
4086 not understand ACLs.
4089 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4091 Description: many clients connecting with IO in progress causes connect timeouts
4092 Details : Avoid synchronous journal commits to avoid delays caused by many
4093 clients connecting/disconnecting when bulk IO is in progress.
4094 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4095 OST_IO_PORTAL to avoid delays behind potentially many pending
4099 Frequency : occasionally with multiple writers to a single file
4101 Description: shared writes to file may result in wrong size reported by stat()
4102 Details : Allow growing of kms when extent lock is cancelled
4105 Frequency : always with random mmap IO to multi-striped file
4107 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4108 Details : Only if the hole is at the end of OST object so that kms is too
4109 small. Fix is to increase kms accordingly in ll_nopage.
4112 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4114 Description: writes to a missing object would leak memory on the OST
4115 Details : If there is an inconsistency between the MDS and OST filesystems,
4116 such that the MDS references an object that doesn't exist, writes
4117 to that object will leak memory due to incorrect cleanup in the
4118 error handling path, eventually running out of memory on the OST.
4123 Description: Creating too long symlink causes lustre errors
4124 Details : Check symlink and name lengths before sending requests to MDS.
4127 Frequency : only if flock is enabled (not on by default)
4129 Description: posix locks not released on fd closure on 2.6.9+
4130 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4131 kernels, this caused such locks not to be released on fd close and
4132 then assertions on fs unmount about still used locks.
4135 Frequency : MDS failover only, very rarely
4137 Description: clients may get ASSERTION(granted_lock != NULL)
4138 Details : When request was taking a long time, and a client was resending
4139 a getattr by name lock request. The were multiple lock requests
4140 with the same client lock handle and
4141 mds_getattr_name->fixup_handle_for_resent_request found one of the
4142 lock handles but later failed with ASSERTION(granted_lock != NULL).
4147 Description: handle->h_buffer_credits > 0, assertion failure
4148 Details : h_buffer_credits is zero after truncate, causing assertion
4149 failure. This patch extends the transaction or creates a new
4153 Frequency : NFS re-export or patchless client
4154 Bugzilla : 11179, 10796
4155 Description: Crash on NFS re-export node (__d_move)
4156 Details : We do not want to hash the dentry if we don't have a lock.
4157 But if this dentry is later used in d_move, we'd hit uninitialised
4158 list head d_hash, so we just do this to init d_hash field but
4159 leave dentry unhashed.
4162 Frequency : NFS re-export or patchless client
4164 Description: NFS exports has problem with symbolic link
4165 Details : lustre client didn't properly install dentry when re-exported
4166 to NFS or running patchless client.
4169 Frequency : NFS re-export or patchless client
4171 Description: Various nfs/patchless fixes.
4172 Details : fixes reuse disconected alias for lookup process - this fixes
4173 warning "find_exported_dentry: npd != pd",
4174 fix permission error with open files at nfs.
4175 fix apply umask when do revalidate.
4178 Frequency : occasional
4180 Description: Crash on NFS re-export node
4181 Details : calling clear_page() on the wrong pointer triggered oops in
4182 generic_mapping_read().
4185 Frequency : rarely, using O_DIRECT IO
4187 Description: unaligned directio crashes client with LASSERT
4188 Details : check for unaligned buffers before trying any requests.
4191 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4193 Description: stale data returned from RAID cache
4194 Details : If only a small amount of IO is done to the RAID device before
4195 reading it again it is possible to get stale data from the RAID
4196 cache instead of reading it from disk.
4199 Frequency : always for sles10 kernel
4201 Description: sles10 support
4202 Details : ll_follow_link: compile fixes and using of nd_set_link
4203 under newer kernels.
4206 Frequency : depends on arch, kernel and compiler version, always on sles10
4209 Description: recursive or deep enough symlinks cause stack overflow
4210 Details : getting rid of large stack-allocated variable in
4214 Frequency : depends on hardware
4216 Description: lustre write performance loss in the SLES10 kernel
4217 Details : the performance loss is caused by using of write barriers in the
4218 ext3 code. The SLES10 kernel turns barrier support on by
4219 default. The fix is to undo that change for ldiskfs.
4221 ------------------------------------------------------------------------------
4223 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4225 * Support for kernels:
4226 2.6.9-42.0.3EL (RHEL 4)
4227 2.6.5-7.276 (SLES 9)
4228 2.4.21-47.0.1.EL (RHEL 3)
4229 2.6.12.6 vanilla (kernel.org)
4233 Frequency : quota enabled and large files being deleted
4235 Description: releasing more than 4GB of quota at once hangs OST
4236 Details : If a user deletes more than 4GB of files on a single OST it
4237 will cause the OST to spin in an infinite loop. Release
4238 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4243 Description: statfs data retrieved from /proc may be stale or zero
4244 Details : When reading per-device statfs data from /proc, in the
4245 {kbytes,files}_{total,free,avail} files, it may appear
4246 as zero or be out of date.
4249 Frequency : systems with MD RAID1 external journal devices
4251 Description: lconf's call to blkid is confused by RAID1 journal devices
4252 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4253 of returning all block devices that match the journal UUID.
4256 Frequency : always, for aggregate stripe size over 4GB
4258 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4259 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4260 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4263 Frequency : always if "lfs find" used on a local file/directory
4265 Description: "lfs find" segfaults if used on a local file/directory
4266 Details : The case where a directory component was not specified wasn't
4267 handled correctly. Handle this properly.
4270 Frequency : always on ppc64
4272 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4273 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4274 perform endianness conversion.
4277 Frequency : rarely (truncate to non-zero file size after write under load)
4278 Bugzilla : 10730, 10687
4279 Description: Files padded with zeros to next 4K multiple
4280 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4281 it is possible that files that are truncated to a non-zero size
4282 immediately after being written are filled with zero bytes beyond
4283 the truncated size. No file data is lost.
4285 Severity : enhancement
4287 Description: Allow recovery/failover for liblustre clients.
4288 Details : liblustre clients were unaware of failover configurations until
4291 Severity : enhancement
4293 Description: user file locks should fail when not mounting with flock option
4294 Details : Set up an error-returning stub in ll_file_operations.lock field
4295 to prevent incorrect behaviour when client is mounted without
4296 flock option. Also, set up properly f_op->flock field for
4300 Frequency : always on ia64
4302 Description: "lfs df" loops on printing out MDS statfs information
4303 Details : The obd_ioctl_data was not initialized and in some systems
4304 this caused a failure during the ioctl that did not return
4305 an error. Initialize the struct and return an error on failure.
4308 Frequency : SLES 9 only
4310 Description: Error of copying files with lustre special EAs as root
4311 Details : Client side always return success for setxattr call for lustre
4312 special xattr (currently only "trusted.lov").
4315 Frequency : rarely on clusters with both ia64+i386 clients
4317 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4318 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4319 clients are doing concurrent writes to the same file it is
4320 possible that the ia64 clients may LASSERT because the OST
4321 extent locks are not PAGE_SIZE aligned. Ensure that grown
4322 locks are always aligned on the request boundary.
4325 Frequency : specific use, occasional
4327 Description: Overwriting in use executable truncates on-disk binary image
4328 Details : If one node attempts to overwrite an executable in use by
4329 another node, we now correctly return ETXTBSY instead of
4330 truncating the file.
4332 Severity : enhancement
4334 Description: Async OSC create to avoid the blocking unnecessarily.
4335 Details : If an OST has no remaining object, system will block on the creating
4336 when need to create a new object on this OST. Now, ways use
4337 pre-created objects when available, instead of blocking on an
4338 empty osc while others are not empty. If we must block, we block
4339 for the shortest possible period of time.
4344 Description: chmod on Lustre root is propagated to other clients
4345 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4346 invalid by the follow_mount time.
4349 Frequency : liblustre clients only
4351 Description: Race in 'instant cancel' lock handling could lead to such locks
4352 never to be granted in case of SMP MDS
4353 Details : Do not destroy not yet granted but cbpending locks in
4357 Frequency : replay/resend of open
4359 Description: non null lock assetion failure in mds_intent_policy
4360 Details : Trying to replay/resend lockless open requests resulted in
4361 mds_open() returning 0 with no lock. Now it sets a flag if
4362 a lock is going to be returned.
4364 Severity : enhancement
4366 Description: Checksum enhancements
4367 Details : New checksum enhancements allow for resending RPCs that failed
4370 Severity : enhancement
4372 Description: Tunables on number of dirty pages in cacche
4373 Details : Allow to set limit on number of dirty pages cached.
4378 Description: client crash on unmount - lock still has references
4379 Details : In some error handling cases it was possible to leak a lock
4380 reference on a client while accessing a file. This was not
4381 harmful to the client during operation, but would cause the
4382 client to crash when the filesystem is unmounted.
4385 Frequency : specific case, rare
4387 Description: ETXTBSY on mds though file not in use by client
4388 Details : ETXTBSY is no longer incorrectly returned when attempting to
4389 chmod or chown a directory that the user previously tried to
4390 execute or a currently-executing binary.
4393 Frequency : extremely rare except on liblustre-based clients
4395 Description: Lustre space not freed when files are deleted
4396 Details : Clean up open-unlinked files after client eviction. Previously
4397 the unlink was skipped and the files remained as orphans.
4402 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4403 Details : In some cases it was possible to send a blocking callback to
4404 a client doing a glimpse, even though that client didn't get
4405 a lock granted. When the glimpse lock is cancelled on the OST
4406 the freed lock is left on the waiting list and corrupted the list.
4409 Frequency : all core dumps
4411 Description: Broke core dumps to lustre
4412 Details : Negative dentry may be unhashed if parent does not have UPDATE
4413 lock, but some callers, e.g. do_coredump, expect dentry to be
4414 hashed after successful create, hash it in ll_create_it.
4416 ------------------------------------------------------------------------------
4418 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4420 * Support for kernels:
4421 2.6.9-42.0.2.EL (RHEL 4)
4422 2.6.5-7.276 (SLES 9)
4423 2.4.21-40.EL (RHEL 3)
4424 2.6.12.6 vanilla (kernel.org)
4428 Frequency : always on RHEL 3
4430 Description: Number of open files grows over time
4431 Details : The number of open files grows over time, whether or not
4432 Lustre is started. This was due to a filp leak introduced
4433 by one of our kernel patches.
4435 ------------------------------------------------------------------------------
4437 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4439 * Support for kernels:
4440 2.6.9-42.EL (RHEL 4)
4441 2.6.5-7.276 (SLES 9)
4442 2.4.21-40.EL (RHEL 3)
4443 2.6.12.6 vanilla (kernel.org)
4448 Bugzilla : 5719, 9635, 9792, 9684
4449 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4450 Details : If a server is under heavy load and cannot reply to new
4451 connection requests before the client resends the (re)connect,
4452 the connection handling code can behave badly if two service
4453 threads are concurrently handing separate (re)connections from
4454 the same client. Add better locking to the connection handling
4455 code, and ensure that only a single connection will be processed
4456 for a given client UUID, even if the lock is dropped.
4458 Severity : enhancement
4460 Description: add TCP zero-copy support to kernel
4461 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4462 sends if the hardware supports scatter-gather and checksumming.
4463 This allows socklnd to do client-write and server-read more
4464 efficiently and reduce CPU utilization from skbuf copying.
4467 Frequency : only if NFS exporting from client
4469 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4470 Details : If setting attributes on a file created under NFS that had
4471 never been opened it would be possible to oops the client
4472 if the file had no objects.
4475 Frequency : always for liblustre
4477 Description: liblustre client does MDS+OSTs setattr RPC for each write
4478 Details : When doing a write from a liblustre client, the client
4479 incorrectly issued an RPC to the MDS and each OST the file was
4480 striped over in order to update the timestamps. When writing
4481 with small chunks and many clients this could overwhelm the MDS
4482 with RPCs. In all cases it would slow down the write because
4483 these RPCs are unnecessary.
4485 Severity : enhancement
4487 Description: allow number of MDS service threads to be changed at module load
4488 Details : It is now possible to change the number of MDS service threads
4489 running. Adding "options mds mds_num_threads={N}" to the MDS's
4490 /etc/modprobe.conf will set the number of threads for the next
4491 time Lustre is restarted (assuming the "mds" module is also
4492 reloaded at that time). The default number of threads will
4493 stay the same, 32 for most systems.
4498 Description: OST crash if filesystem is unformatted or corrupt
4499 Details : If an OST is started on a device that has never been formatted
4500 or if the filesystem is corrupt and cannot even mount then the
4501 error handling cleanup routines would dereference a NULL pointer.
4506 Description: NULL pointer deref in llap_from_page.
4507 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4508 incorrect) mapping if the page was truncated/reclaimed while it was
4509 searched for. Check for this condition and skip such pages when
4510 doing readahead. Introduce extra check to llap_from_page() to
4511 verify page->mapping->host is non-NULL (so page is not anonymous).
4514 Frequency : Sometimes when using sys_sendfile
4516 Description: "page not covered by a lock" warnings from ll_readpage
4517 Details : sendfile called ll_readpage without right page locks present.
4518 Now we introduced ll_file_sendfile that does necessary locking
4519 around call to generic_file_sendfile() much like we do in
4523 Frequency : with certain MDS communication failures at client mount time
4525 Description: NULL pointer deref after failed client mount
4526 Details : a client connection request may delayed by the network layer
4527 and not be sent until after the PTLRPC layer has timed out the
4528 request. If the client fails the mount immediately it will try
4529 to clean up before the network times out the request. Add a
4530 reference from the request import to the obd device and delay
4531 the cleanup until the network drops the request.
4534 Frequency : occasionally during client (re)connect
4536 Description: assertion failure during client (re)connect
4537 Details : processing a client connection request may be delayed by the
4538 client or server longer than the client connect timeout. This
4539 causes the client to resend the connection request. If the
4540 original connection request is replied in this interval, the
4541 client may trip an assertion failure in ptlrpc_connect_interpret()
4542 which thought it would be the only running connect process.
4545 Frequency : only with obd_echo servers and clients that are rebooted
4547 Description: kernel BUG accessing uninitialized data structure
4548 Details : When running an obd_echo server it did not start the ping_evictor
4549 thread, and when a client was evicted an uninitialized data
4550 structure was accessed. Start the ping_evictor in the RPC
4551 service startup instead of the OBD startup.
4553 Severity : enhancement
4554 Bugzilla : 10193 (patchless)
4555 Description: Remove dependency on various unexported kernel interfaces.
4556 Details : No longer need reparent_to_init, exit_mm, exit_files,
4557 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4560 Frequency : rare (only users of deprecated and unsupported LDAP config)
4562 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4563 Details : LDAP apparently contains 'lustreName' attributes instead of
4564 'name'. A simple remapping of the name is sufficient.
4567 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4569 Description: waiting_locks_callback trips kernel BUG if client is evicted
4570 Details : Running with the dump_on_timeout debug flag turned on makes
4571 it possible that the waiting_locks_callback() can try to dump
4572 the Lustre kernel debug logs from an interrupt handler. Defer
4573 this log dumping to the expired_lock_main() thread.
4575 Severity : enhancement
4577 Description: Support NFS exporting on 2.6 kernels.
4578 Details : Implement non-rawops metadata methods for NFS server to use without
4579 changing NFS server code.
4582 Frequency : very rare (synthetic metadata workload only)
4584 Description: two racing renames might cause an MDS thread to deadlock
4585 Details : Running the "racer" program may cause one MDS thread to rename
4586 a file from being the source of a rename to being the target of
4587 a rename at exactly the same time that another thread is doing
4588 so, and the second thread has already enqueued these locks after
4589 doing a lookup of the target and is trying to relock them in
4590 order. Ensure that we don't try to re-lock the same resource.
4593 Frequency : only very large systems with liblustre clients
4595 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4596 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4597 all OSTs in parallel. This allows the eviction of stale liblustre
4598 clients to proceed much faster than if they were done in series,
4599 and also offers similar improvements for other set_info RPCs.
4604 Description: excessive CPU usage during initial read phase on client
4605 Details : During the initial read phase on a client, it would agressively
4606 retry readahead on the file, consuming too much CPU and impacting
4607 performance (since 1.4.5.8). Improve the readahead algorithm
4608 to avoid this, and also improve some other common cases (read
4609 of small files in particular, where "small" is files smaller than
4610 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4615 Description: MDS crash when receiving packet with unknown intent.
4616 Details : Do not LBUG in unknown intent case, just return -EFAULT
4618 Severity : enhancement
4619 Bugzilla : 9293, 9385
4620 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4621 Details : Do not serialize getattr (non-intent version) and statfs.
4624 Frequency : occasional, when OST network is overloaded/intermittent
4626 Description: client evicted by OST after bulk IO timeout
4627 Details : If a client sends a bulk IO request (read or write) the OST
4628 may evict the client if it is unresposive to its data GET/PUT
4629 request. This is incorrect if the network is overloaded (takes
4630 too long to transfer the RPC data) or dropped the OST GET/PUT
4631 request. There is no need to evict the client at all, since
4632 the pinger and/or lock callbacks will handle this, and the
4633 client can restart the bulk request.
4636 Frequency : Always when mmapping file with no objects
4638 Description: client crashes when mmapping file with no objects
4639 Details : Check that we actually have objects in a file before doing any
4640 operations on objects in ll_vm_open, ll_vm_close and
4646 Description: Request leak when working with deleted CWD
4647 Details : Introduce advanced request refcount tracking for requests
4648 referenced from lustre intent.
4650 Severity : Enhancement
4652 Description: Cache open file handles on client.
4653 Details : MDS now will return special lock along with openhandle, if
4654 requested and client is allowed to hold openhandle, even if unused,
4655 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4656 closing files for every read/write openration.
4658 Severity : Enhancement
4660 Description: Cache open negative dentries on client when possible.
4661 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4662 negative dentries on lock revocation.
4667 Description: Remounting a client read-only wasn't possible with a zconf mount
4668 Details : It wasn't possible to remount a client read-only with llmount.
4670 Severity : enhancement
4671 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4672 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4673 MPICH is included to add Lustre-specific ADIO interfaces.
4674 This is based closely on the UFS ADIO layer and only differs
4675 in file creation, in order to allow the OST striping to be set.
4676 This is user-contributed code and not supported by CFS.
4681 Description: extended inode attributes (immutable, append-only) work improperly
4682 when 2.4 and 2.6 kernels are used on client/server or vice versa
4683 Details : Introduce kernel-independent values for these flags.
4685 Severity : enhancement
4688 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4689 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4690 granularity. Now, Fractional MB granularity is be supported,
4691 this is very useful for low memory system.
4693 Severity : enhancement
4695 Description: Getattr by fid
4696 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4697 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4698 have locks and avoids getattr rpc requests for it on every stat.
4701 Frequency : Always, for filesystems larger than 2TB
4703 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4704 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4705 filesystems larger than 2TB to be mounted with mballoc enabled.
4708 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4710 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4711 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4712 filesystems larger than 2TB is mounted then the kernel will
4713 silently corrupt the start of the filesystem. CONFIG_LBD is
4714 enabled for all CFS-supported kernels, but the possibility of
4715 this happening with a modified kernel config exists.
4717 Severity : enhancement
4719 Description: add client O_DIRECT support for 2.6 kernels
4720 Details : It is now possible to do O_DIRECT reads and writes to files
4721 in the Lustre client mountpoint on 2.6 kernel clients.
4723 Severity : enhancement
4725 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4726 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4727 parallel, not waiting for response from every OST before sending
4728 a rpc to the next OST.
4733 Description: setattr vs write race when updating file timestamps
4734 Details : Client processes that update a file timestamp into the past
4735 right after writing to the file (e.g. tar) it is possible that
4736 the updated file modification time can be reset to the current
4737 time due to a race between processing the setattr and write RPC.
4739 Severity : enhancement
4741 Description: Bring 'lfs find' closer in line with regular Linux find.
4742 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4743 -print0 options and obtains all the needed info through the lustre
4746 Severity : enhancement
4748 Description: support up to 1024 configured devices on one node
4749 Details : change obd_dev array from statically allocated to dynamically
4750 allocated structs as they are first used to reduce memory usage
4755 Description: Flush dirty partially truncated pages during truncate
4756 Details : Immediatelly flush partially truncated pages in filter_setattr,
4757 this way we completely avoid having any pages in page cache on OST
4758 and can retire ugly workarounds during writes to flush such pages.
4763 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4764 Details : protect inode from truncation within vfs_unlink() context
4765 just take a reference before calling vfs_unlink() and release it
4766 when parent's i_sem is free.
4771 Description: last_id value checked outside lock on OST caused LASSERT failure
4772 Details : If there were multiple MDS->OST object precreate requests in
4773 flight, it was possible that the OST's last object id was checked
4774 outside a lock and incorrectly tripped an assertion. Move checks
4775 inside locks, and discard old precreate requests.
4778 Frequency : always, if extents are used on OSTs
4780 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4781 Details : OSTs using the extents format would not zero the high 16 bits of
4782 the index physical block number. This is not a problem for any
4783 OST filesystems smaller than 16TB, and no kernels support ext3
4784 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4785 new/modified files) and can be fixed for existing filesystems
4786 with e2fsprogs-1.39-cfs1.
4791 Description: import connection selection may be incorrect if timer wraps
4792 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4793 import connections to be ignored if the 32-bit jiffies counter
4794 wraps. Use a 64-bit jiffies counter.
4797 Frequency : very large clusters immediately after boot
4799 Description: LNET request buffers exhausted under heavy short-term load
4800 Details : If a large number of client requests are generated on a service
4801 that has previously never seen so many requests it is possible
4802 that the request buffer growth cannot keep up with the spike in
4803 demand. Instead of dropping incoming requests, they are held in
4804 the LND until the RPC service can accept more requests.
4807 Frequency : Sometimes during replay
4809 Description: Assertion failure in ll_local_open after replay.
4810 Details : If replay happened on an open request reply before we were able
4811 to set replay handler, reply will become not swabbed tripping the
4812 assertion in ll_local_open. Now we set the handler right after
4813 recognising of open request
4816 Frequency : very rare
4818 Description: kernel reports "badness in vsnprintf"
4819 Details : Reading from the "recovery_status" /proc file in small chunks
4820 may cause a negative length in lprocfs_obd_rd_recovery_status()
4821 call to vsnprintf() (which is otherwise harmless). Exit early
4822 if there is no more space in the output buffer.
4824 Severity : enhancement
4826 Description: clear OBD RPC statistics by writing to them
4827 Details : It is now possible to clear the OBD RPC statistics by writing
4828 to the "stats" file.
4833 Description: Client mtime is not the same on different clients after utimes
4834 Details : In some cases, the client was using the utimes() syscall on
4835 a file cached on another node. The clients now validate the
4836 ctime from the MDS + OSTs to determine which one is right.
4841 Description: Inability to activate failout mode
4842 Details : lconf script incorrectly assumed that in pythong string's numeric
4843 value is used in comparisons.
4846 Frequency : always with multiple stripes per file
4848 Description: Inefficient object allocation for mutli-stripe files
4849 Details : When selecting which OSTs to stripe files over, for files with
4850 a stripe count that divides evenly into the number of OSTs,
4851 the MDS is always picking the same starting OST for each file.
4852 Return the OST selection heuristic to the original design.
4857 Description: mount failures may take full timeout to return an error
4858 Details : Under some heavy load conditions it is possible that a
4859 failed mount can wait for the full obd_timeout interval,
4860 possibly several minutes, before reporting an error.
4861 Instead return an error as soon as the status is known.
4863 Frequency : quota enabled and large files being deleted
4865 Description: releasing more than 4GB of quota at once hangs OST
4866 Details : If a user deletes more than 4GB of files on a single OST it
4867 will cause the OST to spin in an infinite loop. Release
4868 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4873 Description: statfs data retrieved from /proc may be stale or zero
4874 Details : When reading per-device statfs data from /proc, in the
4875 {kbytes,files}_{total,free,avail} files, it may appear
4876 as zero or be out of date.
4879 Frequency : systems with MD RAID1 external journal devices
4881 Description: lconf's call to blkid is confused by RAID1 journal devices
4882 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4883 of returning all block devices that match the journal UUID.
4886 Frequency : always, for aggregate stripe size over 4GB
4888 Description: assertion fails when trying to use 4GB stripe size
4889 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4890 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4893 Frequency : always on ppc64
4895 Description: the first write on an ext3 filesystem with mballoc got stuck
4896 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4897 perform endianness conversion.
4899 ------------------------------------------------------------------------------
4901 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4903 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4904 INTEROPERATE with older versions automatically. Please read the
4905 user documentation before upgrading any part of a live system.
4906 * WARNING: Lustre networking configuration changes are required with
4907 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4910 * Support for kernels:
4911 2.6.9-22.0.2.EL (RHEL 4)
4912 2.6.5-7.244 (SLES 9)
4913 2.6.12.6 vanilla (kernel.org)
4916 Severity : enhancement
4917 Bugzilla : 7981/8208
4918 Description: Introduced Lustre Networking (LNET)
4919 Details : LNET is new networking infrastructure for Lustre, it includes
4920 a reorganized network configuration mode (see the user
4921 documentation for full details) as well as support for routing
4922 between different network fabrics. Lustre Networking Devices
4923 (LNDs) for the supported network fabrics have also been
4924 created for this new infrastructure.
4926 Severity : enhancement
4927 Description: Introduced Access control lists
4928 Details : clients can set ACLs on files and directories in order to have
4929 more fine-grained permissions than the standard Unix UGO+RWX.
4930 The MDS must be started with the "-o acl" mount option.
4932 Severity : enhancement
4933 Description: Introduced filesystem quotas
4934 Details : Administrators may now establish per-user quotas on the
4937 Severity : enhancement
4939 Description: Configuration change for the XT3
4940 The PTLLND is now used to run Lustre over Portals on the XT3
4941 The configure option(s) --with-cray-portals are no longer used.
4942 Rather --with-portals=<path-to-portals-includes> is used to
4943 enable building on the XT3. In addition to enable XT3 specific
4944 features the option --enable-cray-xt3 must be used.
4949 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4950 Details : A race between allocating a new llcd and re-getting the llcd_lock
4951 allowed another thread to grab newly-allocated llcd.
4953 Severity : enhancement
4955 Description: 2.6 OST async journal commit and locking fix to improve performance
4956 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4957 2.6 kernels are now async as they already were in 2.4 kernels so
4958 that they can commit concurrently with the network bulk transfer.
4959 For block-allocated files the filter allocation semaphore is held
4960 to avoid filesystem fragmentation during allocation. BKL lock
4961 removed for 2.6 xattr operations where it is no longer needed.
4966 Description: lconf incorrectly determined whether two IP networks could talk
4967 Details : In some more complicated routing and multiple-network
4968 configurations, lconf will avoid trying to make a network
4969 connection to a disjoint part of the IP space. It was doing the
4970 math incorrectly for one set of cases.
4975 Description: Fix for potential infinite loop processing records in an llog.
4976 Details : If an llog record is corrupted/zeroed, it is possible to loop
4977 forever in llog_process(). Validate the llog record length
4978 and skip the remainder of the block on error.
4981 Frequency : occasional (liblustre only)
4983 Description: liblustre could not open files whose last component is a symlink
4984 Details : sysio_path_walk() would incorrectly pass the open intent to
4985 intermediate path components.
4988 Frequency : rare (liblustre only with non-standard tuning)
4989 Bugzilla : 7201 (7350)
4990 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4991 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4992 after it has been released, only do this for referenced locks
4994 Severity : enhancement
4996 Description: specifying an (invalid) directory default stripe_size of -1
4997 would reset the directory default striping
4998 Details : stripe_size -1 was used internally to signal directory stripe
4999 removal, now use "all default" to signal dir stripe removal
5000 as a directory striping of "all default" is not useful
5003 Frequency : common for large clusters running liblustre clients
5005 Description: doing an ls when liblustre clients are running is slow
5006 Details : sending a glimpse AST to a liblustre client waits for every AST
5007 to time out, as liblustre clients will not respond. Since they
5008 cannot cache data we refresh the OST lock LVB from disk instead.
5010 Severity : enhancement
5012 Description: doing an ls at the same time as file IO can be slow
5013 Details : enqueue and other "small" requests can be blocked behind many
5014 large IO requests. Create a new OST IO portal for non-IO
5015 requests so they can be processed faster.
5018 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
5020 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
5021 created file with mode 000
5022 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
5023 allow an ftruncate() as a non-root user to a file with mode 000.
5024 Lustre now respects this flag to disable mode checks when
5025 truncating a file owned by the user
5028 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
5031 Description: Revoking locks from clients that went dead or catatonic might take
5033 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
5034 cancellation of such locks instant on servers without waiting for
5035 any reply from clients. Clients drops these locks when cancel
5036 notification from server is received without replying.
5039 Frequency : liblustre-only, when liblustre client dies or becomes busy
5041 Description: Doing ls on Linux clients can take a long time with active
5043 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
5044 granting such locks to it in the first place if possible. Locks
5045 are taken by proxy on the OST during the read or write and
5046 dropped immediately afterward. Add connect flags handling, do
5047 not grant locks to liblustre clients for glimpse ASTs.
5049 Severity : enhancement
5051 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
5052 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5053 sequential reads and behaves badly with applications doing random
5054 reads. Improve it by reading ahead at least read region, and
5055 avoiding excessive large RPC for small reads.
5057 Severity : enhancement
5059 Description: Creating more than 1000 files for a single job may cause a load
5060 imbalance on the OSTs if there are also a large number of OSTs.
5061 Details : qos_prep_create() uses an OST index reseed value that is an
5062 even multiple of the number of available OSTs so that if the
5063 reseed happens in the middle of the object allocation it will
5064 still utilize the OSTs as uniformly as possible.
5069 Description: OST or MDS may oops in ping_evictor_main()
5070 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5071 but doesn't restart at beginning of obd_exports_timed list
5074 Severity : enhancement
5076 Description: improve by-nid export eviction on the MDS and OST
5077 Details : allow multiple exports with the same NID to be evicted at one
5078 time without re-searching the exports list.
5081 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5083 Description: MDS may oops in groups_free()
5084 Details : in rare race conditions a newly allocated group_info struct is
5085 freed again, and this can be NULL. The 2.4 compatibility code
5086 for groups_free() checked for a NULL pointer, but 2.6 did not.
5089 Frequency : common for liblustre clients doing little filesystem IO
5090 Bugzilla : 9352, 7313
5091 Description: server may evict liblustre clients accessing contended locks
5092 Details : if a client is granted a lock or receives a completion AST
5093 with a blocking AST already set it would not reply to the AST
5094 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5097 Frequency : lfs setstripe, only systems with more than 160 OSTs
5099 Description: unable to set striping with a starting offset beyond OST 160
5100 Details : llapi_create_file() incorrectly limited the starting stripe
5101 index to the maximum single-file stripe count.
5104 Frequency : LDAP users only
5106 Description: lconf did not handle in-kernel recovery with LDAP properly
5107 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5109 Severity : enhancement
5111 Description: bind OST threads to NUMA nodes to improve performance
5112 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5113 node and do their allocations there to localize memory access
5115 Severity : enhancement
5117 Description: llmount can determine client NID directly from Myrinet (GM)
5118 Details : the client NID code from gmnalnid was moved directly into
5119 llmount, removing the need to use this or specifying the
5120 client NID explicitly when mounting GM clients with zeroconf
5123 Frequency : if client is started with down MDS
5125 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5126 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5127 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5128 ptlrpc_import_delayed_req() because we didn't check if the
5129 request was interrupted, and we also didn't break out of the
5130 event loop if there was no timeout
5135 Description: data loss during non-page-aligned writes to a single file from
5136 both multiple nodes and multiple threads on one node at same time
5137 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5138 inconsistency led to false short-reads, that were cached and later
5139 used by ->prepare_write() to fill in partially written page,
5140 leading to data loss.
5143 Frequency : always, if lconf --abort_recovery used
5145 Description: lconf --abort_recovery fails with 'Operation not supported'
5146 Details : lconf was attempting to abort recovery on the MDT device and not
5149 Severity : enhancement
5151 Description: remove cleanup logs
5152 Details : replace lconf-generated cleanup logs with lustre internal
5153 cleanup routines. Eliminates the need for client-cleanup and
5156 Severity : enhancement
5158 Description: add support for EAs (user and system) on lustre filesystems
5159 Details : it is now possible to store extended attributes in the Lustre
5160 client filesystem, and with the user_xattr mount option it
5161 is possible to allow users to store EAs on their files also
5163 Severity : enhancement
5165 Description: Add possibility (config option) to show minimal available OST free
5167 Details : When compiled with --enable-mindf configure option, statfs(2)
5168 (and so, df) will return least minimal free space available from
5169 all OSTs as amount of free space on FS, instead of summary of
5170 free spaces of all OSTs.
5172 Severity : enhancement
5174 Description: do not expand extent locks acquired on OST-side
5175 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5176 by server: they are not cached anyway.
5179 Frequency : when mmap is used/binaries executed from Lustre
5181 Description: Unmmap pages before throwing them away from read cache.
5182 Details : llap_shrink cache now attempts to unmap pages before discarding
5183 them (if unmapping failed - do not discard). SLES9 kernel has
5184 extra checks that trigger if this unmapping is not done first.
5189 Description: lconf didn't resolve symlinks before checking to see whether a
5190 given mountpoint was already in use
5193 Frequency : when migrating failover services
5194 Bugzilla : 6395, 9514
5195 Description: When migrating a subset of services from a node (e.g. failback
5196 from a failover service node) the remaining services would
5197 time out and evict clients.
5198 Details : lconf --force (implied by --failover) sets the global obd_timeout
5199 to 5 seconds in order to quickly disconnect, but this caused
5200 other RPCs to time out too quickly. Do not change the global
5201 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5203 Severity : enhancement
5204 Frequency : if MDS is started with down OST
5205 Bugzilla : 9439,5706
5206 Description: Allow startup/shutdown of an MDS without depending on the
5207 availability of the OSTs.
5208 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5209 Add appropriate locking and lov-osc refcounts for safe
5210 cleaning. Add osc abort_inflight calls in case the
5211 synchronize never started.
5214 Frequency : occasional (Cray XT3 only)
5216 Description: root not authorized to access files in CRAY_PORTALS environment
5217 Details : The client process capabilities were not honoured on the MDS in
5218 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5219 been accessed by an authorized user then root was able to access
5220 the file on the local client also. The root user capabilities
5221 are now allowed on the MDS, as this environment has secure UID.
5224 Frequency : occasional
5226 Description: ldiskfs "too long searching" message happens too often
5227 Details : A debugging message (otherwise harmless) prints too often on
5228 the OST console. This has been reduced to only happen when
5229 there are fragmentation problems on the filesystem.
5234 Description: Division by zero in statfs when all OSCs are inactive
5235 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5236 lov_statfs divides by value returned by lov_get_stripecnt().
5240 Bugzilla : 9489, 3273
5241 Description: First write from each client to each OST was only 4kB in size,
5242 to initialize client writeback cache, which caused sub-optimal
5243 RPCs and poor layout on disk for the first written file.
5244 Details : Clients now request an initial cache grant at (re)connect time
5245 and so that they can start streaming writes to the cache right
5246 away and always do full-sized RPCs if there is enough data.
5247 If the OST is rebooted the client also re-establishes its grant
5248 so that client cached writes will be honoured under the grant.
5253 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5254 Details : Now I/O RPCs go to different portal number and (presumably) fast
5255 lock requests (and glimses) and other RPCs get their own service
5256 threads pool that should be able to service those RPCs
5259 Severity : enhancement
5261 Description: Ability to exchange lustre version between client and servers and
5262 issue warnings at client side if client is too old. Also for
5263 liblustre clients there is ability to refuse connection of too old
5265 Details : New 'version' field is added to connect data structure that is
5266 filled with version info. That info is later checked by server and
5270 Frequency : rare, liblustre only.
5271 Bugzilla : 9296, 9581
5272 Description: Two simultaneous writes from liblustre at offset within same page
5273 might proceed at the same time overwriting eachother with stale
5275 Details : I/O lock withing llu_file_prwv was released too early, before data
5276 actually was hitting the wire. Extended lock-holding time until
5277 server acknowledges receiving data.
5280 Frequency : extremely rare. Never observed in practice.
5282 Description: avoid generating lustre_handle cookie of 0.
5283 Details : class_handle_hash() generates handle cookies by incrementing
5284 global counter, and can hit 0 occasionaly (this is unlikely, but
5285 not impossible, because initial value of cookie counter is
5286 selected randonly). Value of 0 is used as a sentinel meaning
5287 "unassigned handle" --- avoid it. Also coalesce two critical
5288 sections in this function into one.
5290 Severity : enhancement
5292 Description: allow liblustre clients to delegate truncate locking to OST
5293 Details : To avoid overhead of locking, liblustre client instructs OST to
5294 take extent lock in ost_punch() on client's behalf. New connection
5295 flag is added to handle backward compatibility.
5297 Severity : enhancement
5298 Bugzilla : 4928, 7341, 9758
5299 Description: allow number of OST service threads to be specified
5300 Details : a module parameter allows the number of OST service threads
5301 to be specified via "options ost ost_num_threads={N}" in the
5302 OSS's /etc/modules.conf or /etc/modprobe.conf.
5306 Bugzilla : 6146, 9635, 9895
5307 Description: servers crash with bad pointer in target_handle_connect()
5308 Details : In rare cases when a client is reconnecting it was possible that
5309 the connection request was the last reference for that export.
5310 We would temporarily drop the export reference and get a new
5311 one, but this may have been the last reference and the export
5312 was just destroyed. Get new reference before dropping old one.
5314 Severity : enhancement
5315 Frequency : if client is started with failover MDS
5317 Description: Allow multiple MDS hostnames in the mount command
5318 Details : Try to read the configuration from all specified MDS
5319 hostnames during a client mount in case the "primary"
5322 Severity : enhancement
5324 Description: Stop sending data to evicted clients as soon as possible.
5325 Details : Check if the client we are about to send or are sending data to
5326 was evicted already. (Check is done every second of waiting,
5327 for which l_wait_event interface was extended to allow checking
5328 of exit condition at specified intervals).
5331 Frequency : rare, normally only when NFS exporting is done from client
5333 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5335 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5338 Frequency : always, if srand() is called before liblustre initialization
5340 Description: Liblustre uses system PRNG disturbing its usage by user application
5341 Details : Introduce internal to lustre fast and high-quality PRNG for
5342 lustre usage and make liblustre and some other places in generic
5343 lustre code to use it.
5345 Severity : enhancement
5346 Bugzilla : 9477, 9557, 9870
5347 Description: Verify that the MDS configuration logs are updated when xml is
5348 Details : Check if the .xml configuration logs are newer than the config
5349 logs stored on the MDS and report an error if this is the case.
5350 Request --write-conf, or allow starting with --old_conf.
5352 Severity : enhancement
5354 Description: Handle symlinks in the path when checking if Lustre is mounted.
5355 Details : Resolve intermediate symlinks when checking if a client has
5356 mounted a filesystem to avoid duplicate client mounts.
5361 Description: lconf can hit an error exception but still return success.
5362 Details : The lconf command catches the Command error exception at the top
5363 level script context and will exit with the associated exit
5364 status, but doesn't ensure that this exit status is non-zero.
5369 Description: failure of ptlrpc thread startup can cause oops
5370 Details : Starting a ptlrpc service thread can fail if there are a large
5371 number of threads or the server memory is very fragmented.
5372 Handle this without oopsing.
5375 Frequency : always, only if liblustre and non-default acceptor port was used
5377 Description: liblustre cannot connect to servers with non-default acceptor port
5378 Details : tcpnal_set_default_params() was not called and was therefore
5379 ignoring the environment varaible TCPNAL_PORT, as well as other
5380 TCPNAL_ environment variables
5385 Description: two objects could be created on the same OST for a single file
5386 Details : If an OST is down, in some cases it was possible to create two
5387 objects on a single OST for a single file. No problems other
5388 than potential performance impact and spurious error messages.
5392 Bugzilla : 5681, 9562
5393 Description: Client may oops in ll_unhash_aliases
5394 Details : Client dcache may become inconsistent in race condition.
5395 In some cases "getcwd" can fail if the current directory is
5401 Description: Inode refcounting problems in NFS export code
5402 Details : link_raw functions used to call d_instantiate without obtaining
5403 extra inode reference first.
5407 Bugzilla : 9942, 9903
5408 Description: Referencing freed requests leading to crash, memleaks with NFS.
5409 Details : We used to require that call to ll_revalidate_it was always
5410 followed by ll_lookup_it. Also with revalidate_special() it is
5411 possible to call ll_revalidate_it() twice for the same dentry
5412 even if first occurence returned success. This fix changes semantic
5413 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5414 reference on a request referred from the intent.
5415 ll_intent_release() then releases such a request.
5418 Frequency : rare, normally benchmark loads only
5420 Description: unlinked inodes were kept in memory on the client
5421 Details : If a client is repeatedly creating and unlinking files it
5422 can accumulate a lot of stale inodes in the inode slab cache.
5423 If there is no other client load running this can cause the
5424 client node to run out of memory. Instead flush old inodes
5425 from client cache that have the same inode number as a new inode.
5428 Frequency : SLES9 2.6.5 kernel and long filenames only
5429 Bugzilla : 9969, 10379
5430 Description: utime reports stale NFS file handle
5431 Details : SLES9 uses out-of-dentry names in some cases, which confused
5432 the lustre dentry revalidation. Change it to always use the
5436 Frequency : rare, unless heavy write-truncate concurrency is continuous
5437 Bugzilla : 4180, 6984, 7171, 9963, 9331
5438 Description: OST becomes very slow and/or deadlocked during object unlink
5439 Details : filter_destroy() was holding onto the parent directory lock
5440 while truncating+unlinking objects. For very large objects this
5441 may block other threads for a long time and slow overall OST
5442 responsiveness. It may also be possible to get a lock ordering
5443 deadlock in this case, or run out of journal credits because of
5444 the combined truncate+unlink. Solution is to do object truncate
5445 first in one transaction without parent lock, and then do the
5446 final unlink in a new transaction with the parent lock. This
5447 reduces the lock hold time dramatically.
5450 Frequency : rare, 2.4 kernels only
5452 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5453 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5454 before unmounting their filesystem in order to avoid deadlock.
5455 The kernel_locked() function in 2.4 kernels only checks whether
5456 the kernel lock is held, not whether it is this process that is
5457 holding it as 2.6 kernels do.
5462 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5463 Details : The client ptlrpc code may be trying to reconnect to a down
5464 server before a previous connection attempt has timed out.
5465 Increase the reconnect interval to be longer than the connection
5466 timeout interval to avoid sending duplicate connections to
5470 Frequency : echo_client brw_test command
5472 Description: fix echo_client to work with OST preallocated code
5473 Details : OST preallocation code (5137) didn't take echo_client IO path
5474 into account: echo_client calls filter methods outside of any
5475 OST thread and, hence, there is no per-thread preallocated
5476 pages and buffers to use. Solution: hijack pga pages for IO. As
5477 a byproduct, this avoids unnecessary data copying.
5481 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5482 Description: Client can oops in mdc_commit_close() after open replay
5483 Details : It was possible for the MDS to return an open request with no
5484 transaction number in mds_finish_transno() if the client was
5485 evicted, but without actually returning an error. Clients
5486 would later try to replay that open and may trip an assertion
5487 Simplify the client close codepath, and always return an error
5488 from the MDS in case the open is not successful.
5491 Frequency : rare, 2.6 OSTs only
5493 Description: OST may deadlock under high load on fragmented files
5494 Details : If there was a heavy load and highly-fragmented OST filesystems
5495 it was possible to have all the OST threads deadlock waiting on
5496 allocation of biovecs, because the biovecs were not released
5497 until the entire RPC IO was completed. Instead, release biovecs
5498 as soon as they are complete to ensure forward IO progress.
5500 Severity : enhancement
5502 Description: Support for specifying external journal device at mount
5503 Details : If an OST or MDS device is formatted with an external journal
5504 device, this device major/minor is stored in the ext3 superblock
5505 and may not be valid for failover. Allow detecting and
5506 specifying the external journal at mount time.
5511 Description: Mounting an MDS with pending unlinked files may cause oops
5512 Details : target_finish_recovery() calls mds_postrecov() which returned
5513 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5514 considers this an error and immediately begins cleaning up the
5515 lov, just after starting the mds_lov process
5517 Severity : enhancement
5519 Description: Implement 'lfs df' to report actual free space on per-OST basis
5520 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5521 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5522 usage of objects; '-h' to report in human readable format.
5524 ------------------------------------------------------------------------------
5526 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5533 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5534 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5535 at startup during orphan unlinking. Instead, skip buddy update
5536 before setup, as it will be regenerated after recovery is complete.
5541 Description: If an OST is inactive, its locks might reference stale inodes.
5542 Details : lov_change_cbdata() must iterate over all namespaces, even if
5543 they are inactive to clear inode references from the lock.
5545 Severity : enhancement
5546 Frequency : occasional, if non-standard max_dirty_mb used
5548 Description: Client will block write RPCs if not enough grant
5549 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5550 then the client will block writes while waiting for another RPC
5551 to complete instead of consuming its dirty limit. With change
5552 we get improved performance when max_dirty_mb is small.
5554 Severity : enhancement
5555 Bugzilla : 3389, 6253
5556 Description: Add support for supplementary groups on the MDS.
5557 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5558 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5559 lookups for user supplementary groups into a cache.
5563 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5564 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5565 for liblustre O_EXCL is enforced by MDS.
5569 Description: readdir never returns NULL in liblustre.
5570 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5571 dentry in struct dirent.
5575 Frequency : liblustre only, depends on application IO pattern
5576 Description: liblustre clients evicted if not contacting servers
5577 Details : Don't put liblustre clients into the ping_evictor list, so
5578 they will not be evicted by the pinger ever.
5580 Severity : enhancement
5582 Description: Add ability to evict clients by NID from MDS.
5583 Details : By echoing "nid:$NID" string into
5584 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5585 $NID would be instantly evicted from this MDS and from all active
5586 OSTs connected to it.
5590 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5591 Details : lookup_it_finish() used to query file size from OSTs that was not
5596 Description: service threads change working directory to that of init
5597 Details : Starting lustre service threads may pin the working directory
5598 of the parent thread, making that filesystem busy. Threads
5599 now change to the working directory of init to avoid this.
5603 Frequency : during shutdown only
5604 Description: shutdown with a failed MDS or OST can cause unmount to hang
5605 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5610 Frequency : 2.6 only
5611 Description: chmod/chown may include an extra supplementary group
5612 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5613 supplementary group and if none is specified this is used.
5616 Bugzilla : 5479 (6816)
5618 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5619 Details : If lookup is in progress on a file that is unlinked we might try
5620 to revalidate the inode and fail in revalidate after lookup is
5621 complete and ll_file_open() enqueues the open again but
5622 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5625 Frequency : always, if lconf --abort_recovery used
5627 Description: lconf --abort_recovery fails with 'Operation not supported'
5628 Details : lconf was attempting to abort recovery on the MDT device and not
5631 ------------------------------------------------------------------------------
5633 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5638 Frequency : rare (only unsupported configurations with a node running as an
5640 Bugzilla : 6514, 5137
5641 Description: Mounting a Lustre file system on a node running as an OST could
5643 Details : OSTs now preallocates memory needed to write out data at
5644 startup, instead of when needed, to avoid having to
5645 allocate memory in possibly low memory situations.
5646 Specifically, if the file system is mounted on on OST,
5647 memory pressure could force it to try to write out data,
5648 which it needed to allocate memory to do. Due to the low
5649 memory, it would be unable to do so and the node would
5650 become unresponsive.
5652 Severity : enhancement
5654 Description: Addition of lconf --service command line option
5655 Details : lconf now accepts a '--service <arg>' option, which is
5656 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5658 Severity : enhancement
5660 Description: Failover mode is now the default for OSTs.
5661 Details : By default, OSTs will now run in failover mode. To return to
5662 the old behaviour, add '--failout' to the lmc line for OSTs.
5664 Severity : enhancement
5666 Description: Health checks are now provided for MDS and OSTs
5667 Details : Additional detailed health check information on MSD and OSTs
5668 is now provided through the procfs health_check value.
5671 Frequency : occasional, depends on IO load
5673 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5674 after numerous create/delete cycles
5675 Details : The ext3 inode allocation policy would not allocate new inodes
5676 very well on the OSTs because there are no new directories
5677 being created. Instead we look for groups with free space if
5678 the parent directories are nearly full.
5683 Description: Network or server problems during mount may cause partially
5684 mounted clients instead of returning an error.
5685 Details : The config llog parsing code may overwrite the error return
5686 code during mount error handling, returning success instead
5692 Description: MDS can fail to allocate large reply buffers
5693 Details : After long uptimes the MDS can fail to allocate large reply
5694 buffers (e.g. zconf client mount config records) due to memory
5695 fragmentation or consumption by the buffer cache. Preallocate
5696 some large reply buffers so that these replies can be sent even
5697 under memory pressure.
5701 Frequency : rare (liblustre)
5702 Description: fsx running with liblustre complained that using truncate() to
5703 extend the file doesn't work. This patch corrects that issue.
5704 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5705 ATTR_SIZE and lsm use in llu_setattr_raw.
5709 Frequency : rare, only 2.6 kernels
5710 Description: Unusual file access patterns on the MDS may result in inode
5711 data being lost in very rare circumstances.
5712 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5713 code in the 2.6 kernel caused the inode and/or EA data not to
5714 be read from disk, causing single-file corruption.
5718 Frequency : rare, only 2.6 filesystems using extents
5719 Description: Heavy concurrent write and delete load may cause data corruption.
5720 Details : It was possible under high-load situations to have an extent
5721 metadata block in the block device cache from a just-unlinked
5722 file overwrite a newly-allocated data block. We now unmap any
5723 metadata buffers that alias just-allocated data blocks.
5727 Frequency : filesystems with default stripe_count larger than 77
5728 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5729 Details : lconf specifies an inode size of 4096 bytes when the default
5730 stripe_count is larger than 77. This conflicts with the default
5731 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5732 this case to avoid pinning too much memory for large EAs.
5734 ------------------------------------------------------------------------------
5736 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5741 Frequency : rare (extremely heavy IO load with hundreds of clients)
5743 Description: Client is evicted, gets IO error writing to file
5744 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5745 caused clients to be evicted for AST timeouts. The fixes in
5746 bug 5192 mean we no longer need to have such short AST timeouts
5747 so ldlm_timeout has been increased.
5750 Frequency : occasional during --force or --failover shutdown under load
5751 Bugzilla : 5949, 4834
5752 Description: Server oops/LBUG if stopped with --force or --failover under load
5753 Details : a collection of import/export refcount and cleanup ordering
5754 issues fixed for safer force cleanup
5757 Frequency : only filesystems larger than 120 OSTs
5758 Bugzilla : 5990, 6223
5759 Description: lfs getstripe would oops on a very large filesystem
5760 Details : lov_getconfig used kfree on vmalloc'd memory
5763 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5764 Bugzilla : 6242, 6243
5765 Description: reading from files that had been truncated to a non-zero size
5766 but never opened returned no data
5767 Details : ll_file_read() reads zeros from no-object files to EOF
5772 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5773 Details : There's little harm in aborting MDS/OSS recovery and letting it
5774 try again, so I removed the LASSERT and return an error instead.
5776 Severity : enhancement
5778 Description: New debugging infrastructure for tracking down data corruption
5779 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5780 (b) cover more of the client-side code path, and (c) try to narrow
5781 down where problems occurred
5785 Bugzilla : 3819, 4364, 4397, 6313
5786 Description: Racing close and eviction MDS could cause assertion in mds_close
5787 Details : It was possible to get multiple mfd references during close and
5788 client eviction, leading to one thread referencing a freed mfd.
5790 Severity: : enhancement
5791 Bugzilla : 3262, 6359
5792 Description: Attempts to reconnect to servers are now more aggressive.
5793 Details : This builds on the enhanced upcall-less recovery that was added
5794 in 1.4.2. When trying to reconnect to servers, clients will
5795 now try each server in the failover group every 10 seconds. By
5796 default, clients would previously try one server every 25 seconds.
5801 Description: After recovery, certain operations trigger a failed
5802 assertion on a client.
5803 Details : Failing over an mds, using lconf -d --failover, while a
5804 client was doing a readdir() call would cause the client to
5805 LBUG after recovery completed and the readdir() was resent.
5807 Severity : enhancement
5809 Description: Default groups are now added by lconf
5810 Details : You can now run lconf --group <servicename> without having to
5811 manually add groups with lmc.
5814 Frequency : occasional
5816 Description: Nodes with an elan id of 0 trigger a failed assertion
5819 Frequency : always when accessing e.g. tty/console device nodes
5821 Description: tty and some other devices nodes cannot be used on lustre
5822 Details : file's private_data field is used by device data and lustre
5823 values in there got lost. New field was added to struct file to
5824 store fs-specific private data.
5827 Frequency : when exporting Lustre via NFS
5829 Description: NFSD failed occasionally when looking up a path component
5830 Details : NFSD is looking up ".." which was broken in ext3 directories
5831 that had grown large enough to become hashed.
5834 Frequency : Clusters with multiple interfaces not on the same subnet
5836 Description: Nodes will repeatedly try to reconnect to an interface which it
5837 cannot reach and report an error to the log.
5838 Details : Extra peer list entries will be created by lconf with some peers
5839 unreachable. lconf now validates the peer before adding it.
5842 Frequency : Only if a default stripe is set on the filesystem root.
5844 Description: Setting a default stripe on the filesystem root prevented the
5845 filesystem from being remounted.
5846 Details : The client was sending extra request flags in the root getattr
5847 request and did not allocate a reply buffer for the dir EA.
5850 Frequency : occasional, higher if lots of files are accessed by one client
5851 Bugzilla : 6159, 6097
5852 Description: Client trips assertion regarding lsm mismatch/magic
5853 Details : While revalidating inodes the VFS looks up inodes with ifind()
5854 and in rare cases can find an inode that is being freed.
5855 The ll_test_inode() code will free the lsm during ifind()
5856 when it finds an existing inode and then the VFS later attaches
5857 this free lsm to a new inode.
5861 Bugzilla : 6422, 7030
5862 Description: MDS deadlock between mkdir and client eviction
5863 Details : Creating a new file via mkdir or mknod (starting a transaction
5864 and getting the ns lock) can deadlock with client eviction
5865 (gets ns lock and trying to finish a synchronous transaction).
5868 Frequency : occasional
5869 Description: While starting a server, the fsfilt_ext3 module could not be
5871 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5872 kernels. Previously, lconf would still use the ext3 name
5873 when trying to load modules. Now, it will correctly use
5874 ext3 on 2.4 and ldiskfs on 2.6.
5876 Severity : enhancement
5877 Description: The default stripe count has been changed to 1
5878 Details : The interpretation of the default stripe count (0, to lfs
5879 or lmc) has been changed to mean striping across a single
5880 OST, rather than all available. For general usage we have
5881 found a stripe count of 1 or 2 works best.
5883 Severity : enhancement
5884 Description: Add support for compiling against Cray portals.
5885 Details : Conditional compiling for some areas that are different
5889 Frequency : occasional
5890 Bugzilla : 6409, 6834
5891 Description: Creating files with an explicit stripe count may lead to
5892 a failed assertion on the MDS
5893 Details : If some OSTs are full or unavailable, creating files may
5894 trigger a failed assertion on the MDS. Now, Lustre will
5895 try to use other servers or return an error to the
5899 Frequency : occasional
5901 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5902 caused lock timeout during readahead (since 1.4.2).
5903 Details : Processes doing readahead might match a lock that hasn't been
5904 granted yet if there are overlapping and conflicting lock
5905 requests. The readahead process waits on ungranted lock
5906 (original lock is CBPENDING), while OST waits for that process
5907 to cancel CBPENDING read lock and eventually evicts client.
5909 Severity : enhancement
5911 Description: Initial enabling of flock support for clients
5912 Details : Implements fcntl advisory locking and file status functions.
5913 This feature is provided as an optional mount flag (default
5914 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5915 locking are implemented yet, and those that are are not guaranteed
5916 to be completely correct in production environments.
5917 mount -t lustre -o [flock|noflock] ...
5920 Frequency : occasional
5922 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5923 assertion in the ext3 JBD (journaling) layer.
5924 Details : The b_committed_data struct is protected by the big kernel lock
5925 in 2.4 kernels, serializing journal_commit_transaction() and
5926 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5927 access to this struct. In 2.6 kernels there is finer grained
5928 locking to improve SMP performance of the JBD layer.
5932 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5937 Description: merge and cleanup kernel patches.
5940 -----------------------------------------------------------------------------
5942 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5944 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5945 versions of Lustre. Please update all servers and clients to
5946 version 1.4.2 or later at the same time. You must also run
5947 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5948 to update the configuration logs.
5950 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5951 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5952 - swab llog records of type '0' so we get proper header size/idx (5861)
5953 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5954 - fix rename of one directory over another leaking an inode (5953)
5955 - avoid SetPageDirty on 2.6 (5981)
5956 - don't re-add just-being-destroyed locks to the waiting list (5653)
5957 - when creating new directories, inherit the parent's custom
5958 striping settings if present parent (3048)
5959 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5960 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5961 - don't hold client locks on temporary worklist from l_lru (5666)
5962 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5963 - automatically evict dead clients (5921)
5964 - Update file size properly in create+truncate+fstat case (6196)
5965 - Do not unhash mountpoint dentries, do not allow removal of
5967 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5968 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5969 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5970 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5971 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5972 - Correctly handle reads of files with no objects (6243)
5973 - lctl recover will also mark a device active if deactivate used (5933)
5975 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5976 - allow --write-conf on an MDS with different nettype than client (5619)
5977 - don't write config llogs to MDS for mounts not from that MDS (5617)
5978 - lconf should create multiple TCP connections from a client (5201)
5979 - init scripts are now turned off by default; run chkconfig --on
5980 lustre and chkconfig --on lustrefs to use them
5981 - upcalls are no longer needed for clients to recover to failover
5983 - add --abort-recovery option to lconf to abort recovery on device
5985 - add support for an arbitrary number of OSTs (3026)
5986 - Quota support protocol changes.
5987 - forward compatibility changes to wire structs (6007)
5988 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5989 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5990 - improved "lustre status" script
5991 - initialize blocksize for non-regular files (6062)
5992 - added --disable-server and --disable-client configure options (5782)
5993 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5994 - Vanilla 2.4.29 support
5995 - increase maximum number of obd devices to 520 (6242)
5996 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5997 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5998 SLES 9 2.6.5 kernels
5999 - limit stripes per file to 160 (the maximum EA size) (6093)
6001 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6004 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
6005 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
6006 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
6007 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
6008 - small ext3 extents cleanups and fixes (5733)
6009 - improved mballoc code, several small races and bugs fixed (5733, 5638)
6010 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
6011 - avoid needless client->OST connect, fix handle mismatch (5317)
6012 - fix DLM error path that led to out-of-sync client, long delays (5779)
6013 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
6014 - fix several locking issues related to i_size (5492,5624,5654,5672)
6015 - don't move pending lock onto export if it is already evicted (5683)
6016 - fix kernel oops when creating .foo in unlinked directory (5548)
6017 - fix deadlock in obdfilter statistics vs. object create (5811)
6018 - use time_{before,after} to avoid timer jiffies wrap (5882)
6019 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
6020 - Do not leak request if server was not able to process it (5154)
6021 - If mds_open unable to find parent dir, make that negative lookup(5154)
6022 - don't create new directories with extent-mapping (5909, 5936)
6024 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
6025 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
6026 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
6027 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
6028 - support for RHEL4 kernel on i686 (5773)
6029 - provide error messages when incompatible logs are encountered (5898)
6031 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
6032 * version 1.4.0.10 (1.4.1 release candidate 1)
6034 - don't keep a lock reference when lock is not granted (4238)
6035 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6036 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
6037 - drop import inflight refcount on signal_completed_replay error (5255)
6038 - unlock page after async write error during send (3677)
6039 - handle missing objects in filter_preprw_read properly (5265)
6040 - no transno return for symlink open, don't save no-trasno open (3440)
6041 - don't try to complete elan receive that already failed (4012)
6042 - free RPC server reply state on error (5406)
6043 - clean up thread from ptlrpc_start_thread() on error (5160)
6044 - readahead could read extra page into cache that wasn't ejected (5388)
6045 - prevent races in class_attach/setup/cleanup/detach (5260)
6046 - don't dereference de->d_inode after l_dput of de (5458)
6047 - use "int" for stripe value returned from lock_to_stripe (5544)
6048 - mballoc allocation and error-checking fixes in 2.6 (5504)
6049 - block device patches to fix I/O request sizes in 2.6 (5482)
6050 - look up hostnames for IB nals (5602)
6051 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
6052 - don't start multiple acceptors for the same port (5277)
6053 - fix incorrect LASSERT in mds_getattr_name (5635)
6054 - export a proc file for general "ping" checking (5628)
6055 - fix "lfs check" to not block when the MDS is down (5628)
6057 - service request history (4965)
6058 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6059 - create an "evict_client" /proc entry on OSTs, like the MDS has
6060 - fix mount usage message, return errors per mount(8) (5168)
6061 - change grep [] to grep "[]" in tests so they work in more UMLs
6062 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6063 - remove ancient LOV_MAGIC_V0 EA support (5047)
6064 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6065 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6066 - allow readahead allocations to fail when low on memory (5383)
6067 - mmap locking landed again, after considerable improvement (2828)
6068 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6070 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6073 - send OST transaction number in read/write reply to free req (4966)
6074 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6075 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6076 - the watchdog thread now runs as interruptible (5246)
6077 - flock/lockf fixes (but it's still disabled, pending 5135)
6078 - don't use EXT3 constants in llite code (5094)
6079 - memory shortage at startup could cause assertion (5176)
6081 - reorganization of lov code
6082 - single portals codebase
6084 - add extents/mballoc support (5025)
6085 - direct I/O reads in the obdfilter (4048)
6086 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6088 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6091 - send OST transaction number in read/write reply to free req (4966)
6092 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6093 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6094 - the watchdog thread now runs as interruptible (5246)
6095 - handle missing objects in filter_preprw_read properly (5265)
6096 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6097 - drop import inflight refcount on signal_completed_replay error (5255)
6098 - unlock page after async write error during send (3677)
6099 - return original error code on reconstructed replies (3761)
6100 - no transno return for symlink open, don't save no-trasno open (3440)
6102 - add pid to ldlm debugging output (4922)
6103 - bump the watchdog timeouts -- we can't handle 30sec yet
6104 - extra debugging for orphan dentry/inode bug (5259)
6106 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6109 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6110 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6111 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6112 - fix NULL dereference in /proc/sys/portals/routes (4827)
6113 - allow failed mdc_close() operations to be interrupted (4561)
6114 - stop precreate on OST before MDS would time out on it (4778)
6115 - don't send partial-page writes before EOF from client (4410)
6116 - discard client grant for sub-page writes on large-page clients (4520)
6117 - don't free dentries not owned by NFS code, check generation (4806)
6118 - fix lsm leak if mds_create_objects() fails (4801)
6119 - limit debug_daemon file size, always print CERROR messages (4789)
6120 - use transno after validating reply (3892)
6121 - process timed out requests if import state changes (3754)
6122 - update mtime on OST during writes, return in glimpse (4829)
6123 - add mkfsoptions to LDAP (4679)
6124 - use ->max_readahead method instead of zapping global ra (5039)
6125 - don't interrupt __l_wait_event() during strace
6127 - add software watchdogs to catch hung threads quickly (4941)
6128 - make lustrefs init script start after nfs is mounted
6129 - fix CWARN/ERROR duplication (4930)
6130 - return async write errors to application if possible (2248)
6131 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6132 - print NAL number in %x format (4645)
6133 - update barely-supported suse-2.4.21-171 series (4842)
6134 - support for sles 9 %post scripts
6135 - support for building 2.6 kernel-source packages
6136 - support for sles km_* packages
6138 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6141 - ignore -ENOENT errors in osc_destroy (3639)
6142 - notify osc create thread that OSC is being cleaned up (4600)
6143 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6144 - reconstruct ost_handle() like mds_handle() (4657)
6145 - create a new thread to do import eviction to avoid deadlock (3969)
6146 - let lconf resolve symlinked-to devices (4629)
6147 - don't unlink "objects" from directory with default EA (4554)
6148 - hold socknal file ref over connect in case target is down (4394)
6149 - allow more than 32000 subdirectories in a single directory (3244)
6150 - fix blocks count for O_DIRECT writes (3751)
6151 - OST returns ENOSPC from object create when no space left (4539)
6152 - don't send truncate RPC if file size isn't changing (4410)
6153 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6154 - bind to privileged port in socknal and tcpnal (3689)
6156 - rate limit CERROR/CWARN console message to avoid overload (4519)
6157 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6158 - basic mmap support (3918)
6159 - kernel patch series update from b1_4 (4711)
6161 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6164 - avoid crash during MDS cleanup with OST shut down (2775)
6165 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6166 - don't use bad inodes on the MDS (3744)
6167 - dynamic object preallocation to improve recovery speed (4236)
6168 - don't hold spinlock over lock dumping or change debug flags (4401)
6169 - don't zero obd_dev when it is force cleaned (3651)
6170 - print grants to console if they go negative (4431)
6171 - "lctl deactivate" will stop automatic recovery attempts (3406)
6172 - look for existing locks in ldlm_handle_enqueue() (3764)
6173 - don't resolve lock handle twice in recovery avoiding race (4401)
6174 - revalidate should check working dir is a directory (4134)
6176 - don't always mark "slow" obdfilter messages as errors (4418)
6178 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6181 - don't close LustreDB during write_conf until it is done (3860)
6182 - fix typo in lconf for_each_profile (3821)
6183 - allow dumping logs from multiple threads at one time (3820)
6184 - don't allow multiple threads in OSC recovery (3812)
6185 - fix debug_size parameters (3864)
6186 - fix mds_postrecov to initialize import for llog ctxt (3121)
6187 - replace config semaphore with spinlock (3306)
6188 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6189 - don't allow enqueue to complete on a destroyed export (3822)
6190 - down write_lock before checking llog header bitmap (3825)
6191 - recover from lock replay timeout (3764)
6192 - up llog sem before sending rpc (3652)
6193 - reduce ns lock hold times when setting kms (3267)
6194 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6195 - fix NULL deref and obd_dev leak on setup error (3312)
6196 - replace some LBUG about llog ops with error handling (3841)
6197 - don't match INVALID dentries from d_lookup and spin (3784)
6198 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6199 - fix invalid assertion in ptlrpc_set_wait (3880)
6201 - add libwrap support for the TCP acceptor (3996)
6202 - add /proc/sys/portals/routes for non-root route listing (3994)
6203 - allow setting MDS UUID in .xml (2580)
6204 - print the stack of a process that LBUGs (4228)
6206 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6209 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6210 - make sure to unset replay flag from failed open requests (3440)
6211 - if default stripe count is 0, use OST count for inode size (3636)
6212 - update parent mtime/ctime on client for create/unlink (2611)
6213 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6214 - free recovery state on server during a forced cleanup (3571)
6215 - unregister_reply for resent reqs (3063)
6216 - loop back devices mounting and status check on 2.6 (3563)
6217 - fix resource-creation race that can provoke i_size == 0 (3513)
6218 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6219 - more debugging for page-accounting assertion (3746)
6220 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6221 - avoid deadlock after precreation failure (3758)
6222 - fix race and lock order deadlock in orphan handling (3450, 3750)
6223 - add validity checks when grabbing inodes from l_ast_data (3599)
6225 - add /proc/.../recovery_status to obdfilter (3428)
6226 - lightweight CDEBUG infrastructure, debug daemon (3668)
6227 - change default OSC RPC parameters to be better on small clusters
6228 - turn off OST read cache for files smaller than 32MB
6229 - install man pages and include them in rpms (3100)
6230 - add new init script for (un)mounting lustre filesystems (2593)
6231 - run chkconfig in %post for init scripts (3701)
6232 - drop scimac NAL (unmaintained)
6234 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6237 - clean kiobufs before and after use (3485)
6238 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6239 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6240 - change init script to fail more clearly if not run as root (1528)
6241 - allow clients to reconnect during replay (1742)
6242 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6243 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6244 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6245 - fix cond_resched (3554)
6246 - only evict unfinished clients after recovery (3515)
6247 - allow bulk resend, prevent data loss (3570)
6248 - dynamic ptlrpc request buffer allocation (2102)
6249 - don't allow unlinking open directory if it isn't empty (2904)
6250 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6251 - remove extraneous obd dereference causing LASSERT failure (3334)
6252 - don't use get_cycles() when creating temp. files on the mds (3156)
6253 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6254 - handle EEXIST for set-stripe, set proper directory name (3336)
6256 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6257 - fix ksocknal_fmb_callback() error messages (2918)
6259 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6262 - don't copy lvb into (possibly NULL) reply on error (2983)
6263 - don't deref dentry after dput, don't free lvb on error (2922)
6264 - use the kms to determine writeback rpc length (2947)
6265 - increment oti_logcookies when osc is inactive (2948)
6266 - update client's i_blocks count via lvb messages (2543)
6267 - handle intent open/close of special files properly (1557)
6268 - mount MDS with errors=remount-ro, like obdfilter (2009)
6269 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6270 - don't use cancelling-locks' kms values (2947)
6271 - use highest lock extent for kms, not last one (2925)
6272 - don't dereference ERR_PTR() dentry in error handling path (3107)
6273 - fix thread race in portals_debug_dumplog() (3122)
6274 - create lprocfs device entries at setup instead of at attach (1519)
6275 - common AST error handler, don't evict client on completion race (3145)
6276 - zero nameidata in detach_mnt in 2.6 (3118)
6277 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6278 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6279 - initialize RPC timeout timer earlier for 2.6 (3219)
6280 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6281 - print nal/nid for unknown nid (3258)
6282 - additional checks for oscc recovery before doing precreate (3284)
6283 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6284 - don't crash in mdc_close for bad permissions on open (3285)
6285 - zero i_rdev for non-device files (3147)
6286 - clear page->private before handing to FS, better assertion (3119)
6287 - tune the read pipeline (3236)
6288 - fix incorrect decref of invalidated dentry (2350)
6289 - provide read-ahead stats and refine rpc in flight stats (3328)
6290 - don't hold journal transaction open across create RPC (3313)
6291 - update atime on MDS at close time (3265)
6292 - close LDAP connection when recovering to avoid server load (3315)
6293 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6294 - don't leak open file on MDS after open resend (3325)
6295 - serialize filter_precreate and filter_destroy_precreated (3329)
6296 - loop device shouldn't call sync_dev() for nul device (3092)
6297 - clear page cache after eviction (2766)
6298 - resynchronize MDS->OST in background (2824)
6299 - refuse to mount the same filesystem twice on same mountpoint (3394)
6300 - allow llmount to create routes for mounting behind routers (3320)
6301 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6302 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6303 - fix rare problem with rename on htree directories (3417)
6305 - allow default OST striping configuration per directory (1414)
6306 - fix compilation for qswnal for 2.6 kernels (3125)
6307 - increase maximum number of MDS request buffers for large systems
6308 - change liblustreapi to be useful for external progs like lfsck (3098)
6309 - increase local configuration timeout for slow disks (3353)
6310 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6312 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6315 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6316 - don't overwrite extent policy data in reply if lock was blocked (2901)
6317 - drop filter export grants atomically with removal from device (2663)
6318 - del obd_self_export from work_list in class_disconnect_exports (2908)
6319 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6320 - swab reply message in mdc_close, other PPC fixes (2464)
6321 - fix destroying of named logs (2325)
6322 - overwrite old logs when running lconf --write_conf (2264)
6323 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6324 - fix race in target_handle_connect (2898)
6325 - mds_reint_create() should take same inode create lock (2926)
6326 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6327 - don't close files for self_export to avoid uninitialized obd (2936)
6328 - allow MDS with the same name as client node (2939)
6329 - hold dentry reference for closed log files for unlink (2325)
6330 - reserve space for all logs during transactions (2059)
6331 - don't evict page beyond end of stripe extent (2925)
6332 - don't oops on a deleted current working directory (2399)
6333 - handle hard links to targets without a parent properly (2517)
6334 - don't dereference NULL lock when racing during eviction (2867)
6335 - don't grow lock extents when lots of conflicting locks (2919)
6337 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6340 - account for cache space usage on clients to avoid data loss (974)
6341 - lfsck support in lustre kernel code (2349)
6342 - reduce journal credits needed for BRW writes (2370)
6343 - orphan handling to avoid losing space on client/server crashes
6344 - ptlrpcd can be blocked, stopping ALL progress (2477)
6345 - use lock value blocks to assist in proper KMS, faster stat (1021)
6346 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6347 - recovery for initial connections (2355)
6348 - fixes for mds_cleanup_orphans (1934)
6349 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6350 - block all file creations until orphan recovery completes (1901)
6351 - client remove rq_connection from request struct (2423)
6352 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6353 - recovery timer race (2670)
6354 - mdc_close recovey bug (2532)
6355 - ptlrpc cleanup bug (2710)
6356 - mds timeout on local locks (2588)
6357 - namespace lock held during RPCs (2431)
6358 - handle interrupted sync write properly (2503)
6359 - don't try to handle a message that hasn't been replied to (2699)
6360 - client assert failure during cleanup after abort recovery (2701)
6361 - leak mdc device after failed mount (2712)
6362 - ptlrpc_check_set allows timedout requests to complete (2714)
6363 - wait for inflight reqs when ptlrpcd finishes (2710)
6364 - make sure unregistered services are removed from the srv_list
6365 - reset bulk XID's when resending them (caught by 1138 test)
6366 - unregister_bulk after timeout
6367 - fix lconf error (2694)
6368 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6369 - readahead locks pages, leaves pending causing memory pressure (2673)
6370 - increase OST request buffers to 4096 on large machines (2729)
6371 - fix up permission of existing directories in simple_mkdir (2661)
6372 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6373 - don't assign transno to errored transactions (2742)
6374 - don't delete objects on OST if given a bogus objid from MDS (2751)
6375 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6376 - if rq_no_resend, then timeout request after recovery (2432)
6377 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6378 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6379 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6380 - MDS recovery shouldn't delete objects that it has given out (2730)
6381 - if enqueue arrives after completion, don't clobber LVB (2819)
6382 - don't unlock pages twice when trigger_group_io returns error (2814)
6383 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6384 - don't write pages to disk if there was an error (1450)
6385 - don't ping imports that have recovery disabled (2676)
6386 - take buffered bytes into account when balancing socknal conn (2817)
6387 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6388 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6389 - remove little-endian swabbing of llog records (1987)
6390 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6391 - retry reposting request buffers if they fail (1191)
6392 - grow extent at grant time to avoid granting a revoked lock (2809)
6393 - lock revoke doesn't evict page if covered by a second lock (2765)
6394 - disable VM readahead to avoid reading outside lock extents (2805)
6396 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6397 - updated kernel patches for hp-2.4.20 kernel (2681)
6399 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6402 - fix truncated write corruption (2366)
6403 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6405 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6406 - ensure len(uuid) < 37 in lmc (1171)
6407 - fix ia64 OOPS in llog_test (2255)
6408 - zero end of page at obdfilter for partial page writes (2648)
6409 - don't leave stale dentries around after renames (bug 2428)
6410 - fix timeouts when evicting a client with a single lock held (2642)
6411 - set deadline for the initial HELLO message to drain (2634)
6412 - print out dotted-quad IP addresses in the socknal (2302)
6414 - additional debugging for MDS client eviction problem (2443)
6415 - fix mkfsoptions support for osts (2603, 2604)
6417 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6420 - add series for the vanilla 2.6.0 kernel
6421 - add series for the vanilla 2.4.24 kernel
6422 - add series for a cray x86/64 UL kernel drop
6423 - fix xattr patches for the vanilla 2.4.19 series
6425 - generate true UUIDs in lmc (1171)
6426 - have portals stack dumping break in UML (2466)
6427 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6428 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6429 - wake callback waiting threads on client eviction (2460)
6430 - Add --ptldebug and --subsystem to lmc (1719)
6431 - update assertion to allow safe interrupt allocation
6432 - set rq_no_resend for cancel requests (2432)
6433 - recalculate ptlrpcd timeout after resend (2494)
6434 - call vfs_rmdir when removing pending directories (2368)
6435 - fix renaming a file to itself (2429)
6436 - lmc creates a default one-stripe lov (2454)
6437 - expand procfs space to handle large clusters (2326)
6438 - increase UML stack to avoid overflow
6439 - update lconf's list of debug and subsystem masks
6440 - fix lfs find --obd (2510)
6441 - /proc tunable for disabling filter read caching (2591)
6442 - stop rpm packages from altering slapd.conf (2301)
6443 - disable nagle in the socknal under 0conf (2578)
6444 - choose mds inode size based on stripe count (2572)
6445 - fix kernel-source rpm problems (2516)
6447 - add --disable-doc to avoid pdf generation (2421)
6448 - update documentation, tests, type-os, comments
6449 - avoid format warnings on ia64
6450 - remove the TOE NAL
6451 - tiny code cleanups by removing unused fields
6453 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6456 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6457 - remove the most blatant lies from BUILDING file (2371)
6458 - change default debug level to reasonable production setting
6459 - reduce client side cache size to reduce cache flush time
6460 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6461 - make TCP zerocopy and pinger support enabled by default (2476)
6462 - sync writes completed after process exits caused crashes (2319)
6463 - maintain correct mount count on the MDS (2356)
6464 - backout 1557, because 2316 wasn't really fixed
6465 - better file I/O statistics gathering in /proc
6466 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6467 - another kernel patch to fix zero-copy TCP function export
6468 - don't take duplicate lock when processing re-sent getattr (2420)
6469 - lctl uses obd_self_export instead of creating new conn (2353)
6470 - MDS/OST recovery case which requires object creation asserted (2425)
6471 - move lfs from /usr/sbin to /usr/bin in packages
6472 - fix race between mds_client_add and mds_client_free (2417)
6473 - use kmalloc instead of slabs in portals (2430)
6474 - don't create duplicate records when a failover MDS is present (2442)
6475 - remove unnecessary mount age check (2332)
6476 - don't remove directory inodes from locks prematurely (2451)
6477 - don't break if MDS service name is the same as hostname (2103)
6478 - fix races in client write RPC generation when cache full (2482)
6480 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6483 - remove now-unused request->rq_obd (278)
6484 - if an allocation fails, print out how much memory we've used (1933)
6485 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6486 - add the "configurable stack size" patch to most series files (1256)
6487 - ability to write large log records, for 100+ OST configs (2306)
6488 - fix NULL deref when filter_prep fails (2314)
6489 - fix operator precedence error in filter_sync
6490 - dynamic allocation of socknal TX descriptors (2315)
6491 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6492 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6493 - ensure that CWARN messages in Portals always get to the syslog
6494 - __init/__exit are not for prototype decls (ldlm_init/exit)
6495 - x86-64 compile warning fixes
6496 - fix gateway LMC keyword conflict (2318)
6497 - fix MDS lock inversions in getattr/reint paths (1844)
6498 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6499 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6500 - DLM race condition prevented some lock evictions (2328)
6501 - ENOMEM detection and retry on socknal sends (2230)
6502 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6503 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6504 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6505 - better error messages when a client is rejected during recovery (1505)
6506 - avoid cancelling locks which were never granted, after failure (2330)
6507 - fix i_sem/journal inversion in mds_client_add (2333)
6508 - fix truncate/getattr lock cycle deadlock (2334)
6509 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6510 - fix two rare exit paths which could leak an l_lock() ref (2321)
6511 - fencepost error in MDS/OST orphan recovery (2226)
6512 - make log record alignment 8 bytes (1988)
6513 - lstripe now fails when requested offset > ost_count (2237)
6514 - ensure that all kernel series have a complete list.h (1607)
6515 - fix crashes in special-file operations (2316)
6516 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6518 - allow configurable automake binary, for testing new versions
6519 - small update to the lfs documentation
6521 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6523 * fix negative export reference count in fsfilt_sync (2312)
6525 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6526 * release candidate 0.9.1
6528 - orphans are moved into the PENDING directory for possible recovery
6529 - replayed opens now open by fid for orphan/rename safety (1042)
6530 - last close of an orphan inode generates a transno (683)
6531 - chdir() and mount() now pin the directory entry (1020)
6532 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6533 - discard very old requests without processing them (1502)
6534 - remove some common, well-understood CERRORs (1505)
6535 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6536 - clear "grant" flags in OST replies until OST grant code lands (1644)
6537 - fix read performance by not clobbering i_blksize on client (1598)
6538 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6539 - make LRU size a /proc tunable, clears locks when reduced (707)
6540 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6541 - server locks take references on exports now (1558)
6542 - build fixes for 2.4.20-rh trees (1663)
6543 - return an error from lov_create if all OSCs are inactive (1751)
6544 - fix import levels when a reconnect happens without a timeout (1597)
6545 - exit early from mds_open if we get a lookup error (1749)
6546 - partial page read at EOF wouldn't wait for disk before sending (1642)
6547 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6548 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6549 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6550 - write/truncate lock inversion (1639)
6551 - Don't auto-load obdclass, portals modules during cleanup (1495)
6552 - fix timestamps from jumping to "now" (1763)
6553 - extra journal assertions (1648)
6554 - add an extra multiunlink test (1771)
6555 - fix read_record/write_record API (1776)
6556 - fix leak of offset_extent, possible incorrect i_size later (1772)
6557 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6558 - Debugging for the kqswnal_get_idle_tx problems (1820)
6559 - Allow recovery to be attempted multiple times (1536)
6560 - Write out MDS last_rcvd file after it is first created (1600)
6561 - Fix tx_descriptor leak in failed transmit situations (1827)
6562 - ext3 journaling fixes for assertion failure after IO error (1871)
6563 - class_export_put() on freed export after completion AST error (1896)
6564 - Fix revalidate looping in VFS (1322)
6565 - Don't access a freed export during MDS_REINT timeout (1521)
6566 - Add open-unlink recovery support on the MDS (1673,1764)
6567 - Return an error if no MDS data was read from last_rcvd (1946)
6568 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6569 - Don't setup a disk device that doesn't match exported UUID (317)
6570 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6571 - avoid committing NULL handle in force close
6572 - local.sh is now a one-stripe LOV configuration
6573 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6574 - fix ext3 htree duplicate directory entry corruption (1516)
6575 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6576 - update to new LOV EA format (2097)
6577 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6579 2003-06-15 Phil Schwan <phil@clusterfs.com>
6582 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6583 - per-import/export recovery handling (958, 931, 959)
6584 - multiple last-rcvd slots, for serving multiple FSes (949)
6585 - connections are again shared between multiple imp/exports (963, 964)
6586 - "umount -f" would hang if any requests needed to be sent (393, 978)
6587 - avoid pinning large req buffer by copying for queued messages (989)
6588 - add "uuid" to "lctl device" command to help upcalls (991)
6589 - "open" RPCs with transnos would confuse recovery counters (1037)
6590 - do proper endian conversion of all wire messages (288, 340, 891)
6591 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6592 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6593 - fix ptlrpc_connection leak in target_handle_connect (1174)
6594 - fix import refcounting bug in OST and MDS cleanup (1134)
6595 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6596 - fix very unlikely obd_types race condition (501)
6597 - remove osc_open hack for echo_client (1187)
6598 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6599 - a failure in read_inode2 leads to deadlock (1139)
6600 - cancel ack-locks as soon as transaction is committed (1072)
6601 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6602 - make sure to commitrw after any preprw to avoid deadlock (1162)
6603 - failing to execute a file in a lustre FS would lock inode (1203)
6604 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6605 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6606 - obd_uuid2tgt would walk off the end of the list (1255)
6607 - on IA64 the osc would give portals incorrect bulk size (1258)
6608 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6609 - fix lock inversion caused by new llite matching code (1282)
6610 - limit the number of dirty pages on a client to 10MB (1286)
6611 - timed out locks were not being corrected cancelled (1289)
6612 - fix O_DIRECT above 4GB on IA-32 (1292)
6613 * major user-visible changes
6614 - fail out/fail over policy now controlled by the upcall (993)
6616 - add OBD_PING to check server availability and failure (954)
6617 - lustre messages are now sent in sending host order (288, 340, 891)
6618 - add eadatalen to MDS getattr reply (340)
6619 - OST read replies may contain second buffer, with per-page status (593)
6621 2003-03-11 Phil Schwan <phil@clusterfs.com>
6624 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6625 - failed open()s could cause deadlock; fixed (867, 869)
6626 - stop cancelling OST locks when files are closed (481)
6627 - overlapping XID spaces caused network corruption (851, 853)
6628 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6629 - setattr_raw added, to do single-RPC, server-side setattrs
6630 - lmc/lconf syntax change for OST UUIDs
6631 - fix crashy race condition between ptlrpc_free_req and osc_close
6632 - don't use request in mdc_enqueue if we hit a timeout (889)
6633 - don't set the inode i_size for regular files from the MDS (896)
6634 - handle out of order completion AST (842)
6635 - don't LBUG if a lock request times out after receiving AST (913)
6636 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6637 - if a bad lock AST arrives, send an error instead of dropping entirely
6638 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6639 - fix leak in bulk IO when only partially completed (899, 900, 926)
6640 - fix O_DIRECT for ia64 (55)
6641 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6642 - C-z after timeout could hang a process forever; fixed (977)
6644 - client-side I/O cache (678, 924, 929, 941, 970)
6646 - READPAGE and SETATTRs which don't take server-side locks get
6649 2003-02-11 Phil Schwan <phil@clusterfs.com>
6652 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6653 - Fix fsfilt_extN_readpage() to read a full page of directory
6654 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6655 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6656 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6657 - Most intent operations were removed, in favour of a new RPC mode
6658 that does a single RPC to the server and bypasses most of the VFS
6659 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6660 - Aggressively cancel local locks on DLM servers
6661 - mds_reint_unlink sends EA to the client if it's the last nlink.
6662 client uses that EA to unlink OST objects.
6663 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6664 - recursive symlinks were fixed (439)
6665 - fixed NULL deref in DEBUG_REQ
6666 - filter_update_lastobjid no longer calls sync, which annoyed extN
6667 - fixed multi-client small-writes to a single file problem (445)
6668 - fixed mtime updates during file writes (607)
6669 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6670 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6671 - fixed Linux OST size reporting problem (444, 656)
6672 - OST now updates object mtime with writes or setattr (607, 619)
6673 - client verifies file size before zeroing page past EOF (445)
6674 - OST now writes last allocated objid to disk with allocation (108)
6675 - LOV on echo now works (409)
6677 - mds_reint_unlink sends a new buffer, with the EA included. this
6678 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6679 set if a regular file was being unlinked, and it was the last link
6680 - use PtlGet from the target for bulk writes (315)
6681 - OST now updates object mtime with writes or setattr (607, 619)
6682 - LDLM now has a grant-time callback to revalidate locked items, if
6684 - Many MDS operations were reorganized to combat race conditions
6686 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6687 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6689 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6692 - Fully reactivate OST imports after reconnection (512, others)
6693 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6694 - More graceful error handling for truncating on dead OST (515)
6695 - Don't error out unless we're actually accessing dead stripes (474)
6696 - Fix garbage sizes when stripes are missing (410)
6697 - LRU counters were broken, causing constant lock purge (433, 432)
6698 - garbage on read from stripes with failed OSTs (441)
6699 - mark OSCs as active before reconnecting during recovery (438)
6700 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6701 - lfind did not preserve OST order in output (443)
6702 - symlinks cause hung clients, incorrect data (439)
6703 - stop dereferencing request after dropping refcount (457)
6704 - don't LASSERT(spin_is_locked) on non-SMP (455)
6705 - fixes for many rename() bugs
6706 - fstat didn't correctly synchronize attributes (399)
6707 - server must handle lock cancellation during blocking AST prep (487)
6708 - bulk descriptors were free()d too soon (511)
6709 - fix paths in lconf, which would load incorrect modules (451, 507)
6710 - fix confusing lconf 'host not found' error message (386)
6711 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6712 - fix race condition in mdc_blocking_ast() for inode access (526)
6713 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6714 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6715 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6716 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6717 - fix lstripe in several ways (kernel side) (527)
6718 - fix request leak in ldlm_cli_enqueue (262)
6719 - incorrect OSC was marked inactive after OST failure
6720 - call mds_fs_cleanup before unmounting filesystem (524)
6721 - fix races between taking ns_lock and ldlm_lock_change_resource
6722 - fix races updating LOV export open file list
6723 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6724 - fix recovery NULL deref in ldlm_cli_cancel_unused
6725 - fix some DLM races by using new hash table for lock handles (419)
6726 - permit the client to specify desired inodes, at replay
6727 - duplicate requests when we queue them for replay reintegration
6728 - fix last_rcvd offset calculation
6729 - sync after each recovered transaction, so we always make progress
6730 - never, not always, ERESTART requests without transnos
6731 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6733 - skip replay if the MDS says that the client is already connected
6734 - don't check for a recovery-enabled export to match lctl's UUID
6735 - don't INC_USE_COUNT for phantom exports
6736 - don't crash when cleaning up phantom exports (567)
6737 - don't double-finish or set replay data for errored mdc_open requests
6738 - abort requests when they time out, so we don't get old replies
6739 - send/receive replies for AST messages again
6740 - if the client says that it doesn't have the lock, cancel it on the
6742 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6743 mark it as destroyed, it will all work out when decref is called
6744 - fix module use counts (22, 581)
6746 - ASTs now expect a reply (server cancels lock on error reply)
6748 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6751 - fix many simultaneous client startup (392)
6752 - fix dentry->d_it clobbering
6753 - credentials weren't being shipped for readdir/getattr operations
6754 - remove invalid assertions triggered during some concurrent MD
6756 - proper Lustre versions added (336, 389)
6757 - fix memory leak for create error case (398)
6758 - fix LOV locking bug that would get cli/srv out of sync
6759 - fix echo client over LOV (409)
6760 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6761 - fix double-O_EXCL intent crash (424)
6762 - avoid sending multiple lock CANCELs (352)
6764 - MDS can do multi-client recovery (modulo bugs in new code)
6766 - many updates, edits, cleanups
6768 2002-11-18 Phil Schwan <phil@clusterfs.com>
6771 - fix null d_it dereference (346)
6772 - fix full OST/dbench hang (333)
6773 - fix permission problem with file removal (286)
6774 - fix removal of OSCs from LOV when they fail
6775 - fix NULL deref during bulk timeout (214)
6776 - fix problems related to multiple filesystems on one MDS (241)
6777 - fixed serious subtle metadata locking bugs
6778 - free locks on clients when inodes are removed due to memory
6780 - fix inode pointer in lock data (285)
6781 - partial support for multiple MDS on a single host (241)
6782 - data locks weren't cancelled at clear_inode time (290, 311)
6783 - intent locks could lead to unbounded lock growth (205)
6784 - added a maximum lock count, an LRU list, and a flusher
6785 - fix multiple rename (365)
6786 - properly abstracted the echo client
6787 - OSC locked 1 byte too many; fixed
6788 - rewrote brw callback code:
6789 - fixed recovery bugs related to LOVs (306)
6790 - fixed too-many-pages-in-one-write crash (191)
6791 - fixed (again) crash in sync_io_timeout (214)
6792 - probably fixed callback-related race (385)
6794 - Add capability to MDS protocol
6795 - LDLM cancellations and callbacks on different portals
6797 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6800 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6801 - defer open object destruction to close time (601981, 138)
6802 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6803 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6804 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6805 - don't call obd_setattr in ll_file_release for destroyed objects
6806 * protocol change to lustre_msg: move |version| and add |flags|
6807 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6808 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6809 * added replay of create, unlink, link and rename operations during
6810 MDS failover; recovery should be much more robust now
6811 * remove failed OSCs from LOVs (only lov_create uses this so far)
6812 * the lustre-HOWTO was brought (more) up to date (582544)
6814 2002-10-23 Phil Schwan <phil@clusterfs.com>
6817 - in-use dentries weren't being reused properly (617851)
6818 - prevent multiple LDLM setup (599178)
6819 - fix LOV size calculations for truncate (617853)
6820 - fix client handling of MDS intent errors (POSIX)
6821 - fix permission bug in lovstripe.c test (624321)
6822 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6823 - truncate past end of file could corrupt data
6824 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6825 - a race in recovery could return ETIMEDOUT to apps (623947)
6826 - building outside the source directory was fixed
6827 * the lustre-HOWTO was brought (more) up to date (582544)
6828 * major progress was made on recovery functionality
6830 2002-10-10 Phil Schwan <phil@clusterfs.com>
6833 - recovery deadlock fix
6834 - rm -rf causes LBUG fix (617817)
6835 - file open by multiple tasks fix (618962)
6836 - directory permissions bugs (602707 and 620007)
6837 - journal_stop fixed with locking (611313)
6838 - O_APPEND failures resolved (618273, perhaps 614459)
6839 - lconf PATH fix (619770)
6840 - IA64 build fix (621450)
6841 - RPC buffer sizes scale with amount of memory
6843 2002-10-01 Phil Schwan <phil@clusterfs.com>
6846 - locks would be cancelled without throwing away data pages,
6847 resulting in inconsistent data (605627)
6848 - inode attributes were not always being refreshed (605627, 612449)
6849 - lconf now continues to cleanup after lctl reports an error
6850 - MDS now enforces user permissions (602707)
6851 - lprocfs cleanup fixed, but not yet enabled (614157)
6852 - fixed infinite server hang, should a client not respond to an AST
6853 - avoid going into recovery if user calls readlink() with a buffer
6854 that's too small (613941)
6855 - AST RPCs no longer require replies (614867) -- this may be changed
6856 - don't crash server if client sends an IOV that's too big (611336)
6857 - fixed lock conversion deadlock (611892)
6858 - fixed the following of symlinks (614622)
6859 * recovery: the server can remove locks from a client that dies, other
6860 clients can make progress
6861 * more extN patch fixes
6862 * compile-time configurable ptlrpc buffer allocations
6864 - collaborative read cache document
6865 - Lustre Lite Performance CDR document-in-progress
6867 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6870 - fix typo in patch-2.4.18
6872 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6875 - clear ptlrpc request each time in handle_incoming_request()
6876 - unlink of files now destroys the object on the OST
6878 2002-09-19 Peter Braam <braam@clusterfs.com>
6880 * add hard link support
6881 * change obdfile creation method
6882 * kernel patch changed
6884 2002-09-19 Peter Braam <braam@clusterfs.com>
6887 - stack overflow bug in extN fixed
6889 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6891 * documentation updates
6892 - add man pages for config tools
6893 - update tests/README to describe testing with new config tools
6894 - finish metadata API descriptions
6895 * bug fixes and cleanups
6896 - statfs workaround for 16TB limit
6897 - LOV stripe allocation improved, can stripe on subset of OSTs
6898 - LOV file size/IO offset was wrong for files > 4GB in size
6899 - object EA data was being dropped, caused files to be unreadable
6900 - memory overflow with non-LOV OST caused memory corruption
6901 - fixed regression tests to work with new config tools, obdfilter
6902 - fixed bug when directory size became larger than 1 block
6903 - fixed bug (for single client case) when PWD was deleted
6904 - invalidate local directory pages when doing intent-based ops
6905 - avoid LDLM oops when lock callback contained bad data
6907 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6909 * documentation updates
6910 * bug fixes and cleanups
6911 - configuration tools
6914 - 64-bit compile warnings
6915 - 64-bit internal statfs data
6917 * test_brw on persistent OST devices
6919 * lprocfs (disabled)
6921 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6923 * documentation updates
6924 * bug fixes and cleanups
6925 * configuration tools
6927 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6930 * many small fixes to 0_5_4
6931 * io/network handling
6932 * thinkos in MDS operations
6934 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6937 * crucial basic fixes to 0.5.3
6938 * IOR, Iozone work over Elan
6941 2002-08-07 Phil Schwan <phil@clusterfs.com>
6942 * version 0_5_3, our first alpha
6943 * we use the new Portals iovs
6944 * documentation updates
6945 * bug fixes and cleanups
6946 * small changes in the DLM wire protocol
6948 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6949 * version 0_5_1 with some initial stability,
6950 * locking on MD and file I/O.
6951 * documentation updates
6952 * several bug fixes since 0.5.0
6953 * small changes in wire protocol
6955 2002-07-18 Phil Schwan <phil@clusterfs.com>
6957 * delivered as Lustre Light Alpha
6958 * fixed a crash after handling invalid MDS requests
6959 * fixed directory pages for architectures with non-4k pages sizes
6961 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6962 * release version v0_4_4
6963 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6965 * Fixes a number of interruption problems with OST operations.
6966 * Update documentation for portals header changes
6967 * Move all wire protocol structs/defines to lustre_idl.h
6968 * Fixes symlink length bug.
6969 * Add tcpdump to repository.
6971 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6972 * release version v0_4_3
6973 * Fixes statfs for inodes on extN.
6974 * Fixes bug in runtests which would delete /etc/hosts.
6975 * Use 64-bit object IDs wherever possible (not into VFS though)
6976 Remove ost_get_info, which is unused by lustre, and out of date.
6978 2002-07-03 Peter Braam <braam@clusterfs.com>
6979 * release version v0_4_2 Fixes a lookup error (type not passed)
6980 * move forward to head of Portals
6981 * move forward to latest Lustre kernel
6983 2002-06-25 Peter Braam <braam@clusterfs.com>
6984 * release version v0_4_1. Hopefully stable on single node use.