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 3.10.0-1062.9.1.el7 (RHEL7.7)
8 * Server known to build on the following kernels (others may also work):
9 2.6.32-696.18.7.el6 (RHEL6.9)
10 3.10.0-862.14.4.el7 (RHEL7.5)
11 4.4.120-92.70 (SLES12 SP2)
12 4.4.180-94.97 (SLES12 SP3)
13 4.4.0-142 (Ubuntu 16.04)
14 4.15.0-45 (Ubuntu 18.04)
15 vanilla linux 4.6.7 (ZFS only)
16 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
17 * Client primary unpatched kernels built and tested during release cycle:
18 3.10.0-957.27.2.el7 (RHEL7.6)
19 3.10.0-1062.9.1.el7 (RHEL7.7)
20 3.10.0-1127.el7 (RHEL7.8)
21 4.18.0-80.11.2.el8 (RHEL8.0)
22 4.4.180-94.97 (SLES12 SP3)
23 4.12.14-95.45 (SLES12 SP4)
24 4.15.0-45 (Ubuntu 18.04)
25 * Client known to build on unpatched kernels (others may also work):
26 2.6.32-696.18.7.el6 (RHEL6.9)
27 2.6.32-754.6.3.el6 (RHEL6.10)
28 3.10.0-862.14.4.el7 (RHEL7.5)
29 4.14.0-49.13.1.el7a (RHEL7.5)
30 4.14.0-115.2.2.el7a (RHEL7.6)
31 4.18.0-147.3.1.el8 (RHEL8.1)
32 4.4.120-92.70 (SLES12 SP2)
33 4.4.0-142 (Ubuntu 16.04)
34 vanilla linux 4.15.0 (ZFS only)
35 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
36 * Recommended ZFS / SPL version: 0.7.13
37 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
38 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
39 stack overflow. For more information, please refer to bugzilla 17630.
40 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
41 the re-exporting nodes to be mounted with "32bitapi" mount option
42 * Note for Ubuntu 14.04.5:
43 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
44 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
48 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
49 for currently supported client and server kernel versions.
50 * Server primary kernels built and tested during release cycle:
51 3.10.0-957.27.2.el7 (RHEL7.6)
52 3.10.0-1062.9.1.el7 (RHEL7.7)
53 * Server known to build on the following kernels (others may also work):
54 2.6.32-696.18.7.el6 (RHEL6.9)
55 3.10.0-862.14.4.el7 (RHEL7.5)
56 4.4.120-92.70 (SLES12 SP2)
57 4.4.180-94.97 (SLES12 SP3)
58 4.4.0-142 (Ubuntu 16.04)
59 4.15.0-45 (Ubuntu 18.04)
60 vanilla linux 4.6.7 (ZFS only)
61 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
62 * Client primary unpatched kernels built and tested during release cycle:
63 3.10.0-957.27.2.el7 (RHEL7.6)
64 3.10.0-1062.9.1.el7 (RHEL7.7)
65 4.18.0-80.11.2.el8 (RHEL8.0)
66 4.4.180-94.97 (SLES12 SP3)
67 4.12.14-95.29 (SLES12 SP4)
68 4.15.0-45 (Ubuntu 18.04)
69 * Client known to build on unpatched kernels (others may also work):
70 2.6.32-696.18.7.el6 (RHEL6.9)
71 2.6.32-754.6.3.el6 (RHEL6.10)
72 3.10.0-862.14.4.el7 (RHEL7.5)
73 4.14.0-49.13.1.el7a (RHEL7.5)
74 4.14.0-115.2.2.el7a (RHEL7.6)
75 4.18.0-147.3.1.el8 (RHEL8.1)
76 4.4.120-92.70 (SLES12 SP2)
77 4.4.0-142 (Ubuntu 16.04)
78 vanilla linux 4.15.0 (ZFS only)
79 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
80 * Recommended ZFS / SPL version: 0.7.13
81 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
82 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
83 stack overflow. For more information, please refer to bugzilla 17630.
84 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
85 the re-exporting nodes to be mounted with "32bitapi" mount option
86 * Note for Ubuntu 14.04.5:
87 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
88 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
92 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
93 for currently supported client and server kernel versions.
94 * Server primary kernels built and tested during release cycle:
95 3.10.0-957.27.2.el7 (RHEL7.6)
96 3.10.0-1062.1.1.el7 (RHEL7.7)
97 * Server known to build on the following kernels (others may also work):
98 2.6.32-696.18.7.el6 (RHEL6.9)
99 3.10.0-862.14.4.el7 (RHEL7.5)
100 4.4.120-92.70 (SLES12 SP2)
101 4.4.180-94.97 (SLES12 SP3)
102 4.4.0-142 (Ubuntu 16.04)
103 4.15.0-45 (Ubuntu 18.04)
104 vanilla linux 4.6.7 (ZFS only)
105 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
106 * Client primary unpatched kernels built and tested during release cycle:
107 3.10.0-957.27.2.el7 (RHEL7.6)
108 3.10.0-1062.1.1.el7 (RHEL7.7)
109 4.18.0-80.7.1.el8 (RHEL8.0)
110 4.4.180-94.97 (SLES12 SP3)
111 4.12.14-95.29 (SLES12 SP4)
112 4.15.0-45 (Ubuntu 18.04)
113 * Client known to build on unpatched kernels (others may also work):
114 2.6.32-696.18.7.el6 (RHEL6.9)
115 2.6.32-754.6.3.el6 (RHEL6.10)
116 3.10.0-862.14.4.el7 (RHEL7.5)
117 4.14.0-49.13.1.el7a (RHEL7.5)
118 4.14.0-115.2.2.el7a (RHEL7.6)
119 4.4.120-92.70 (SLES12 SP2)
120 4.4.0-142 (Ubuntu 16.04)
121 vanilla linux 4.15.0 (ZFS only)
122 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
123 * Recommended ZFS / SPL version: 0.7.13
124 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
125 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
126 stack overflow. For more information, please refer to bugzilla 17630.
127 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
128 the re-exporting nodes to be mounted with "32bitapi" mount option
129 * Note for Ubuntu 14.04.5:
130 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
131 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
135 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
136 for currently supported client and server kernel versions.
137 * Server known to build on the following kernels (others may also work):
138 2.6.32-696.18.7.el6 (RHEL6.9)
139 3.10.0-862.14.4.el7 (RHEL7.5)
140 3.10.0-957.10.1.el7 (RHEL7.6)
141 4.4.120-92.70 (SLES12 SP2)
142 4.4.162-94.69 (SLES12 SP3)
143 4.4.0-142 (Ubuntu 16.04)
144 4.15.0-45 (Ubuntu 18.04)
145 vanilla linux 4.6.7 (ZFS only)
146 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
147 * Client known to build on unpatched kernels (others may also work):
148 2.6.32-696.18.7.el6 (RHEL6.9)
149 2.6.32-754.6.3.el6 (RHEL6.10)
150 3.10.0-862.14.4.el7 (RHEL7.5)
151 3.10.0-957.10.1.el7 (RHEL7.6)
152 4.14.0-49.13.1.el7a (RHEL7.5)
153 4.14.0-115.2.2.el7a (RHEL7.6)
154 4.4.120-92.70 (SLES12 SP2)
155 4.4.162-94.69 (SLES12 SP3)
156 4.12.14-95.13 (SLES12 SP4)
157 4.4.0-142 (Ubuntu 16.04)
158 4.15.0-45 (Ubuntu 18.04)
159 vanilla linux 4.15.0 (ZFS only)
160 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
161 * Recommended ZFS / SPL version: 0.7.13
162 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
163 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
164 stack overflow. For more information, please refer to bugzilla 17630.
165 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
166 the re-exporting nodes to be mounted with "32bitapi" mount option
167 * Note for Ubuntu 14.04.5:
168 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
169 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
174 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
175 for currently supported client and server kernel versions.
176 * Server known to build on the following kernels (others may also work):
177 2.6.32-696.18.7.el6 (RHEL6.9)
178 3.10.0-862.14.4.el7 (RHEL7.5)
179 3.10.0-957.10.1.el7 (RHEL7.6)
180 4.4.120-92.70 (SLES12 SP2)
181 4.4.162-94.69 (SLES12 SP3)
182 4.4.0-142 (Ubuntu 16.04)
183 4.15.0-45 (Ubuntu 18.04)
184 vanilla linux 4.6.7 (ZFS only)
185 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
186 * Client known to build on unpatched kernels (others may also work):
187 2.6.32-696.18.7.el6 (RHEL6.9)
188 2.6.32-754.6.3.el6 (RHEL6.10)
189 3.10.0-862.14.4.el7 (RHEL7.5)
190 3.10.0-957.10.1.el7 (RHEL7.6)
191 4.14.0-49.13.1.el7a (RHEL7.5)
192 4.14.0-115.2.2.el7a (RHEL7.6)
193 4.4.120-92.70 (SLES12 SP2)
194 4.4.162-94.69 (SLES12 SP3)
195 4.12.14-95.13 (SLES12 SP4)
196 4.4.0-142 (Ubuntu 16.04)
197 4.15.0-45 (Ubuntu 18.04)
198 vanilla linux 4.15.0 (ZFS only)
199 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
200 * Recommended ZFS / SPL version: 0.7.13
201 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
202 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
203 stack overflow. For more information, please refer to bugzilla 17630.
204 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
205 the re-exporting nodes to be mounted with "32bitapi" mount option
206 * Note for Ubuntu 14.04.5:
207 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
208 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
212 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
213 for currently supported client and server kernel versions.
214 * Server known to build on the following kernels (others may also work):
215 2.6.32-696.18.7.el6 (RHEL6.9)
216 3.10.0-862.14.4.el7 (RHEL7.5)
217 3.10.0-957.el7 (RHEL7.6)
218 4.4.120-92.70 (SLES12 SP2)
219 4.4.162-94.69 (SLES12 SP3)
220 4.4.0-131 (Ubuntu 16.04)
221 4.15.0-32 (Ubuntu 18.04)
222 vanilla linux 4.6.7 (ZFS only)
223 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
224 * Client known to build on unpatched kernels (others may also work):
225 2.6.32-696.18.7.el6 (RHEL6.9)
226 2.6.32-754.6.3.el6 (RHEL6.10)
227 3.10.0-862.14.4.el7 (RHEL7.5)
228 3.10.0-957.el7 (RHEL7.6)
229 4.14.0-49.13.1.el7a (RHEL7.5)
230 4.14.0-115.2.2.el7a (RHEL7.6)
231 4.4.120-92.70 (SLES12 SP2)
232 4.4.162-94.69 (SLES12 SP3)
233 4.4.0-131 (Ubuntu 16.04)
234 4.15.0-32 (Ubuntu 18.04)
235 vanilla linux 4.15.0 (ZFS only)
236 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
237 * Recommended ZFS / SPL version: 0.7.9
238 * Tested with ZFS / SPL version: 0.7.9
239 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
240 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
241 stack overflow. For more information, please refer to bugzilla 17630.
242 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
243 the re-exporting nodes to be mounted with "32bitapi" mount option
244 * Note for Ubuntu 14.04.5:
245 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
246 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
248 --------------------------------------------------------------------------------
251 2018-04-03 Intel Corporation
253 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
254 for currently supported client and server kernel versions.
255 * Server known to build on patched kernels:
256 2.6.32-431.29.2.el6 (RHEL6.5)
257 2.6.32-504.30.3.el6 (RHEL6.6)
258 2.6.32-573.26.1.el6 (RHEL6.7)
259 2.6.32-642.15.1.el6 (RHEL6.8)
260 2.6.32-696.18.7.el6 (RHEL6.9)
261 3.10.0-693.21.1.el7 (RHEL7.4)
262 3.0.101-0.47.71 (SLES11 SP3)
263 3.0.101-107 (SLES11 SP4)
264 3.12.74-60.64.40 (SLES12 SP1)
265 4.4.114-92.64 (SLES12 SP2)
266 4.4.114-94.11 (SLES12 SP3)
267 3.13.0-101 (Ubuntu 14.04, ZFS only)
268 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
269 4.4.0-116 (Ubuntu 16.04)
270 vanilla linux 4.6.7 (ZFS only)
271 * Client known to build on unpatched kernels:
272 2.6.32-431.29.2.el6 (RHEL6.5)
273 2.6.32-504.30.3.el6 (RHEL6.6)
274 2.6.32-573.26.1.el6 (RHEL6.7)
275 2.6.32-642.15.1.el6 (RHEL6.8)
276 2.6.32-696.18.7.el6 (RHEL6.9)
277 3.10.0-693.21.1.el7 (RHEL7.4)
278 3.0.101-0.47.71 (SLES11 SP3)
279 3.0.101-107 (SLES11 SP4)
280 3.12.74-60.64.40 (SLES12 SP1)
281 4.4.114-92.64 (SLES12 SP2)
282 4.4.114-94.11 (SLES12 SP3)
283 3.13.0-101 (Ubuntu 14.04)
284 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
285 4.4.0-116 (Ubuntu 16.04)
287 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
288 * Recommended ZFS / SPL version: 0.7.6
289 * Tested with ZFS / SPL version: 0.7.6
290 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
291 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
292 stack overflow. For more information, please refer to bugzilla 17630.
293 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
294 the re-exporting nodes to be mounted with "32bitapi" mount option
295 * Note for Ubuntu 14.04.5:
296 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
297 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
299 --------------------------------------------------------------------------------
301 2017-07-05 Intel Corporation
303 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
304 for currently supported client and server kernel versions.
305 * Server known to build on patched kernels:
306 2.6.32-431.29.2.el6 (RHEL6.5)
307 2.6.32-504.30.3.el6 (RHEL6.6)
308 2.6.32-573.26.1.el6 (RHEL6.7)
309 2.6.32-642.15.1.el6 (RHEL6.8)
310 2.6.32-696.3.1.el6 (RHEL6.9)
311 3.10.0-514.21.1.el7 (RHEL7.3)
312 3.0.101-0.47.71 (SLES11 SP3)
313 3.0.101-100 (SLES11 SP4)
314 3.12.74-60.64.40 (SLES12 SP1)
315 4.4.59-92.17 (SLES12 SP2)
316 vanilla linux 4.6.7 (ZFS only)
317 * Client known to build on unpatched kernels:
318 2.6.32-431.29.2.el6 (RHEL6.5)
319 2.6.32-504.30.3.el6 (RHEL6.6)
320 2.6.32-573.26.1.el6 (RHEL6.7)
321 2.6.32-642.15.1.el6 (RHEL6.8)
322 2.6.32-696.3.1.el6 (RHEL6.9)
323 3.10.0-514.21.1.el7 (RHEL7.3)
324 3.0.101-0.47.71 (SLES11 SP3)
325 3.0.101-100 (SLES11 SP4)
326 3.12.74-60.64.40 (SLES12 SP1)
327 4.4.59-92.17 (SLES12 SP2)
329 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
330 * Recommended ZFS / SPL version: 0.6.5.9
331 * Tested with ZFS / SPL version: 0.6.5.9
332 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
333 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
334 stack overflow. For more information, please refer to bugzilla 17630.
335 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
336 the re-exporting nodes to be mounted with "32bitapi" mount option
338 --------------------------------------------------------------------------------
340 12-07-2016 Intel Corporation
342 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
343 for currently supported client and server kernel versions.
344 * Server known to build on patched kernels:
345 2.6.32-431.29.2.el6 (RHEL6.5)
346 2.6.32-504.30.3.el6 (RHEL6.6)
347 2.6.32-573.26.1.el6 (RHEL6.7)
348 2.6.32-642.11.1.el6 (RHEL6.8)
349 3.10.0-514.el7 (RHEL7.3)
350 3.0.101-0.47.71 (SLES11 SP3)
351 3.0.101-80 (SLES11 SP4)
352 3.12.62-60.64.8 (SLES12 SP1)
353 vanilla linux 4.5.7 (ZFS only)
354 * Client known to build on unpatched kernels:
355 2.6.32-431.29.2.el6 (RHEL6.5)
356 2.6.32-504.30.3.el6 (RHEL6.6)
357 2.6.32-573.26.1.el6 (RHEL6.7)
358 2.6.32-642.11.1.el6 (RHEL6.8)
359 3.10.0-514.el7 (RHEL7.3)
360 3.0.101-0.47.71 (SLES11 SP3)
361 3.0.101-80 (SLES11 SP4)
362 3.12.62-60.64.8 (SLES12 SP1)
364 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
365 * Recommended ZFS / SPL version: 0.6.5.8
366 * Tested with ZFS / SPL version: 0.6.5.8
367 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
368 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
369 stack overflow. For more information, please refer to bugzilla 17630.
370 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
371 the re-exporting nodes to be mounted with "32bitapi" mount option
373 --------------------------------------------------------------------------------
375 02-29-2016 Intel Corporation
377 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
378 for currently supported client and server kernel versions.
379 * Server known to build on patched kernels:
380 2.6.32-431.29.2.el6 (RHEL6.5)
381 2.6.32-504.30.3.el6 (RHEL6.6)
382 2.6.32-573.12.1.el6 (RHEL6.7)
383 3.10.0-327.3.1.el7 (RHEL7.2)
384 3.0.101-0.47.71 (SLES11 SP3)
385 3.0.101-68 (SLES11 SP4)
386 vanilla linux 4.2.1 (ZFS only)
387 * Client known to build on unpatched kernels:
388 2.6.32-431.29.2.el6 (RHEL6.5)
389 2.6.32-504.30.3.el6 (RHEL6.6)
390 2.6.32-573.12.1.el6 (RHEL6.7)
391 3.10.0-327.3.1.el7 (RHEL7.2)
392 3.0.101-0.47.71 (SLES11 SP3)
393 3.0.101-68 (SLES11 SP4)
396 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
397 * Recommended ZFS / SPL version: 0.6.4.2
398 * Tested with ZFS / SPL version: 0.6.4.2
399 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
400 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
401 stack overflow. For more information, please refer to bugzilla 17630.
402 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
403 the re-exporting nodes to be mounted with "32bitapi" mount option
405 --------------------------------------------------------------------------------
407 03-10-2015 Intel Corporation
409 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
410 for currently supported client and server kernel versions.
411 * Server known to build on patched kernels:
412 2.6.32-431.29.2.el6 (RHEL6.5)
413 2.6.32-504.8.1.el6 (RHEL6.6)
414 3.0.101-0.46 (SLES11 SP3)
415 * Client known to build on unpatched kernels:
416 2.6.32-431.29.2.el6 (RHEL6.5)
417 2.6.32-504.8.1.el6 (RHEL6.6)
418 3.10.0-123.20.1.el7 (RHEL7)
419 3.0.101-0.46 (SLES11 SP3)
420 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
421 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
422 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
423 stack overflow. For more information, please refer to bugzilla 17630.
424 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
425 the re-exporting nodes to be mounted with "32bitapi" mount option
427 Severity : enhancement
429 Description: control OST-index in IDIF via ROCOMPAT flag.
430 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
431 last_rcvd file. For new formatted OST device, it will be auto set;
432 for the case of upgrading from old OST device, you can enable it
433 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
434 enabled, for new created OST-object, its IDIF-in-LMA will contain
435 the OST-index; for the existing OST-object, the OSD will convert
436 old format IDIF as new format IDIF with OST-index stored in the
437 LMA EA when accessing such OST-object or via OI scrub. Once such
438 flag is enabled, it cannot be reverted back, so the system cannot
439 be downgraded to the orignal incompatible version.
441 --------------------------------------------------------------------------------
443 07-30-2014 Intel Corporation
445 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
446 for currently supported client and server kernel versions.
447 * Server known to build on patched kernels:
448 2.6.32-431.20.3.el6 (RHEL6.5)
449 3.0.101-0.31 (SLES11 SP3)
450 * Client known to build on unpatched kernels:
451 2.6.32-431.20.3.el6 (RHEL6.5)
452 3.0.101-0.31 (SLES11 SP3)
455 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
456 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
457 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
458 stack overflow. For more information, please refer to bugzilla 17630.
459 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
460 the re-exporting nodes to be mounted with "32bitapi" mount option
462 --------------------------------------------------------------------------------
464 10-31-2013 Intel Corporation
466 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
467 * Server builds for kernels:
468 2.6.32-358.18.1.el6 (RHEL6.4)
469 * Client builds for unpatched kernels:
470 2.6.32-358.18.1.el6 (RHEL6.4)
471 3.0.92-0.8 (SLES11 SP3)
473 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
474 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
475 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
476 stack overflow. For more information, please refer to bugzilla 17630.
477 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
478 the re-exporting nodes to be mounted with "32bitapi" mount option
480 --------------------------------------------------------------------------------
482 05-30-2013 Intel Corporation
484 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
485 * Server builds for kernels:
486 2.6.32-358.6.2.el6 (RHEL6.4)
487 * Client builds for unpatched kernels:
488 2.6.32-358.6.2.el6 (RHEL6.4)
489 3.0.74-0.6.6 (SLES11 SP2)
491 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
492 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
493 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
494 stack overflow. For more information, please refer to bugzilla 17630.
495 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
496 the re-exporting nodes to be mounted with "32bitapi" mount option
498 --------------------------------------------------------------------------------
500 9-30-2012 Whamcloud, Inc.
502 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
503 * Server support for kernels:
504 2.6.32-279.5.1.el6 (RHEL6)
505 * Client support for unpatched kernels:
506 2.6.32-279.5.1.el6 (RHEL6)
508 2.6.18-238.19.1.el5 (RHEL5)
509 2.6.16.60-0.69.1 (SLES10)
510 * Recommended e2fsprogs version: 1.41.90.wc4
511 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
512 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
513 stack overflow. For more information, please refer to bugzilla 17630.
514 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
515 the re-exporting nodes to be mounted with "32bitapi" mount option
517 Severity : enhancement
519 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
521 Severity : enhancement
523 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
525 Severity : enhancement
527 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
529 Severity : enhancement
531 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
533 Severity : enhancement
535 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
537 --------------------------------------------------------------------------------
539 04-30-2012 Whamcloud, Inc.
541 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
542 * Server support for kernels:
543 2.6.32-220.4.2.el6 (RHEL6)
544 * Client support for unpatched kernels:
545 2.6.32-220.4.2.el6 (RHEL6)
547 2.6.18-238.19.1.el5 (RHEL5)
548 2.6.16.60-0.69.1 (SLES10)
549 * Recommended e2fsprogs version: 1.41.90.wc4
550 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
551 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
552 stack overflow. For more information, please refer to bugzilla 17630.
553 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
554 the re-exporting nodes to be mounted with "32bitapi" mount option
556 Severity : enhancement
558 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
560 Severity : enhancement
562 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
564 Severity : enhancement
566 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
568 Severity : enhancement
570 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
572 --------------------------------------------------------------------------------
574 09-30-2011 Whamcloud, Inc.
576 * Server support for kernels:
577 2.6.18-238.19.1.el5 (RHEL5)
578 2.6.32-131.6.1.el6 (RHEL6)
579 * Client support for unpatched kernels:
580 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
581 2.6.18-238.19.1.el5 (RHEL5)
582 2.6.32-131.6.1.el6 (RHEL6)
583 2.6.29.4-167.fc11 (FC11)
584 2.6.16.60-0.69.1 (SLES10)
586 * Recommended e2fsprogs version: 1.41.90.wc3
587 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
588 Lustre filesystem with 4K stack may cause a stack overflow. For more
589 information, please refer to bugzilla 17630.
590 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
591 mounted with "32bitapi" mount option
593 Severity : enhancement
594 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
596 Severity : enhancement
597 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
599 Severity : enhancement
600 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
603 Frequency : only with SLES10
605 Description: Use OFED "KMP" provided by Novell
606 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
607 Lustre is now built against this package. That means you need to
608 install the ofed-kmp package from Novell for the patchless client.
609 Note that the ofed-kmp that Novell ships may not exactly match the
610 kernel version but should still be compatible.
612 Severity : enhancement
614 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
615 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
617 --------------------------------------------------------------------------------
619 2010-07-15 Oracle, Inc.
621 * Server support for kernels:
622 2.6.18-164.11.1.el5 (RHEL 5)
623 2.6.18-164.11.1.0.1.el5 (OEL 5)
624 * Client support for unpatched kernels:
625 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
626 2.6.18-164.11.1.el5 (RHEL 5),
627 2.6.18-164.11.1.0.1.el5 (OEL 5)
628 2.6.16.60-0.42.8 (SLES 10),
630 2.6.29.4-167.fc11 (FC11)
631 * Recommended e2fsprogs version: 1.41.10-sun2
633 Severity : enhancement
635 Description: Distribute lfs_migrate script with lustre
636 Details : The lfs_migrate mentioned in the manual is now included in the
637 lustre tree in lustre/scripts and will distributed in lustre RPMs.
639 Severity : enhancement
641 Description: Failover nid list added to import proc file.
643 Severity : enhancement
645 Description: "som_preview" mount option required to use SOM
646 Details : The clients will need to use som_preview mount option
650 Description: Lustre does not do 1MB IOs to HW RAID
651 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
653 Severity : enhancement
655 Description: lfs quota output is not very convenient for awk/sed-parsing
656 Details : Some positions in lfs quota output table could be empty or
657 non-empty which made it hard to parse it with scripts, now
658 a dash is put instead of space where there is not supposed
663 Description: Update to OFED 1.5.1
664 Details : OFED 1.5.1 will be used in environments that don't use
665 vendor supplied OFED.
669 Description: Remove lnet. prefix from lctl params display
670 Details : Also renamed lnet.memused to lnet_memused to
675 Description: Use conf_param -d to remove permanant settings.
679 Description: Add list_param -R
680 Details : lctl list_param now supports recursive mode.
683 Bugzilla : 21829, 22312
684 Description: Fix llobdstat.
685 Details : 1. Now updating the write_bytes counter
686 2. Port script fixes from 1.8.
690 Description: decrease the usage of memory on clients.
691 Details : 1. On clients, recycle dentries and inodes unused.
692 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
697 Description: llobdstat fix and enhancement
698 Details : add a counter to set a limit to how many samples will be returned
699 fix a wildcard in the path to limit to obdfilter stats only
703 Description: lfs check only allowed for root
704 Details : Code cleanups and added a regresion test of non-root use of
709 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
710 Details : Fix hr->hr_index overload.
714 Description: Fix FIEMAP functionality.
718 Description: FC11 patchless client support added
722 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
723 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
727 Description: New acc-sm script and various test-framework enhancments.
731 Description: Previously landed patch was reverted.
735 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
739 Description: Add ldiskfs maxdirsize mount option.
743 Description: Fix mkfs.lustre support for 16TB LUNs.
747 Description: Print mmp_check_interval
748 Details : print mmp_check_interval and make it possible to abort
749 mount operation in case it takes too long.
753 Description: Add cascading_rw to lustre mpi tests.
757 Description: Store SOM data in EA.
758 Details : Also several SOM related cleanups.
762 Description: Add refcount to md_open_data
766 Description: Use refcount on lov_request_set to prevent memory corruption.
770 Description: Statahead: use parent dir's i_mutex for synchronization.
774 Description: loadgen improvements.
778 Description: Enable peer health detection in ksocklnd and ko2iblnd.
782 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
786 Description: Fix race on obd_devcie:md_stats
790 Description: Fix SLES 11 client builds.
791 Details : Resolved uninitalized pointer gcc warnings.
795 Description: Allow non-root access for "lfs check".
796 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
800 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
802 Severity : enhancement
804 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
805 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
809 Description: Sanity test_27p fails.
810 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
811 Sanity tests 27 were cleaned up and improved.
815 Description: Unify req format on client/servers
816 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
817 buffers and swabbers approach..
822 Description: drop join file stuff.
824 Severity : enhancement
826 Description: Add 2.6.27 and SLES11 (patchless client) support
830 Description: Fix in ptlrpc_expire_one_request() to print the signed time
833 Severity : enhancement
835 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
838 Severity : enhancement
840 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
842 Severity : enhancement
844 Description: correctly handle big reply message.
845 Details : send LNet event if reply is bigger than buffer and adjust this buffer
850 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
852 Severity : enhancement
854 Description: Add support for OEL5.
858 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
860 Severity : enhancement
862 Description: "weak-modules" support
863 Details : Implement "weak-modules" support which enables kernel modules
864 to be used with any kernel that implements the same kABI. In
865 order to achieve this modules are now installed in
866 /lib/modules/$(uname -r)/updates/kernel on all distributions.
870 Description: client could not reconnect to OST because of an active request.
871 Details : The client now retries to reconnect to the same server, if a
872 connect request failed with EBUSY or -EAGAIN. Test case 35b
873 conf-sanity.sh has been created.
877 Description: Don't allow make backward step on assignin osc next id.
878 Details : race between allocation next id and ll_sync thread can be cause
879 of set wrong osc next id and can be kill valid ost objects.
883 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
887 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
891 Description: Error handling in osc_statfs_interpret() has been improved.
892 Details : Check in osc_statfs_interpret() for EBADR.
896 Description: Do not send statfs() requests to OSTs disabled by administrator.
897 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
899 Severity : enhancement
901 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
903 Severity : enhancement
905 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
907 Severity : enhancement
909 Description: Enhancement for lfs(1) command to use numeric uid/gid.
913 Description: Improved cleanup procedure for conf-sanity test_46a.
917 Description: Do not start bulk transfer, if client tried to reconnect.
918 Details : Do not start bulk transfer, if client tried to reconnect,
919 but failed due to active rpcs. Check for exp_abort_active_req
920 has been implemented.
924 Description: Prevent inconsistences between linux and lustre mount structures.
925 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
926 Make the sleep interruptible.
928 Severity : enhancement
930 Description: provide server to client comms path
931 Details : server to client communications path via new LDLM_SET_INFO rpc
935 Description: truncate starts GFP_FS allocation under transaction causing deadlock
936 Details : ldiskfs_truncate calls grab_cache_page which may start page
937 allocation under an open transaction. This may lead to
938 calling prune_icache with consequent lustre reentrance.
942 Description: send correctly lsm on open replay
943 Details : MDS is trust to LSM size on replay open, but client can set wrong size
948 Description: enable client interface failover
949 Details : When a child reconnects from another NID, properly update export
950 nid hash position and ldlm reverse import.
952 Severity : enhancement
954 Description: Communicate OST degraded/readonly state via statfs to MDS
955 Details : Flags in the statfs returned from OSTs indicate whether the
956 OST is in a degraded RAID state, or if the filesystem has
957 turned read-only after a filesystem error is detected.
962 Description: don't panic if EPROTO was hit when reading symlink
963 Details : correctly handling request reference in error cases.
965 Severity : enhancement
967 Description: Add LustreNetLink, a kernel-userspace communcation path.
969 Severity : enhancement
971 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
976 Description: access to llog context before init.
977 Details : move handling CATALOGS file at osc layer and forbid access to llog
981 Frequency : rare, only when client is writing to OST on local node
983 Description: Avoid deadlock for local client writes
984 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
985 memory freeing context. This allows OST threads to set the
986 PF_MEMALLOC flag on task structures in order to allocate memory
987 from reserved pools and complete IO.
988 Use GFP_HIGHUSER for OST allocations for non-local client writes,
989 so that the OST threads generate memory pressure and allow
990 inactive pages to be reclaimed.
992 Severity : enhancement
994 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
999 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1000 Details : move ldlm namespace creation in setup phase to avoid grab
1001 _lprocfs_lock with cli_sem held.
1005 Description: Temporarily disable grant shrink.
1006 Details : Disable the feature for debugging.
1009 Frequency : only when formatting test filesystems
1011 Description: Unable to run several mkfs.lustre on loop devices at the same time
1012 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1013 formatting. The solution is to proper handle the error.
1015 Severity : enhancement
1017 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1019 Severity : enhancement
1021 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1026 Description: resolve race between obd_disconnect and class_disconnect_exports
1027 Details : if obd_disconnect will be called to already disconnected export he
1028 forget release one reference and osc module can't unloaded.
1030 Severity : enhancement
1032 Description: Allow tuning service thread via /proc
1033 Details : For each service a new
1034 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1035 created that can be used to set min/max thread counts, and get the
1036 current number of running threads.
1041 Descriptoin: don't return error if have partially created objects for file.
1042 Details : lov_update_create_set uses set->set_success as index for created
1043 objects, so if some requests will be failed, they will have hole
1044 at end of array and qos_shrink_lsm for allocate correct lsm.
1046 Severity : enhancement
1048 Description: Update OFED support to 1.4.1
1053 Description: Reduce small size read RPC
1054 Details : Set read-ahead limit for every file and only do read-ahead when
1055 available read-ahead pages are bigger than 1M to avoid small size
1058 Severity : enhancement
1060 Description: move AT tunable parameters for more consistent usage
1061 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1063 Severity : enhancement
1065 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1066 Details : allow skip disconnected ost for send statfs request and hide error
1072 Description: fix racy locking of mballoc block bitmaps causing BUG
1073 Details : The locking of the mballoc buddy bitmap and the in-memory
1074 block bitmap was using two different spin locks in some
1075 cases. This made it possible to incorrectly access the
1076 mballoc bitmap while another process was modifying it,
1077 causing a sanity assertion to fail. While no on-disk corruption
1078 was reported, there was some risk of this happening.
1081 Frequency : rare, on llog test 6
1083 Descriptoin: don't allow connect to already connected import
1084 Details : allowing connect to already connected import is hide connecting problem.
1089 Descriptoin: allow kill process which wait statahead result
1090 Details : for some reasons 'ls' can stick in waiting result from statahead,
1091 in this case need way for kill this process.
1093 Severity : enhancement
1095 Description: Add state history info file, enhance import info file
1096 Details : Track import connection state changes in a new osc/mdc proc file;
1097 add overview-type data to the osc/mdc import proc file.
1099 Severity : enhancement
1101 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1103 Severity : enhancement
1105 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1110 Descriptoin: OST grant too much space to client even there are not enough space.
1111 Details : Client will shrink its grant cache to OST if there are no write
1112 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1113 this grant cache if there are already not enough avaible space
1114 (left_space < total_clients * 32M).
1117 Frequency : start MDS on uncleanly shutdowned MDS device
1119 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1120 Details : stay in waiting mds<>ost recovery finished produce random bugs
1121 due race between two ll_sync thread for one lov target. send
1122 ACTIVATE event only if connect realy finished and import have
1126 Frequency : rare, connect and disconnect target at same time
1128 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1129 Details : don't call obd_disconnect under lov_lock. this long time
1130 operation and can block ptlrpcd which answer to connect request.
1135 Descriptoin: don't lose wakeup for imp_recovery_waitq
1136 Details : recover_import_no_retry or invalidate_import and import_close can
1137 both sleep on imp_recovery_waitq, but we was send only one wakeup
1141 Frequency : always with long access acl
1143 Descriptoin: mds can't pack reply with long acl.
1144 Details : mds don't control size of acl but they limited by reint/getattr
1147 Severity : enhancement
1149 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1151 Severity : enhancement
1153 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1156 Frequency : start MDS on uncleanly shutdowned MDS device
1158 Descriptoin: aborting recovery hang on MDS
1159 Details : don't throttle destroy RPCs for the MDT.
1162 Frequency : on remount
1164 Description: external journal device not working after the remount
1165 Details : clear dev_rdonly flag for external journal devices in
1171 Description: shutdown vs evict race
1172 Details : client_disconnect_export vs connect request race.
1173 if client will evicted at this time - we start invalidate
1174 thread without referece to import and import can be freed
1178 Frequency : rare, need acl's on inode.
1180 Description: client can't handle ost additional correctly
1181 Details : if ost was added after client connected to mds client can have
1182 hit lnet_try_match_md ... to big messages to wide striped files.
1183 in this case need teach client to handle config events about add
1184 lov target and update client max ea size at that event.
1186 Severity : enhancement
1188 Description: Changelogs
1189 Details : Changelogs are a lightweight mechanism to track filesystem
1190 metadata and namespace changes. The changelog is recorded
1191 permanently on the MDTs, and is periodically "consumed" / purged
1192 when records are no longer needed.
1194 Severity : enhancement
1196 Description: compact fld format with extents
1197 Details : Store range of seq rather than every seq in FLD. Seq
1198 controller update FLD rather than clients. In Case of CMD, mdt0
1199 has FLD, all other metadata server act as non persistent proxy
1200 for FLD queries and cache fld entries in fld cache.
1205 Description: don't skip ost target if they assigned to file
1206 Details : Drop slow OSCs if we can, but not for requested start idx.
1207 This means "if OSC is slow and it is not the requested
1208 start OST, then it can be skipped, otherwise skip it only
1209 if it is inactive/recovering/out-of-space.
1213 Description: more cleanup in mds_lov
1214 Details : not send LOV EA under replay, we can't know about they size at this
1215 time. Don't allow client connect to mds before any ost connected,
1216 for avoid problems with LOV EA size and returning EIO to client.
1218 Severity : enhancement
1220 Description: Interoperability at server side (Disk interoperability)
1222 Severity : enhancement
1224 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1226 Severity : enhancement
1228 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1230 Severity : enhancement
1232 Description: New client IO stack (CLIO).
1234 Severity : enhancement
1236 Description: Commit on sharing. Eliminate inter-client dependencies between
1237 uncommitted transactions by doing transaction commits.
1238 Thereby clients may recovery independently.
1241 Frequency : Create a symlink file with a very long name
1243 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1244 Details : If there is no extra space in the request for early cancels,
1245 ldlm_req_handles_avail() returns 0 instead of a negative value.
1247 Severity : enhancement
1249 Description: Add /proc entry for import status
1250 Details : The mdc, osc, and mgc import directories now have
1251 an import directory that contains useful import data for debugging
1252 connection problems.
1254 Severity : enhancement
1256 Description: Re-disable certain /proc logging
1257 Details : Enable and disable client's offset_stats, extents_stats and
1258 extents_stats_per_process stats logging on the fly.
1261 Frequency : Only on FC kernels 2.6.22+
1263 Description: oops in statahead
1264 Details : Do not drop reference count for the dentry from VFS when lookup,
1265 VFS will do that by itself.
1267 Severity : enhancement
1269 Description: Generic /proc file permissions
1270 Details : Set /Proc file permissions in a more generic way to enable non-
1271 root users operate on some /proc files.
1275 Description: Hitting mdc_commit_close() ASSERTION
1276 Details : Properly handle request reference release in
1277 ll_release_openhandle().
1281 Description: quota recovery deadlock during mds failover
1282 Details : This patch includes att18982, att18236, att18237 in bz14840.
1284 1. fix osts hang when mds does failover with quotaon
1285 2. prevent watchdog storm when osts threads wait for the
1290 Frequency : only patchless client
1291 Description: add workaround for race between add/remove dentry from hash
1293 Severity : enhancement
1295 Description: Allow OST glimpses to return PW locks
1299 Description: LBUG when llog conf file is full
1300 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1304 Description: Prevent import from entering FULL state when server in recovery
1308 Description: service mount cannot take device name with ":"
1309 Details : Only when device name contains ":/" will mount treat it as
1315 Description: replace ptlrpcd with the statahead thread to interpret the async
1316 statahead RPC callback
1320 Frequency : on recovery
1321 Description: I/O failures after umount during fail back
1322 Details : if client reconnected to restarted server we need join to recovery
1323 instead of find server handler is changed and process self
1324 eviction with cancel all locks.
1326 Severity : enhancement
1328 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1330 Severity : enhancement
1332 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1334 Severity : enhancement
1336 Description: Upcall on Lustre log has been dumped
1337 Details : Allow for a user mode script to be called once a Lustre log has
1338 been dumped. It passes the filename of the dumped log to the
1339 script, the location of the script can be specified via
1340 /proc/sys/lnet/debug_log_upcall.
1345 Description: avoid idr_remove called for id which is not allocated.
1346 Details : Move assigment s_dev for clustered nfs to end of initialization,
1347 to avoid problem with error handling.
1352 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1353 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1354 into CONN_USED_HASH and this prodice warning when put connection
1355 again in unused hash.
1358 Severity : enhancement
1360 Description: File striping can now be set to use an arbitrary pool of OSTs.
1362 Severity : enhancement
1364 Description: Export bytes_read/bytes_write count on OSC/OST.
1368 Description: Early reply size mismatch, MGC loses connection
1369 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1370 the connect flags are properly negotiated.
1375 Description: quotas are not honored with O_DIRECT
1376 Details : all writes with the flag O_DIRECT will use grants which leads to
1377 this problem. Now using OBD_BRW_SYNC to guard this.
1381 Description: add quota statistics
1382 Details : 1. sort out quota proc entries and proc code.
1383 2. add quota statistics
1385 Severity : enhancement
1387 Description: enable quota support for HEAD.
1391 Description: Properly propagate oinfo flags from lov to osc for statfs
1392 Details : restore missing copy oi_flags to lov requests.
1394 Severity : enhancement
1396 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1397 : l_getgroups(8), lst(8), routerstat(8)
1398 Details : included man pages for llobdstat(8), llstat(8),
1399 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1401 Severity : enhancement
1403 Description: configure's --enable-quota should check the
1404 : kernel .config for CONFIG_QUOTA
1405 Details : configure is terminated if --enable-quota is passed but
1406 : no quota support is in kernel
1410 Description: Remove portals compatibility
1411 Details : Remove portals compatibility, not interoperable with releases
1416 Description: Resolve device initialization race
1417 Details : Prevent proc handler from accessing devices added to the
1418 obd_devs array but yet be intialized.
1420 Severity : enhancement
1422 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1424 Severity : enhancement
1426 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1431 Description: Using wrong pointer in osc_brw_prep_request
1432 Details : Access to array[-1] can produce panic if kernel compiled with
1433 CONFIG_PAGE_ALLOC enabled
1437 Description: Client runs out of low memory
1438 Details : Consider only lowmem when counting initial number of llap pages
1442 Description: *optional* service tags registration
1443 Details : if the "service tags" package is installed on a Lustre node
1444 When the filesystem is mounted, a local-node service tag will
1445 be created. See http://inventory.sun.com/ for more information
1446 about the Service Tags asset management system.
1450 Description: Kernel BUG tries to release flock
1451 Details : Lustre does not destroy flock lock before last reference goes
1452 away. So always drop flock locks when client is evicted and
1453 perform unlock regardless of successfulness of speaking to MDS.
1457 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1461 Description: sanity test 65a fails if stripecount of -1 is set
1462 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1467 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1468 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1471 Severity : enhancement
1473 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1477 Description: do not process already freed flock
1478 Details : flock can possibly be freed by another thread before it reaches
1479 to ldlm_flock_completion_ast.
1483 Description: LBUG during stress test
1484 Details : Need properly lock accesses the flock deadlock detection list.
1488 Description: oops in page fault handler
1489 Details : kernel page fault handler can return two special 'pages' in error
1490 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1494 Description: timeout with invalidate import.
1495 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1496 should be handled by ptlrpcd. This produce long age waiting and
1497 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1499 Severity : enhancement
1501 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1505 Description: enable MGS and MDT services start separately
1506 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1507 a co-located MGS without starting the MGS, which is a complement
1508 to 'nosvc' mount option.
1512 Frequency : after recovery
1513 Description: precreate to many object's after del orphan.
1514 Details : del orphan st in oscc last_id == next_id and this triger growing
1515 count of precreated objects. Set flag LOW to skip increase count
1516 of precreated objects.
1520 Frequency : rare, on clear nid stats
1521 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1522 Details : when clean nid stats sometimes try destroy live entry,
1523 and this produce panic in free.
1527 Description: Stack overflow during MDS log replay
1528 ease stack pressure by using a thread dealing llog_process.
1532 Description: wait until IO finished before start new when do lock cancel.
1533 Details : VM protocol want old IO finished before start new, in this case
1534 need wait until PG_writeback is cleared until check dirty flag and
1535 call writepages in lock cancel callback.
1537 Severity : enhancement
1539 Description: using special macro for print time and cleanup in includes.
1543 Description: mds_mfd_close() ASSERTION(rc == 0)
1544 Details : In mds_mfd_close(), we need protect inode's writecount change
1545 within its orphan write semaphore to prevent possible races.
1549 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1553 Frequency : rare, on shutdown ost
1554 Description: don't hit live lock with umount ost.
1555 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1556 use shrink_dcache_sb instead.
1560 Description: don't panic with use echo client
1561 Details : echo client pass NULL as client nid pointer and this produce null
1562 pointer dereference.
1566 Description: fix build on ppc32
1567 Details : compile code with -m64 flag produce wrong object file for ppc32.
1571 Description: add message levels for liblustreapi
1575 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1576 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1577 will result in an -EAGAIN and in the exit call path we call
1578 alloc_rr() which will with increasing aggressiveness attempt to
1579 aquire precreated objects on the minimum number of required OSCs.
1583 Description: Use old size assignment to avoid deadlock
1584 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1585 the scheduling while holding a spinlock. We do not need locking
1586 for size in ll_update_inode() because size is only updated from
1587 the MDS for directories or files without objects, so there is no
1588 other place to do the update, and concurrent access to such inodes
1589 are protected by the inode lock.
1593 Description: resolve "_IOWR redefined" build error on SLES10
1597 Description: dump the memory debugging after all modules are unloaded to
1598 suppress false negative in conf_sanity test 39
1600 Severity : enhancement
1602 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1607 Description: fix SLES kernel versioning
1608 Details : the kernel version for our SLES 10 kernel did not include a "-"
1609 before the "smp" at the end. while this was not a problem in
1610 general, it did mean that software trying to use the kernel
1611 version to try to detect a vendor specific kernel would fail.
1612 this was most evident by the OFED build scripts.
1616 Description: Don't update lov_desc members until making sure they are valid
1617 Details : When updating lov_desc members via proc fs, need fix their
1618 validities before doing the real update.
1622 Description: don't put request into delay list while invalidate in flight.
1623 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1624 import in flight. this produce timeout for invalidate and sometimes
1625 can cause stale data.
1629 Frequency : on ppc only
1630 Description: not convert ost objects for directory because it's not exist.
1631 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1635 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1637 Severity : enhancement
1639 Description: Update to support 2.6.22.14 vanilla kernel.
1643 Frequency : rare, on recovery
1644 Description: read procfs can produce deadlock in some situation
1645 Details : Holding lprocfs lock which send rpc can produce block for destroy
1646 obd objects and this also block reconnect with -EALREADY. This
1647 isn't fix all lprocfs bugs - but make it rare.
1649 Severity : enhancement
1651 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1654 Frequency : frequent on X2 node
1656 Description: mdc_set_open_replay_data LBUG
1657 Details : Set replay data for requests that are eligible for replay.
1661 Description: lustre_mgs: operation 101 on unconnected MGS
1662 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1663 MGC, and late on MGC cannot successfully connect to MGS and a lot
1664 of the error messages complaining that MGS is not connected.
1668 Frequency : on network error
1669 Description: panic with double free request if network error
1670 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1671 but it's true only for synchronus enqueue, for async enqueue
1672 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1675 Severity : enhancement
1677 Description: client-side metadata stat-ahead during readdir(directory readahead)
1678 Details : perform client-side metadata stat-ahead when the client detects
1679 readdir and sequential stat of dir entries therein
1682 Frequency : on start mds
1684 Description: Implement get_info(last_id) in obdfilter.
1687 Frequency : occasional
1689 Description: Correctly check stale fid, not start epoch if ost not support SOM
1690 Details : open with flag O_CREATE need set old fid in op_fid3 because
1691 op_fid2 was overwritten with new generated fid, but mds can answer
1692 with one of these two fids and both is not stale. Setattr
1693 incorrectly started an epoch and assume will be called
1694 done_writing, but without SOM done_writing ever being called.
1697 Frequency : rare, depends on device drivers and load
1699 Description: MDS or OSS nodes crash due to stack overflow
1700 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1701 In some cases, in conjunction with device drivers that use a lot
1702 of stack the MDS (or possibly OSS) service threads could overflow
1703 the stack. One change which was identified to consume additional
1704 stack has been reworked to avoid the extra stack usage.
1707 Frequency : occasional
1709 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1710 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1711 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1712 oscc_has_objects and see OSC has no objests and call
1713 oscc_internal_create to resend create request.
1715 Severity : enhancement
1717 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1719 Severity : enhancement
1721 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1724 Frequency : very rare
1726 Description: Fix replay if there is an un-replied request and open
1727 Details : In some cases, older replay request will revert the
1728 mcd->mcd_last_xid on MDS which is used to record the client's
1729 latest sent request.
1731 Severity : enhancement
1733 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1735 Severity : enhancement
1737 Description: Add rhel5 support to HEAD.
1739 Serverity : enhancement
1741 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1746 Description: Don't allow skipping OSTs if index has been specified.
1747 Details : Don't allow skipping OSTs if index has been specified, make
1748 locking in internal create lots better.
1752 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1753 Details : access to bitfield in structure is always rounded to long
1754 and this produce problem with not atomic change any bit.
1758 Description: Lustre make rpms failed.
1759 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1760 builds Lustre rpms from tarball.
1763 Frequency : rare on shutdown ost
1765 Description: If llog cancel was not send before clean_exports phase, this can
1766 produce deadlock in llog code.
1767 Details : If llog thread has last reference to obd and call class_import_put
1768 this produce deadlock because llog_cleanup_commit_master wait when
1769 last llog_commit_thread exited, but this never success because was
1770 called from llog_commit_thread.
1774 Description: allow userland application know is lost one of stripes.
1775 Details : fill lvb_blocks with error code on ost and return it to
1776 application if error flag found.
1780 Description: NULL lov_tgts causing MDS oops
1781 Details : more safe checks for NULL lov_tgts for avoid oops.
1783 Severity : enhancement
1785 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1789 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1790 Details : If an OST is down the MDS will hang indefinitely in
1791 obd_statfs_rqset() waiting for the statfs data. While for
1792 MDS QOS usage of statfs, it should not stuck in waiting.
1794 Severity : enhancement
1796 Description: remote_acl support
1797 Details : Support ACL-based permission check for remote user.
1798 Support setfacl/getfacl for remote user with the utils
1799 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1800 the system "{s,g}etfacl" utils.
1802 Severity : enhancement
1804 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1806 Severity : enhancement
1808 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1812 Description: make mgs_setparam() handle fsname containing dash
1813 Details : fsname containing a dash does not work with lctl conf_param
1815 Severity : enhancement
1817 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1819 Severity : enhancement
1821 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1825 Frequency : rare, at shutdown
1826 Description: access already free / zero obd_namespace.
1827 Details : if client_disconnect_export was called without force flag set,
1828 and exist connect request in flight, this can produce access to
1829 NULL pointer (or already free pointer) when connect_interpret
1830 store ocd flags in obd_namespace.
1834 Frequency : only at startup
1835 Description: not alloc memory with spinlock held.
1836 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1837 if any spinlock held.
1839 Severity : enhancement
1841 Description: make lustre randomly fail allocating memory
1842 Details : Make lustre randomly failed allocating memory for testing purpose.
1844 Severity : enhancement
1846 Description: lost problems with lov objid file
1847 Details : Fixes some scability and access to not inited memory problems
1848 in work with lov objdid file.
1853 Description: lfs find does not continue on file error
1854 Details : Continue other files processing when a file/dir is absent.
1858 Description: Inconsistent usage of lustre_pack_reply()
1859 Details : Standardize the usage of lustre_pack_reply() such that it
1860 always generate a CERROR on failure.
1863 Frequency : occasional
1865 Description: MDS hang or stay in waiting lock
1866 Details : If client receive lock with CBPENDING flag ldlm need send lock
1867 cancel as separate rpc, to avoid situation when cancel request
1868 can't processed due all i/o threads stay in wait lock.
1872 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1876 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1880 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1884 Description: Update to RHEL4 latest kernel.
1886 Severity : enhancement
1888 Description: Build SLES10 patchless client fails
1889 Details : The configure was broken by run ./configure with
1890 --with-linux-obj=.... argument for patchless client. When the
1891 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1892 can't search header adequately. Use absolute path such as
1893 -I($LINUX)/include instead.
1897 Description: interrupt oig_wait produce painc on resend.
1898 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1899 private set, and this produce situation when rq_ptlrpcd_data not
1900 copyed to new allocated request and triggered LBUG on assert
1901 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1902 ptlrpcd set if request is changed and can be safety changed to use
1907 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1908 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1909 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1910 these macros and replace them with their difinitions instead.
1914 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1915 Details : Update which_patch & target file for SLES10 latest kernel.
1918 Description: Remove client patches from SLES 10 kernel.
1919 Details : This causes SLES 10 clients to behave as patchless clients
1920 even on a Lustre-patched (server) kernel.
1922 Severity : enhancement
1924 Description: self-adjustable client's lru lists
1925 Details : use adaptive algorithm for managing client cached locks lru
1926 lists according to current server load, other client's work
1927 pattern, memory activities, etc. Both, server and client
1928 side namespaces provide number of proc tunables for controlling
1931 Severity : enhancement
1933 Description: light-weight GSS support
1934 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1935 and reduce performance overhead.
1937 Severity : enhancement
1939 Description: Linux keyring support
1940 Details : Support using service of Linux keyring for Lustre GSS internal
1941 context refresh/cache mechanism.
1945 Description: Fix errors in lfs documentation
1946 Details : Fixes man pages
1950 Description: don't use GFP_* in generic Lustre code.
1951 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1955 Description: obdclass is limited by single OBD_ALLOC(idarray)
1956 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1960 Description: warnings with build patchless client with vanila 2.6.19 and up
1961 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1962 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1966 Frequency : Only for SLES
1967 Description: sanity_quota fail test_1
1968 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1969 file, which makes TSTID[2] inunique.
1973 Frequency : Only for SLES9
1974 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1975 Details : Update target/ChangeLog/which_patch .
1979 Frequency : Only for test_10 in sanity-quota.sh
1980 Description: a bug in quota test code
1981 Details : checking a test flag in a wrong place
1985 Frequency : Only for test_13 in sanity-quota.sh
1986 Description: a bug in quota test script
1987 Details : shouldn't check overfull allocated quota
1991 Description: jbd statistics
1992 Details : Port older jbd statistics patch for sles10
1994 Severity : enhancement
1996 Description: Add printing inode info into message about error in writepage.
2001 Description: reply_lock_interpret crash due to race with it and lock cancel.
2002 Details : Do not replay locks that are being cancelled. Do not reference
2003 locks by their address during replay, just by their handle.
2007 Frequency : When flocks are used.
2008 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2009 Details : Flock locks might destroy just granted lock if it could be merged
2010 with another existing flock, this is done in completion handler,
2011 so teach ldlm_cli_enquque_fini that this is a valid case for
2017 Description: Oops in read and write path when failing to allocate lock.
2018 Details : Check if lock allocation failed and return error back.
2022 Description: lstripe command fails for valid OST index
2023 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2024 instead of lov->desc.ld_active_tgt_count.
2028 Description: sanity.sh failed test 103
2029 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2034 Description: df doesn't work properly if diskfs blocksize != 4K
2035 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2039 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2040 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2042 Severity : enhancement
2044 Description: Adaptive timeouts
2045 Details : RPC timeouts adapt to changing server load and network
2046 conditions to reduce resend attempts and improve recovery time.
2050 Description: llapi_file_create() does not allow some changes
2051 Details : add llapi_file_open() that allows specifying the mode and
2052 open flags, and also returns an open file handle.
2056 Description: merge and cleanup kernel patches.
2057 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2061 Description: Add journal checksum support.(Kernel part)
2062 Details : The journal checksum feature adds two new flags i.e
2063 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2064 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2065 indicates that the commit block contains the checksum for
2066 the blocks described by the descriptor blocks. Now commit
2067 record can be sent to disk without waiting for descriptor
2068 blocks to be written to disk. This behavior is controlled
2069 using JBD2_FEATURE_ASYNC_COMMIT flag.
2073 Description: OSS needs mutliple precreate threads
2074 Details : Add ability to start more than one create thread per OSS.
2078 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2079 Details : execute lfs setstripe on client
2083 Description: mds_obd_create error creating tmp object
2084 Details : When the user sets quota on root, llog will be affected and can't
2085 create files and write files.
2088 Frequency : Always on ia64 patchless client, and possibly others.
2090 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2091 Details : This allows the patchless client to be loaded on architectures
2092 without this export.
2096 Description: RedHat Update kernel for RHEL5
2097 Details : Add the kernel config file for RHEL5.
2101 Description: RedHat Update kernel for RHEL5
2102 Details : Modify the kernel config file more closer RHEL5.
2106 Description: Build failure against Centos5 (RHEL5)
2107 Details : Define PAGE_SIZE when it isn't present.
2111 Description: test_7 on sanity-quota.sh can't be used on separate machines
2112 Details : add facet to handle it
2116 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2117 Details : with NFS, the anon dentry's parent was set to itself in
2118 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2119 corresponding dentry other than use rec->ur_name.
2121 Severity : enhancement
2123 Description: lfs setstripe enhancement
2124 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2128 Description: enable data checksumming by default
2129 Details : enable checksum by default, allow --disable-checksum
2130 configure option and "-o nochecksum" mount option. Checksums
2131 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2135 Description: lustre support for RHEL5
2136 Details : Add support for RHEL5.
2140 Description: Client eviction due to failover config
2141 Details : after a connection loss, the lustre client should attempt to
2142 reconnect to the last active server first before trying the
2143 other potential connections.
2147 Description: when mds and osts use different quota unit(32bit and 64bit),
2148 quota will be released repeatly.
2149 Details : void sending multiple quota reqs to mds, which will keep the
2150 status between the reqs.
2154 Description: osts not allocated evenly to files
2155 Details : change the condition to increase offset_idx
2158 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2160 Description: Data corruption for OSTs that are formatted larger than 2TB
2162 Details : When generating the bio request for lustre file writes the
2163 sector number would overflow a temporary variable before being
2164 used for the IO. The data reads correctly from Lustre (which
2165 will overflow in a similar manner) but other file data or
2166 filesystem metadata may be corrupted in some cases.
2170 Description: Tune the kernel for good SCSI performance.
2171 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2172 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2177 Description: rewrite ext2-derived code in obdclass/uuid.c
2178 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2179 from scratch preserving functionality.
2183 Description: "lfs find -obd UUID" prints directories
2184 Details : "lfs find -obd UUID" will return all directory names instead
2185 of just file names. It is incorrect because the directories
2186 do not reside on the OSTs.
2190 Description: use i_size_read and i_size_write in 2.6 port
2191 Details : replace inode->i_size access with i_size_read/write()
2194 Frequency : when removing large files
2196 Description: scheduling issue during removal of large Lustre files
2197 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2198 It causes scheduling issues when removing large files (17TB in the
2202 Frequency : only with liblustre clients on XT3
2204 Description: evictions taking too long
2205 Details : allow llrd to evict clients directly on OSTs
2208 Frequency : only on ppc
2210 Description: /proc/fs/lustre/devices broken on ppc
2211 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2212 We should make sure the type of 'index' is loff_t and then cast
2213 explicitly as needed below. Do not assign an explicitly cast
2217 Frequency : only for rhel5
2219 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2220 Details : Modify the target file & which_kernel.
2222 Severity : enhancement
2224 Description: omit set fsid for export NFS
2225 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2226 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2230 Frequency : Always, for kernels after 2.6.16
2231 Description: Fix warning idr_remove called for id=.. which is not allocated.
2232 Details : Last kernels save old s_dev before kill super and not allow
2233 to restore from callback - restore it before call kill_anon_super.
2237 Description: buffer overruns could theoretically occur
2238 Details : llapi_semantic_traverse() modifies the "path" argument by
2239 appending values to the end of the origin string, and a
2240 overrun may occur. Adding buffer overrun check in liblustreapi.
2244 Description: Fix error on 'ls .' at the top of the Lustre mount.
2245 Details : Don't revalidate dentry if it is a root dentry.
2249 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2250 Details : Modify vm-tunables-rhel4.patch.
2254 Description: Kernel config for 2.6.18-vanilla.
2255 Details : Modify targets/2.6-vanilla.target.in.
2256 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2257 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2258 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2259 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2263 Description: improve handling recoverable errors
2264 Details : if request processig with error which can be recoverable on server
2265 request should be resend, otherwise page released from cache and
2270 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2271 Details : Modify target file & which_patch.
2272 A flaw was found in the IA32 system call emulation provided
2273 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2274 value could be stored in the %RAX register, which could trigger an
2275 out-of-bounds system call table access. An untrusted local user
2276 could exploit this flaw to run code in the kernel
2277 (ie a root privilege escalation). (CVE-2007-4573).
2281 Description: O_DIRECT bypasses client statistics.
2282 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2283 read_ahead_stats were updated. copied the stats section from
2284 osc_send_oap_rpc() into async_internal().
2288 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2292 Description: change order of libsysio includes
2293 Details : '#include sysio.h' should always come before '#include xtio.h'
2297 Description: handle "serious error: objid * already exists" more gracefully
2298 Details : If LAST_ID value on disk is smaller than the objects existing in
2299 the O/0/d* directories, it indicates disk corruption and causes an
2300 LBUG(). If the object is 0-length, then we should use the existing
2301 object. This will help to avoid a full fsck in most cases.
2303 Severity : enhancement
2305 Description: adapt the lustre_config script to support the upgrade case
2306 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2307 server targets to 1.6 in parallel.
2311 Description: To avoid grant space > avaible space when the disk is almost
2312 full. Without this patch you might see the error "grant XXXX >
2313 available" or some LBUG about grant, when the disk is almost
2315 Details : In filter_check_grant, for non_grant cache write, we should
2316 check the left space by if (*left > ungranted + bytes), instead
2317 of (*left > ungranted), because only we are sure the left space
2318 is enough for another "bytes", then the ungrant space should be
2319 increase. In client, we should update cl_avail_grant only there
2320 is OBD_MD_FLGRANT in the reply.
2324 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2325 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2326 described in CVE-2007-4573. This problem could allow untrusted
2327 local users to gain root access.
2330 Frequency : when using O_DIRECT and quotas
2332 Description: Incorrect file ownership on O_DIRECT output files
2333 Details : block usage reported by 'lfs quota' does not take into account
2334 files that have been written with O_DIRECT.
2339 Description: touch file failed when fs is not full
2340 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2341 otherwise we can get ENOSP while fs is not full.
2345 Description: parallel lock callbacks
2346 Details : Instead of sending blocking and completion callbacks as separated
2347 requests, adding them to a set and sending in parallel.
2350 Frequency : only for Cray XT3
2351 Bugzilla : 12829/13455
2352 Description: Changing primary group doesn't change the group lustre assigns to
2354 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2355 overridden with the primary group provided by the group upcall,
2356 whereas the supplied fsgid can be trusted if it is in the list of
2357 supplementary groups returned by the group upcall.
2359 Severity : enhancement
2361 Description: Allow masking D_WARNING, D_ERROR messages from console
2362 Details : Console messages can now be disabled via lnet.printk.
2366 Description: User code with malformed file open parameter crashes client node
2367 Details : Before packing join_file req, all the related reference should be
2368 checked carefully in case some malformed flags cause fake
2369 join_file req on client.
2373 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2374 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2375 mode is equal to the granted mode. However, it should also check
2376 that LDLM_FL_CP_REQD is not set.
2380 Description: Heavy nfs access might result in deadlocks
2381 Details : After ELC code landed, it is now improper to enqueue any mds
2382 locks under och_sem, because enqueue might want to decide to
2383 cancel open locks for same inode we are holding och_sem for.
2387 Description: Client eviction while running blogbench
2388 Details : A lot of unlink operations with concurrent I/O can lead to a
2389 deadlock causing evictions. To address the problem, the number of
2390 oustanding OST_DESTROY requests is now throttled to
2391 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2392 callbacks are processed in priority.
2396 Description: enable ACLs on MDS by default
2397 Details : ACLs must be enabled on MDS by default.
2400 Frequency : PPC/PPC64 only
2402 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2403 Details : fix duplicated definitions between asm-ppc64/types.h and
2404 lustre_types.h on PPC.
2407 Frequency : PPC/PPC64 only
2409 Description: asm-ppc/segment.h does not exist
2410 Details : fix compile issue on PPC.
2414 Description: better handle error messages in extents code
2417 Frequency : RHEL4 only
2419 Description: mkfs is very slow on IA64/RHEL4
2420 Details : A performance regression has been discovered in the MPT Fusion
2421 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2422 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2423 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2425 Severity : enhancement
2427 Description: SNMP support enhancement
2428 Details : Adding total number of sampled request for an MDS node in snmp
2431 Severity : enhancement
2433 Description: Optimize ldlm waiting list processing for PR extent locks
2434 Details : When processing waiting list for read extent lock and meeting read
2435 lock that is same or wider to it that is not contended, skip
2436 processing rest of the list and immediatelly return current
2437 status of conflictness, since we are guaranteed there are no
2438 conflicting locks in the rest of the list.
2442 Description: Time out and refuse to reconnect
2443 Details : When the failover node is the primary node, it is possible
2444 to have two identical connections in imp_conn_list. We must
2445 compare not conn's pointers but NIDs, otherwise we can defeat
2446 connection throttling.
2450 Description: port llog fixes from b1_6 into HEAD
2451 Details : Port llog reference couting and some llog cleanups from b1_6
2452 (bug 10800) into HEAD, for protect from panic and access to already
2453 free llog structures.
2457 Description: Detect stride IO mode in read-ahead
2458 Details : When a client does stride read, read-ahead should detect that and
2459 read-ahead pages according to the detected stride pattern.
2463 Description: data checksumming impacts single node performance
2464 Details : add support for several checksum algorithm. Currently, only CRC32
2465 and Adler-32 are supported. The checksum type can be changed on
2466 the fly via /proc/fs/lustre/osc/*/checksum_type.
2470 Description: use adler32 for page checksums
2471 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2476 Description: build for x2 fails
2477 Details : fix compile issue on Cray systems.
2481 Description: Properly match for duplicate locks
2482 Details : Due to different lock order from skiplists code, we need to
2483 traverse entire list for now
2486 Frequency : only on PPC/SLES10
2488 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2489 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2490 be defined. Add a hack in mkfs_lustre.c to work around this compile
2495 Description: LASSERT on MDS when client holding flock lock dies
2496 Details : ldlm pool logic depends on number of granted locks equal to
2497 number of released locks which is not true for flock locks, so
2498 just exclude such locks from consideration.
2502 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2503 Details : Use fsfilt_sync() for both the whole filesystem sync and
2504 individual file sync to eliminate dangerous inode locking
2505 with I_LOCK that can lead to a deadlock.
2509 Description: performance in 1.6.3
2510 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2511 sure that 1MB requests are not fragmented by the block layer.
2513 Severity : enhancement
2515 Description: organize the server-side client stats on per-nid basis
2516 Details : Change the structure of stats under obdfilter and mds to
2524 The "uuid"s file would list the uuids of _active_ exports.
2525 And the clear entry is to clear all stats and stale nids.
2527 Severity : enhancement
2529 Description: eliminate client locks in face of contention
2530 Details : file contention detection and lockless i/o implementation
2531 for contended files.
2535 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2538 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2540 Description: MDS may lose file striping (and hence file data) in some cases
2541 Details : If there are additional extended attributes stored on the MDS,
2542 in particular ACLs, SELinux, or user attributes (if user_xattr
2543 is specified for the client mount options) then there is a risk
2544 of attribute loss. Additionally, the Lustre file striping
2545 needs to be larger than default (e.g. striped over all OSTs),
2546 and an additional attribute must be stored initially in the
2547 inode and then increase in size enough to be moved to the
2548 external attribute block (e.g. ACL growing in size) for file
2553 Description: skiplist implementation simplification
2554 Details : skiplists are used to group compatible locks on granted list
2555 that was implemented as tracking first and last lock of each
2556 lock group the patch changes that to using doubly linked lists
2560 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2561 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2562 because LASSERT on both the data supplied by a client, and the
2563 data on disk is dangerous and incorrect.
2565 Severity : enhancement
2567 Description: Slow truncate/writes to huge files at high offsets.
2568 Details : Directly associate cached pages to lock that protect those pages,
2569 this allows us to quickly find what pages to write and remove
2570 once lock callback is received.
2574 Description: more ldlm soft lockups
2575 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2576 starve other threads from the resource lock for a long time in
2577 case of long waiting queue, so change the debug level from
2578 D_OTHER to the less frequently used D_INFO.
2580 Severity : enhancement
2582 Description: add -gid, -group, -uid, -user options to lfs find
2586 Description: Hung threads in invalidate_inode_pages2_range
2587 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2588 once one is completed. As a result, some RPCs are stuck in the
2589 queue and are never sent.
2593 Description: filter threads hungs on waiting journal commit
2594 Details : Cleanup filter group llog code, then only filter group llog will
2595 be only created in the MDS/OST syncing process.
2599 Description: Procfs and llog threads access destoryed import sometimes.
2600 Details : Sync the import destoryed process with procfs and llog threads by
2601 the import refcount and semaphore.
2603 Severity : enhancement
2605 Description: openlock cache of b1_6 port to HEAD
2610 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2611 Details : If insertion of an extent fails, then discard the inode
2612 preallocation and free data blocks else it can lead to duplicate
2617 Description: don't always update ctime in ext3_xattr_set_handle()
2618 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2619 In some cases the ctime should not be updated, for example for
2620 2.0->1.8 compatibility it is necessary to delete an xattr and it
2621 should not update the ctime.
2625 Bugzilla : 15713/16362
2626 Description: Assertion in iopen_connect_dentry in 1.6.3
2627 Details : looking up an inode via iopen with the wrong generation number can
2628 populate the dcache with a disconneced dentry while the inode
2629 number is in the process of being reallocated. This causes an
2630 assertion failure in iopen since the inode's dentry list contains
2631 both a connected and disconnected dentry.
2635 Description: assertion failure in ldlm_handle2lock()
2636 Details : fix a race between class_handle_unhash() and class_handle2object()
2637 introduced in lustre 1.6.5 by bug 13622.
2642 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2643 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2648 Description: oops in sd_iostats_seq_show()
2649 Details : unloading/reloading the scsi low level driver triggers a kernel
2650 bug when trying to access the sd iostat file.
2655 Description: Kernel panics during QLogic driver reload
2656 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2657 patch, causing memory corruption.
2662 Description: journal_dev option does not work in b1_6
2663 Details : pass mount option during pre-mount.
2665 Severity : enhancement
2667 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2668 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2669 extent information of a file. It can be used to map logical blocks
2670 in a file to physical blocks in the block device.
2674 Description: LDLM soft lockups - improvement
2675 Details : It is be possible to send the lock handle along with each read
2676 or write request because the client is already doing a lock match
2677 itself so there isn't any reason the OST should have to re-do that
2681 Frequency : only on Cray X2
2683 Description: X2 build failures
2684 Details : fix build failures on Cray X2.
2688 Description: xid & resent requests
2689 Details : Initialize RPC XID from clock at startup (randomly if clock is
2692 Severity : enhancement
2694 Description: Add lustre_start utility to start or stop multiple Lustre servers
2699 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2700 Details : In case of memory pressure, list_del() can be called twice on
2701 req->rq_history_list, causing a kernel oops.
2705 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2706 Details : fix stack overflow in the distributed lock manager by defering
2707 export eviction after a failed AST to the elt thread instead of
2708 handling it in the dlm interpret routine.
2712 Description: Convert some comments to new format.
2713 Details : Update documenting comments to match doxygen conventions.
2717 Description: Grammar fixes.
2718 Details : A couple of trivial spelling fixes.
2722 Description: OSD_COUNTERS-mandatory
2723 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2724 to simplify the code.
2728 Description: simplify cmm_device freeing logic.
2729 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2733 Description: Add lockdep support to dt_object_operations locking interface.
2734 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2735 parameter indicating lock ordering. Update mdd code to use new
2740 Description: Introduce failloc constants for lockless IO tests.
2741 Details : Add two new failloc constants to test lockless IO. Only one of
2742 them in implemented---another is checked in yet to be landed
2747 Description: Add lockdep support for inode mutex.
2748 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2749 in the situations where Lustre has to lock more than one inode
2754 Description: Add optional invariants checking support.
2755 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2756 configure switch. This macro is to be used for consistency and
2757 sanity checks that are too expensive to be left in `production'
2762 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2763 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2768 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2769 Details : Use spin_lock_nested() in (the only) situation where more than
2770 one ldlm_lock is locked simultaneously. Also, fix possible
2771 dead-lock in ldlm_lock_change_resource() by enforcing particular
2776 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2777 Details : Instead of specifying each ldlm_lock call-back through separate
2778 parameter, wrap them into struct ldlm_callback_suite.
2782 Description: Kill join_lru obd method and its callers.
2783 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2784 in memory---a requirement implemented through ->o_join_lru() obd
2785 method in HEAD. Remove this method and its users.
2789 Description: Add asynchronous ldlm ENQUEUE completion handler.
2790 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2791 case that stock ldlm_completion_ast() cannot handle as it waits
2792 until lock is granted. Introduce new ldlm_completion_ast_async()
2793 for this. Also comment ldlm_completion_ast().
2797 Description: ldlm_error <-> errno conversion.
2798 Details : Add functions to map (rather arbitrary) between LDLM error codes
2799 and standard errno values. CLIO needs this to prevent LDLM specific
2800 constants from escaping ldlm and osc.
2804 Description: Kill unused ldlm_handle2lock_ns() function.
2805 Details : Kill unused ldlm_handle2lock_ns() function.
2809 Description: Add lu_ref support to ldlm_lock
2810 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2811 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2812 and ldlm_resource. LDLM interface has to be changed, because code that
2813 releases a reference on a lock, has to "know" what reference this is.
2814 In the most frequent case
2816 lock = ldlm_handle2lock(handle);
2818 LDLM_LOCK_PUT(lock);
2820 no changes are required. When any other reference (received _not_ from
2821 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2822 instead of LDLM_LOCK_PUT().
2824 Arguably, changes are pervasive, and interface requires some discipline
2825 for proper use. On the other hand, it was very instrumental in finding
2826 a few leaked lock references.
2830 Description: Add ldlm_lock_addref_try().
2831 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2832 attempts to addref a lock that might be being canceled
2837 Description: Add ldlm_weigh_callback().
2838 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2839 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2840 hard-coded `number of pages' logic.
2844 Description: Add lockdep annotations to llog code.
2845 Details : Use appropriately tagged _nested() locking calls in the places
2846 where llog takes more than one ->lgh_lock lock.
2850 Description: Add loi_kms_set().
2851 Details : Wrap kms updates into a helper function.
2855 Description: Constify instances of struct lsm_operations.
2856 Details : Constify instances of struct lsm_operations.
2860 Description: lu_conf support.
2861 Details : On a server, a file system object is uniquely identified
2862 by a fid, which is sufficient to locate and load all object
2863 state (inode). On a client, on the other hand, more data are
2864 necessary instantiate an object. Change lu_object_find() and
2865 friends to take additional `lu_conf' argument describing object.
2866 Typically this includes layout information.
2870 Description: lu_context fixes.
2871 Details : Introduce new lu_context functions that are needed on the client
2872 side, where some system threads (ptlrpcd) are shared by multiple
2873 modules, and so cannot be stopped during module shutdown.
2877 Description: Add start and stop methods to lu_device_type_operations.
2878 Details : Introduce two new methods in lu_device_type_operations, that are
2879 invoked when first instance of a given type is created and last one
2880 is destroyed respectively. This is need by CLIO.
2884 Description: Add lu_ref support to struct lu_device.
2885 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2886 track leaked references.
2890 Description: Introduce lu_kmem_descr.
2891 Details : lu_kmem_descr and its companion interface allow to create
2892 and destroy a number of kmem caches at once.
2896 Description: Fix lu_object finalization race.
2897 Details : Fix a race between lu_object_find() finding an object and its
2898 concurrent finalization. This race is (most likely) not possible
2899 on the server, but might happen on the client.
2903 Description: Introduce lu_ref interface.
2904 Details : lu_ref is a debugging module allowing to track references to
2905 a given object. It is quite cpu expensive, and has to be
2906 explicitly enabled with --enable-lu_ref. See usage description
2911 Description: Factor lu_site procfs stats into a separate function.
2912 Details : Separate lu_site stats printing code into a separate function
2913 that can be reused on a client.
2917 Description: Constify instances of struct {lu,dt,md}_device_operations.
2918 Details : Constify instances of struct {lu,dt,md}_device_operations.
2922 Description: Introduce struct md_site and move meta-data specific parts of
2923 struct lu_site here.
2924 Details : Move md-specific fields out of struct lu_site into special struct
2925 md_site, so that lu_site can be used on a client.
2929 Description: Kill mdd_lov_destroy().
2930 Details : Remove unused mdd code.
2934 Description: Add st_block checking to multistat.c.
2935 Details : Add st_block checking to multistat.c.
2939 Description: Add lu_ref support to struct obd_device.
2940 Details : Add lu_ref tracking to obd_device.
2944 Description: Kill obd_set_fail_loc().
2945 Details : Remove unused code.
2949 Description: Add special type for ptlrpc_request interpret functions.
2950 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2951 there. Actual usage will be in CLIO.
2955 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2956 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2957 doesn't work with lockdep.
2961 Description: Add rwv.c test program.
2962 Details : New testing program exercising readv(2) and writev(2) (Qian).
2966 Description: Add sendfile.c test program.
2967 Details : New testing program exercising sendfile(2) (Jay).
2971 Description: Ratelimit a message that can be very frequent.
2972 Details : Ratelimit a memory allocation failure message that can
2977 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2978 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2982 Description: Liblustre build fixes.
2983 Details : Liblustre build fixes.
2987 Description: libcfs: add cfs_{need,cond}_resched() interface.
2988 Details : libcfs: add cfs_{need,cond}_resched() definition and
2989 implementations for Linux, NT, and liblustre.
2991 Severity : enhancement
2993 Description: More exported tunables for mballoc
2994 Details : Add support for tunable preallocation window and new tunables for
2995 large/small requests
2999 Description: Detect corruption of block bitmap and checking for preallocations
3000 Details : Checks validity of on-disk block bitmap. Also it does better
3001 checking of number of applied preallocations. When corruption is
3002 found, it turns filesystem readonly to prevent further corruptions.
3006 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3007 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3008 window to 0. If it does hit cache miss, and read-pattern is still
3009 stride-io mode, does not reset the stride window, but also does not
3010 increase the stride window length in this case.
3014 Frequency : only for big-endian servers
3015 Description: Check if system is big-endian while mounting fs with extents feature
3016 Details : Mounting a filesystem with extents feature will fail on big-endian
3017 systems since ext3-based ldiskfs is not supported on big-endian
3018 systems. This can be over-riden with "bigendian_extents" mount option.
3020 Severity : enhancement
3022 Description: The root squash functionality
3023 Details : A security feature, which is to prevent users from being able
3024 to mount lustre on their desktop, run as root, and delete
3025 all of the files in the filesystem. The goal is accomplished by
3026 remapping user id (UID) and group id (GID) of the root user to
3027 a UID and GID specified by the system administartor via Lustre
3028 configuration management server (MGS). The functionality also
3029 allows to specify sets of clients for which the remapping does
3034 Description: Excessive recovery window
3035 Details : With AT enabled, the recovery window can be excessively long (6000+
3036 seconds). To address this problem, we no longer use
3037 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3038 timeout no longer depends on the service time, it is set to
3039 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3040 time via pb_service_time.
3044 Description: Watchdog triggered on MDS failover
3045 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3046 the OSTs know that the MDS "UUID" can be reused for the same export
3047 from a different NID, so we do not need to wait for the export to be
3051 Frequency : rare, only if using MMP with Linux RAID
3053 Description: MMP doesn't work with Linux RAID
3054 Details : While using HA for Lustre servers with Linux RAID, it is possible
3055 that MMP will not detect multiple mounts. To make this work we
3056 need to unplug the device queue in RAID when the MMP block is being
3057 written. Also while reading the MMP block, we should read it from
3058 disk and not the cached one.
3060 Severity : enhancement
3062 Description: open file using fid
3063 Details : A file can be opened using just its fid, like
3064 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3067 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3070 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3071 Details : add proper locking for creation of jbd_4k slab cache
3075 Description: MMP check in ext3_remount() fails without displaying any error
3076 Details : When multiple mount protection fails during remount, proper error
3079 Severity : enhancement
3081 Description: Allow stripe size to be up to 4G-64k
3082 Details : Fix math logic to allow large stripe sizes.
3086 Description: add check for >8TB ldiskfs filesystems
3087 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3088 Don't allow >8TB ldiskfs filesystems to be mounted without
3089 force_over_8tb mount option
3094 Descriptoin: OSS complains "too many blocks" and crashes
3095 Details : In filter_commitrw_write():
3096 1. lquota_chkquota();
3097 2. obdo_from_inode();
3098 3. lquota_pending_commit();
3099 "2." has changed uid and gid in this case, then caused
3100 lquota_chkquota() and lquota_pending_commit() to use
3101 different uid and gid. That is the root of the bug.
3105 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3106 Details : Detect and print out the 'stripe count = -1' case properly.
3110 Description: Out or order replies might be lost on replay
3111 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3112 request with tid smaller than one currently being added, add it
3113 to the start, not end of the list.
3115 --------------------------------------------------------------------------------
3117 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3119 * Support for kernels:
3120 2.6.5-7.283 (SLES 9),
3121 2.6.9-55.EL (RHEL 4),
3122 2.6.16.46-0.14 (SLES 10),
3123 2.6.18.8 vanilla (kernel.org)
3124 * Client support for unpatched kernels:
3125 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3126 2.6.16 - 2.6.22 vanilla (kernel.org)
3127 * Due to recently discovered recovery problems, we do not recommend
3128 using patchless RHEL 4 clients with this or any earlier release.
3129 * Recommended e2fsprogs version: 1.39.cfs8
3130 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3131 * Starting with this release, the ldiskfs backing filesystem required
3132 by Lustre is now in its own package, lustre-ldiskfs. This package
3133 should be installed. It is versioned separately from Lustre and
3134 may be released separately in future.
3138 Description: block reactivating mgc import until all deactivates complete
3139 Details : Fix race when failing back MDT/MGS to itself (testing)
3142 Frequency : at statup only
3144 Description: mds_lov_synchronize race leads to various problems
3145 Details : simultaneous MDT->OST connections at startup can cause the
3146 sync to abort, leaving the OSC in a bad state.
3148 Severity : enhancement
3150 Description: add optional extra BUILD_VERSION info
3151 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3152 to override the lustre version.
3155 Frequency : 2.6.18 servers only
3157 Description: ll_kern_mount() doesn't release the module reference
3158 Details : The ldiskfs module reference count never drops down to 0
3159 because ll_kern_mount() doesn't release the module reference.
3164 Description: server LBUG when using old ost_num_threads parameter
3165 Details : Accept the old ost_num_threads parameter but warn that it
3166 is deprecated, and fix an off-by-one error that caused an LBUG.
3171 Description: Transient SCSI error results in persistent IO issue
3172 Details : iobuf->dr_error is not reinitialized to 0 between two
3176 Frequency : sometimes when underlying device returns I/O errors
3178 Description: OSTs not going read-only during write failures
3179 Details : OSTs are not remounted read-only when the journal commit threads
3180 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3181 instead of the ext3 wrappers.
3185 Description: poor connect scaling with increasing client count
3186 Details : Don't run filter_grant_sanity_check for more than 100 exports
3187 to improve scaling for large numbers of clients.
3190 Frequency : SLES10 only
3192 Description: sanity-quota.sh quotacheck failed: rc = -22
3193 Details : Quotas cannot be enabled on SLES10.
3196 Frequency : liblustre clients only
3198 Description: getdirentries does not give error when run on compute nodes
3199 Details : getdirentries does not fail when the size specified as an argument
3200 is too small to contain at least one entry
3202 Severity : enhancement
3204 Description: Add LNET router traceability for debug purposes
3205 Details : If a checksum failure occurs with a router as part of the
3206 IO path, the NID of the last router that forwarded the bulk data
3207 is printed so it can be identified.
3212 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3213 Details : Due to a race condition, liblustre clients were occasionally
3214 evicted incorrectly.
3216 Severity : enhancement
3218 Description: lfs setstripe use optional parameters instead of postional
3221 Severity : enhancement
3223 Description: Nanosecond timestamp support for ldiskfs
3224 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3225 resolution timestamps. There is not yet support for this at
3226 the Lustre filesystem level.
3229 Frequency : during server recovery
3231 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3232 Details : request with rq_no_resend flag not awake l_wait_event if they get
3236 Frequency : nfs export on patchless client
3238 Description: connectathon hang when test nfs export over patchless client
3239 Details : Disconnected dentry cannot be found with lookup, so we do not need
3240 to unhash it or make it invalid
3243 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3244 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3245 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3250 Description: open req refcounting wrong on reconnect
3251 Details : If reconnect happened between getting open reply from server and
3252 call to mdc_set_replay_data in ll_file_open, we will schedule
3253 replay for unreferenced request that we are about to free.
3254 Subsequent close will crash in variety of ways.
3255 Check that request is still eligible for replay in
3256 mdc_set_replay_data().
3261 Description: disable writes to filesystem when reading health_check file
3262 Details : the default for reading the health_check proc file has changed
3263 to NOT do a journal transaction and write to disk, because this
3264 can cause reads of the /proc file to hang and block HA state
3265 checking on a healthy but otherwise heavily loaded system. It
3266 is possible to return to the previous behaviour during configure
3267 with --enable-health-write.
3269 Severity : enhancement
3271 Description: 64-bit inode version
3272 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3273 to the inode. This will be required for version-based recovery.
3278 Description: MDS fails to start if a duplicate client export is detected
3279 Details : in some rare cases it was possible for a client to connect to
3280 an MDS multiple times. Upon recovery the MDS would detect this
3281 and fail during startup. Handle this more gracefully.
3283 Severity : enhancement
3285 Description: Add -o localflock option to simulate old noflock
3287 Details : This will achieve local-only flock/fcntl locks
3293 Description: log_commit_thread vs filter_destroy race leads to crash
3294 Details : Take import reference before releasing llog record semaphore
3299 Description: Wrong request locking in request set processing
3300 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3301 imp_delayed_list, in this place should be used imp_lock.
3304 Frequency : when reconnection
3306 Description: Grant Leak when osc reconnect to OST
3307 Details : When osc reconnect ost, OST(filter) should check whether it
3308 should grant more space to client by comparing fed_grant and
3309 cl_avail_grant, and return the granted space to client instead
3310 of "new granted" space, because client will call osc_init_grant
3311 to update the client grant space info.
3314 Frequency : when client reconnect to OST
3316 Description: Grant Leak when osc do resend and replay bulk write
3317 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3318 bulk write request, because the grant info while be sync between
3319 OSC and OST when reconnect, and we should ignore the grant info
3320 these of resend/replay write req.
3325 Description: Grant space more than available left space sometimes.
3326 Details : When then OST is about to be full, if two bulk writing from
3327 different clients came to OST. Accord the avaliable space of the
3328 OST, the first req should be permitted, and the second one
3329 should be denied by ENOSPC. But if the seconde arrived before
3330 the first one is committed. The OST might wrongly permit second
3331 writing, which will cause grant space > available space.
3334 Frequency : when client is evicted
3336 Description: Grant might be wrongly erased when osc is evicted by OST
3337 Details : when the import is evicted by server, it will fork another
3338 thread ptlrpc_invalidate_import_thread to invalidate the
3339 import, where the grant will be set to 0. While the original
3340 thread will update the grant it got when connecting. So if
3341 the former happened latter, the grant will be wrongly errased
3342 because of this race.
3347 Description: Checking Stale with correct fid
3348 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3349 to check whether it is stale, because sometimes, we want the
3350 enqueue happened anyway, and op_data.fid2 will not be initialized.
3352 Severity : enhancement
3354 Description: update patchless client
3355 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3358 Frequency : only with 2.4 kernel
3360 Description: random memory corruption
3361 Details : size of struct ll_inode_info is to big for union inode.u and this
3362 can be cause of random memory corruption.
3367 Description: Memory leak in recovery
3368 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3369 It should also check obd_fail before fsfilt_start, otherwise if
3370 fsfilt_start return -EROFS,(failover mds during mds recovery).
3371 then the req will return with repmsg->transno = 0 and rc = EROFS.
3372 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3373 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3374 be freed no matter whether fsfilt_commit success or not.
3377 Frequency : only with huge count clients
3379 Description: Prevents from taking the superblock lock in llap_from_page for
3381 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3382 for llap_from_page call in ll_removepage prevents from taking the
3383 superblock lock for a soon died page.
3388 Description: Not check open intent error before release open handle
3389 Details : in some rare cases, the open intent error is not checked before
3390 release open handle, which may cause
3391 ASSERTION(open_req->rq_transno != 0), because it tries to release
3392 the failed open handle.
3397 Description: Set cat log bitmap only after create log success.
3398 Details : in some rare cases, the cat log bitmap is set too early. and it
3399 should be set only after create log success.
3403 Description: Accessing a block bevice can re-enable I/O when Lustre is
3404 tearing down a device.
3405 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3410 Description: service threads may hog cpus when there are a lot of requests
3411 Details : Insert cond_resched to give other threads a chance to use some CPU
3416 Description: the cat log was not initialized in recovery
3417 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3418 unlink log on mds will not be initialized until mds post
3419 recovery. And also in mds post recovery, the unlink log will
3420 initialization will be done asynchronausly, so there will be race
3421 between add unlink log and unlink log initialization.
3425 Description: brw_stats were being printed incorrectly
3426 Details : brw_stats were being printed as log2 but all of them were not
3427 recorded as log2. Also remove some code duplication arising from
3428 filter_tally_{read,write}.
3432 Frequency : rare, only in recovery.
3433 Description: ASSERTION(req->rq_type != LI_POISON) failed
3434 Details : imp_lock should be held while iterating over imp_sending_list for
3435 prevent destroy request after get timeout in ptlrpc_queue_wait.
3439 Description: replay-single.sh test 52 fails
3440 Details : A lock's skiplist need to be cleanup when it being unlinked
3441 from its resource list.
3445 Description: Short directio read returns full requested size rather than
3447 Details : Direct I/O operations should return actual amount of bytes
3448 transferred rather than requested size.
3450 Severity : enhancement
3452 Description: metadata RPC reduction (e.g. for rm performance)
3453 Details : decrease the amount of synchronous RPC between clients and servers
3454 by canceling conflicing lock before the operation on the client
3455 and packing thier handles into the main operation RPC to server.
3457 Severity : enhancement
3459 Description: Async OSC create to avoid the blocking unnecessarily.
3460 Details : If an OST has no remaining object, system will block on the
3461 creation when it needs to create a new object on this OST. Now,
3462 ways use pre-created objects when available, instead of blocking on
3463 an empty osc while others are not empty. If we must block, we
3464 block for the shortest possible period of time.
3468 Description: improve handling recoverable errors
3469 Details : if request processig with error which can be recoverable on server
3470 request should be resend, otherwise page released from cache and
3473 Severity : enhancement
3475 Description: refine locking for avoid write wrong info into lov_objid file
3476 Details : fix possible races with add new target and write/update data in
3479 --------------------------------------------------------------------------------
3481 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3486 Frequency : on some architectures
3488 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3489 Details : Uninitialized flags sometimes cause configuration commands to
3493 Frequency : patchless clients only
3495 Description: missing __iget() symbol export
3496 Details : The __iget() symbol export is missing. To avoid the need for
3497 this on patchless clients the deathrow inode reaper is turned
3498 off, and we depend on the VM to clean up old inodes. This
3499 dependency was during via the fix for bug 12181.
3501 --------------------------------------------------------------------------------
3503 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3505 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3506 INTEROPERATE with older versions automatically. In many cases a
3507 special upgrade step is needed. Please read the
3508 user documentation before upgrading any part of a live system.
3509 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3510 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3511 * WARNING: Lustre configuration and startup changes are required with
3512 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3516 Severity : enhancement
3518 Description: MountConf
3519 Details : Lustre configuration is now managed via mkfs and mount
3520 commands instead of lmc and lconf. New obd types (MGS, MGC)
3521 are added for dynamic configuration management. See
3522 https://mail.clusterfs.com/wikis/lustre/MountConf for
3525 Severity : enhancement
3527 Description: dynamic OST addition
3528 Details : OSTs can now be added to a live filesystem
3530 Severity : enhancement
3532 Description: startup order invariance
3533 Details : MDTs and OSTs can be started in any order. Clients only
3534 require the MDT to complete startup.
3536 Severity : enhancement
3538 Description: parallel, asynchronous orphan cleanup
3539 Details : orphan cleanup is now performed in separate threads for each
3540 OST, allowing parallel non-blocking operation.
3542 Severity : enhancement
3544 Description: optimized stripe assignment
3545 Details : stripe assignments are now made based on ost space available,
3546 ost previous usage, and OSS previous usage, in order to try
3547 to optimize storage space and networking resources.
3549 Severity : enhancement
3551 Description: Permanently set tunables
3552 Details : All writable /proc/fs/lustre tunables can now be permanently
3553 set on a per-server basis, at mkfs time or on a live system.
3555 Severity : enhancement
3557 Description: Lustre message v2
3558 Details : Add lustre message format v2.
3560 Severity : enhancement
3562 Description: client OST exclusion list
3563 Details : Clients can be started with a list of OSTs that should be
3564 declared "inactive" for known non-responsive OSTs.
3568 Description: SPEC SFS validation failure on NFS v2 over lustre.
3569 Details : Changes the blocksize for regular files to be 2x RPC size,
3570 and not depend on stripe size.
3572 Severity : enhancement
3574 Description: Multiple MD RPCs in flight.
3575 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3576 To avoid overly-overloading MDS, introduce a limit on number of
3577 MDS RPCs in flight for a single client and add /proc controls
3578 to adjust this limit.
3580 Severity : enhancement
3582 Description: client read/write statistics
3583 Details : Add client read/write call usage stats for performance
3584 analysis of user processes.
3585 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3586 file access. extents_stats shows chunk size distribution.
3587 extents_stats_per_process show chunk size distribution per
3590 Severity : enhancement
3592 Description: mds statistics
3593 Details : Add detailed mds operations statistics in
3594 /proc/fs/lustre/mds/*/stats.
3598 Description: Failure of copying files with lustre special EAs.
3599 Details : Client side always return success for setxattr call for lustre
3600 special xattr (currently only "trusted.lov").
3605 Description: Refcount LNET uuids
3606 Details : The global LNET uuid list grew linearly with every startup;
3607 refcount repeated list entries instead of always adding to
3610 Severity : enhancement
3612 Description: Dynamic service threads
3613 Details : Within a small range, start extra service threads
3614 automatically when the request queue builds up.
3617 Frequency : mixed-endian client/server environments
3619 Description: mixed-endian crashes
3620 Details : The new msg_v2 system had some failures in mixed-endian
3623 Severity : enhancement
3625 Description: Easy OST removal
3626 Details : OSTs can be permanently deactivated with e.g. 'lctl
3627 conf_param lustre-OST0001.osc.active=0'
3629 Severity : enhancement
3631 Description: MGS proc entries
3632 Details : Added basic proc entries for the MGS showing what filesystems
3635 Severity : enhancement
3637 Description: provide MGS failover
3638 Details : Added config lock reacquisition after MGS server failover.
3640 Severity : enhancement
3642 Description: add Linux 2.4 support
3643 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3647 Description: a large application tries to do I/O to the same resource and dies
3648 in the middle of it.
3649 Details : Check the req->rq_arrival time after the call to
3650 ost_brw_lock_get(), but before we do anything about
3651 processing it & sending the BULK transfer request. This
3652 should help move old stale pending locks off the queue as
3653 quickly as obd_timeout.
3656 Frequency : when an incorrect nid is specified during startup
3658 Description: ptlrpc connect to non-existant node causes kernel crash
3659 Details : LNET can't be re-entered from an event callback, which
3660 happened when we expire a message after the export has been
3661 cleaned up. Instead, hand the zombie cleanup off to another
3664 Severity : enhancement
3666 Description: plain/inodebits lock performance improvement
3667 Details : Grouping plain/inodebits in granted list by their request modes
3668 and bits policy, thus improving the performance of search through
3672 Frequency : only if OST filesystem is corrupted
3674 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3675 Details : for a short time RPCs with bulk IO are in the replay list,
3676 but replay of bulk IOs is unimplemented. If the OST filesystem
3677 is corrupted due to disk cache incoherency and then replay is
3678 started it is possible to trip an assertion. Avoid putting
3679 committed RPCs into the replay list at all to avoid this issue.
3682 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3684 Description: System hang on startup
3685 Details : This bug allowed the liblustre (e.g. catamount) client to
3686 return to the app before handling all startup RPCs. This
3687 could leave the node unresponsive to lustre network traffic
3688 and manifested as a server ptllnd timeout.
3690 Severity : enhancement
3692 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3693 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3694 boolean to control whether to print peer debug info when a
3695 client's RPC times out.
3698 Frequency : only for kernels with patches from Lustre below 1.4.3
3700 Description: Remove old rdonly API
3701 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3704 Frequency : only for devices with external journals
3706 Description: Set external device read-only also
3707 Details : During a commanded failover stop, we set the disk device
3708 read-only while the server shuts down. We now also set any
3709 external journal device read-only at the same time.
3712 Frequency : when upgrading from 1.4 while trying to change parameters
3714 Description: The wrong (new) MDC name was used when setting parameters for
3715 upgraded MDT's. Also allows changing of OSC (and MDC)
3716 parameters if --writeconf is specified at tunefs upgrade time.
3719 Frequency : when setting specific ost indicies
3721 Description: QOS code breaks on skipped indicies
3722 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3723 created with --index need not be sequential.
3725 Severity : enhancement
3727 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3728 Details : The uninit_groups feature works in conjunction with the kernel
3729 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3730 up the pass1 processing of e2fsck. This is a read-only feature
3731 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3732 mount filesystems that have had this feature enabled.
3734 Severity : enhancement
3736 Description: Improve multi-block allocation algorithm to avoid fragmentation
3737 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3738 allocation locality and avoid filesystem fragmentation.
3740 ------------------------------------------------------------------------------
3742 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3744 * Support for kernels:
3745 2.4.21-47.0.1.EL (RHEL 3)
3746 2.6.5-7.283 (SLES 9)
3747 2.6.9-42.0.10.EL (RHEL 4)
3748 2.6.12.6 vanilla (kernel.org)
3749 2.6.16.27-0.9 (SLES 10)
3750 * Recommended e2fsprogs version: 1.39.cfs5
3752 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3756 Frequency : occasional, depends on client load and configuration
3757 Bugzilla : 12181, 12203
3758 Description: data loss for recently-modified files
3760 Details : In some cases it is possible that recently written or created
3761 files may not be written to disk in a timely manner (this should
3762 normally be within 30s unless client IO load is very high).
3763 The problem appears as zero-length files or files that are a
3764 multiple of 1MB in size after a client crash or client eviction
3765 that are missing data at the end of the file.
3767 This problem is more likely to be hit on clients where files are
3768 repeatedly created and unlinked in the same directory, clients
3769 have a large amount of RAM, have many CPUs, the filesystem has
3770 many OSTs, the clients are rebooted frequently, and/or the files
3771 are not accessed by other nodes after being written.
3773 The presence of the problem can be detected by looking at
3774 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3775 smaller than the second (nr_unused) then dirty files will not
3776 be flushed automatically to disk. "sync; sleep 10" should be
3777 run several times on the node before unmounting it to update
3778 Lustre (this is also safe to run on nodes without this problem).
3780 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3781 that can cause this same problem, so customers using that kernel
3782 also need to update the kernel in addition to Lustre. In order
3783 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3785 It is normal that files written just before a client crash (less
3786 than 30s) may not yet have been flushed to disk, even for local
3790 Frequency : frequent on thin XT3 nodes
3792 Description: UUID collision on thin XT3 Linux nodes
3793 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3794 randomly, since we relied on an insufficiently-seeded PRNG.
3799 Description: OSS hangs after "All ost request buffers busy"
3800 Details : A deadlock between quota and journal operations caused OSS
3801 hangs after printing "All ost request buffers busy."
3804 Frequency : always on liblustre builds
3806 Description: Cleanup compiler warnings on liblustre
3809 Frequency : always on liblustre builds on XT3
3811 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3815 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3816 > 32000 subdirectories
3817 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3818 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3819 subdirectory count crosses 32000. This will aid e2fsck to
3820 correctly handle more than 32000 subdirectories.
3823 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3825 Description: System hang on startup
3826 Details : This bug allowed the liblustre (e.g. catamount) client to
3827 return to the app before handling all startup RPCs. This
3828 could leave the node unresponsive to lustre network traffic
3829 and manifested as a server ptllnd timeout.
3831 Severity : enhancement
3833 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3834 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3835 boolean to control whether to print peer debug info when a
3836 client's RPC times out.
3841 Description: make O_SYNC working on 2.6 kernels
3842 Details : 2.6 kernels use different method for mark pages for write,
3843 so need add a code to lustre for O_SYNC work.
3848 Description: Failure to close file and release space on NFS
3849 Details : Put inode details into lock acquired in ll_intent_file_open.
3850 Use mdc_intent_lock in ll_intent_open to properly
3851 detect all kind of errors unhandled by mdc_enqueue.
3856 Description: proc file read during shutdown sometimes raced obd removal,
3858 Details : Add lock to prevent obd access after proc file removal.
3861 Frequency : Only for files larger than 4GB on 32-bit clients.
3863 Description: improperly doing page alignment of locks
3864 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3865 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3870 Description: RPCs being resent when they shouldn't be
3871 Details : Some RPCs that should not be resent are being resent. This
3872 can cause inconsistencies in the RPC state machine. Do not
3873 resend such requests.
3876 Frequency : rare, only with NFS export
3878 Description: Crash on NFS re-export node
3879 Details : under very unusual load conditions an assertion is hit in
3880 ll_intent_file_open()
3883 Frequency : only if OST filesystem is corrupted
3885 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3886 Details : for a short time RPCs with bulk IO are in the replay list,
3887 but replay of bulk IOs is unimplemented. If the OST filesystem
3888 is corrupted due to disk cache incoherency and then replay is
3889 started it is possible to trip an assertion. Avoid putting
3890 committed RPCs into the replay list at all to avoid this issue.
3895 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3896 Details : Large single O_DIRECT read and write calls can fail to allocate
3897 a sufficiently large buffer to process the request. In case of
3898 allocation failure the allocation is retried with a smaller
3899 buffer and broken into smaller requests.
3901 Severity : enhancement
3903 Description: Add -o localflock option to simulate old noflock behaviour.
3904 Details : This will achieve local-only flock/fcntl locks coherentness.
3909 Description: versioning check is incomplete
3910 Details : Checking the version difference of client vs. server, report
3911 error if the gap is too big.
3916 Description: add support PG_writeback bit
3917 Details : add support for PG_writeback bit for Lustre, for more carefull
3918 work with page cache in 2.6 kernel. This also fix some deadlocks
3919 and remove hack for work O_SYNC with 2.6 kernel.
3921 Severity : enhancement
3923 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3924 Details : The uninit_groups feature works in conjunction with the kernel
3925 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3926 up the pass1 processing of e2fsck. This is a read-only feature
3927 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3928 mount filesystems that have had this feature enabled.
3930 Severity : enhancement
3932 Description: Improve multi-block allocation algorithm to avoid fragmentation
3933 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3934 allocation locality and avoid filesystem fragmentation.
3936 ------------------------------------------------------------------------------
3938 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3940 * Support for kernels:
3941 2.6.9-42.0.3.EL (RHEL 4)
3942 2.6.5-7.276 (SLES 9)
3943 2.4.21-47.0.1.EL (RHEL 3)
3944 2.6.12.6 vanilla (kernel.org)
3945 2.6.16.21-0.8 (SLES10)
3946 * Recommended e2fsprogs version: 1.39.cfs2-0
3948 * The backwards-compatible /proc/sys/portals symlink has been removed
3949 in this release. Before upgrading, please ensure that you change
3950 any configuration scripts or /etc/sysctl.conf files that access
3951 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3952 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3953 in advance of the upgrade on any system running Lustre 1.4.6 or
3954 newer, since /proc/sys/lnet was added in that version.
3955 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3959 Frequency : only when quota is used
3961 Description: avoid scanning export list for quota master
3962 Details : Change the algorithms to avoid scanning export list in order
3963 to improve the efficiency.
3966 Frequency : MDS failover only, very rarely
3968 Description: "went back in time" messages on mds failover
3969 Details : The greatest transno may be lost when the current operation
3970 finishes with an error (transno==0) and the client's last_rcvd
3971 record is over-written. Save the greatest transno in the
3972 mds_last_transno for this case.
3975 Frequency : always for specific kernels and striping counts
3977 Description: client may get "Matching packet too big" without ACL support
3978 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3979 when trying to access files in certain configurations. The
3980 clients should in fact be denied when mounting because they do
3981 not understand ACLs.
3984 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3986 Description: many clients connecting with IO in progress causes connect timeouts
3987 Details : Avoid synchronous journal commits to avoid delays caused by many
3988 clients connecting/disconnecting when bulk IO is in progress.
3989 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3990 OST_IO_PORTAL to avoid delays behind potentially many pending
3994 Frequency : occasionally with multiple writers to a single file
3996 Description: shared writes to file may result in wrong size reported by stat()
3997 Details : Allow growing of kms when extent lock is cancelled
4000 Frequency : always with random mmap IO to multi-striped file
4002 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4003 Details : Only if the hole is at the end of OST object so that kms is too
4004 small. Fix is to increase kms accordingly in ll_nopage.
4007 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4009 Description: writes to a missing object would leak memory on the OST
4010 Details : If there is an inconsistency between the MDS and OST filesystems,
4011 such that the MDS references an object that doesn't exist, writes
4012 to that object will leak memory due to incorrect cleanup in the
4013 error handling path, eventually running out of memory on the OST.
4018 Description: Creating too long symlink causes lustre errors
4019 Details : Check symlink and name lengths before sending requests to MDS.
4022 Frequency : only if flock is enabled (not on by default)
4024 Description: posix locks not released on fd closure on 2.6.9+
4025 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4026 kernels, this caused such locks not to be released on fd close and
4027 then assertions on fs unmount about still used locks.
4030 Frequency : MDS failover only, very rarely
4032 Description: clients may get ASSERTION(granted_lock != NULL)
4033 Details : When request was taking a long time, and a client was resending
4034 a getattr by name lock request. The were multiple lock requests
4035 with the same client lock handle and
4036 mds_getattr_name->fixup_handle_for_resent_request found one of the
4037 lock handles but later failed with ASSERTION(granted_lock != NULL).
4042 Description: handle->h_buffer_credits > 0, assertion failure
4043 Details : h_buffer_credits is zero after truncate, causing assertion
4044 failure. This patch extends the transaction or creates a new
4048 Frequency : NFS re-export or patchless client
4049 Bugzilla : 11179, 10796
4050 Description: Crash on NFS re-export node (__d_move)
4051 Details : We do not want to hash the dentry if we don't have a lock.
4052 But if this dentry is later used in d_move, we'd hit uninitialised
4053 list head d_hash, so we just do this to init d_hash field but
4054 leave dentry unhashed.
4057 Frequency : NFS re-export or patchless client
4059 Description: NFS exports has problem with symbolic link
4060 Details : lustre client didn't properly install dentry when re-exported
4061 to NFS or running patchless client.
4064 Frequency : NFS re-export or patchless client
4066 Description: Various nfs/patchless fixes.
4067 Details : fixes reuse disconected alias for lookup process - this fixes
4068 warning "find_exported_dentry: npd != pd",
4069 fix permission error with open files at nfs.
4070 fix apply umask when do revalidate.
4073 Frequency : occasional
4075 Description: Crash on NFS re-export node
4076 Details : calling clear_page() on the wrong pointer triggered oops in
4077 generic_mapping_read().
4080 Frequency : rarely, using O_DIRECT IO
4082 Description: unaligned directio crashes client with LASSERT
4083 Details : check for unaligned buffers before trying any requests.
4086 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4088 Description: stale data returned from RAID cache
4089 Details : If only a small amount of IO is done to the RAID device before
4090 reading it again it is possible to get stale data from the RAID
4091 cache instead of reading it from disk.
4094 Frequency : always for sles10 kernel
4096 Description: sles10 support
4097 Details : ll_follow_link: compile fixes and using of nd_set_link
4098 under newer kernels.
4101 Frequency : depends on arch, kernel and compiler version, always on sles10
4104 Description: recursive or deep enough symlinks cause stack overflow
4105 Details : getting rid of large stack-allocated variable in
4109 Frequency : depends on hardware
4111 Description: lustre write performance loss in the SLES10 kernel
4112 Details : the performance loss is caused by using of write barriers in the
4113 ext3 code. The SLES10 kernel turns barrier support on by
4114 default. The fix is to undo that change for ldiskfs.
4116 ------------------------------------------------------------------------------
4118 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4120 * Support for kernels:
4121 2.6.9-42.0.3EL (RHEL 4)
4122 2.6.5-7.276 (SLES 9)
4123 2.4.21-47.0.1.EL (RHEL 3)
4124 2.6.12.6 vanilla (kernel.org)
4128 Frequency : quota enabled and large files being deleted
4130 Description: releasing more than 4GB of quota at once hangs OST
4131 Details : If a user deletes more than 4GB of files on a single OST it
4132 will cause the OST to spin in an infinite loop. Release
4133 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4138 Description: statfs data retrieved from /proc may be stale or zero
4139 Details : When reading per-device statfs data from /proc, in the
4140 {kbytes,files}_{total,free,avail} files, it may appear
4141 as zero or be out of date.
4144 Frequency : systems with MD RAID1 external journal devices
4146 Description: lconf's call to blkid is confused by RAID1 journal devices
4147 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4148 of returning all block devices that match the journal UUID.
4151 Frequency : always, for aggregate stripe size over 4GB
4153 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4154 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4155 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4158 Frequency : always if "lfs find" used on a local file/directory
4160 Description: "lfs find" segfaults if used on a local file/directory
4161 Details : The case where a directory component was not specified wasn't
4162 handled correctly. Handle this properly.
4165 Frequency : always on ppc64
4167 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4168 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4169 perform endianness conversion.
4172 Frequency : rarely (truncate to non-zero file size after write under load)
4173 Bugzilla : 10730, 10687
4174 Description: Files padded with zeros to next 4K multiple
4175 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4176 it is possible that files that are truncated to a non-zero size
4177 immediately after being written are filled with zero bytes beyond
4178 the truncated size. No file data is lost.
4180 Severity : enhancement
4182 Description: Allow recovery/failover for liblustre clients.
4183 Details : liblustre clients were unaware of failover configurations until
4186 Severity : enhancement
4188 Description: user file locks should fail when not mounting with flock option
4189 Details : Set up an error-returning stub in ll_file_operations.lock field
4190 to prevent incorrect behaviour when client is mounted without
4191 flock option. Also, set up properly f_op->flock field for
4195 Frequency : always on ia64
4197 Description: "lfs df" loops on printing out MDS statfs information
4198 Details : The obd_ioctl_data was not initialized and in some systems
4199 this caused a failure during the ioctl that did not return
4200 an error. Initialize the struct and return an error on failure.
4203 Frequency : SLES 9 only
4205 Description: Error of copying files with lustre special EAs as root
4206 Details : Client side always return success for setxattr call for lustre
4207 special xattr (currently only "trusted.lov").
4210 Frequency : rarely on clusters with both ia64+i386 clients
4212 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4213 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4214 clients are doing concurrent writes to the same file it is
4215 possible that the ia64 clients may LASSERT because the OST
4216 extent locks are not PAGE_SIZE aligned. Ensure that grown
4217 locks are always aligned on the request boundary.
4220 Frequency : specific use, occasional
4222 Description: Overwriting in use executable truncates on-disk binary image
4223 Details : If one node attempts to overwrite an executable in use by
4224 another node, we now correctly return ETXTBSY instead of
4225 truncating the file.
4227 Severity : enhancement
4229 Description: Async OSC create to avoid the blocking unnecessarily.
4230 Details : If an OST has no remaining object, system will block on the creating
4231 when need to create a new object on this OST. Now, ways use
4232 pre-created objects when available, instead of blocking on an
4233 empty osc while others are not empty. If we must block, we block
4234 for the shortest possible period of time.
4239 Description: chmod on Lustre root is propagated to other clients
4240 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4241 invalid by the follow_mount time.
4244 Frequency : liblustre clients only
4246 Description: Race in 'instant cancel' lock handling could lead to such locks
4247 never to be granted in case of SMP MDS
4248 Details : Do not destroy not yet granted but cbpending locks in
4252 Frequency : replay/resend of open
4254 Description: non null lock assetion failure in mds_intent_policy
4255 Details : Trying to replay/resend lockless open requests resulted in
4256 mds_open() returning 0 with no lock. Now it sets a flag if
4257 a lock is going to be returned.
4259 Severity : enhancement
4261 Description: Checksum enhancements
4262 Details : New checksum enhancements allow for resending RPCs that failed
4265 Severity : enhancement
4267 Description: Tunables on number of dirty pages in cacche
4268 Details : Allow to set limit on number of dirty pages cached.
4273 Description: client crash on unmount - lock still has references
4274 Details : In some error handling cases it was possible to leak a lock
4275 reference on a client while accessing a file. This was not
4276 harmful to the client during operation, but would cause the
4277 client to crash when the filesystem is unmounted.
4280 Frequency : specific case, rare
4282 Description: ETXTBSY on mds though file not in use by client
4283 Details : ETXTBSY is no longer incorrectly returned when attempting to
4284 chmod or chown a directory that the user previously tried to
4285 execute or a currently-executing binary.
4288 Frequency : extremely rare except on liblustre-based clients
4290 Description: Lustre space not freed when files are deleted
4291 Details : Clean up open-unlinked files after client eviction. Previously
4292 the unlink was skipped and the files remained as orphans.
4297 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4298 Details : In some cases it was possible to send a blocking callback to
4299 a client doing a glimpse, even though that client didn't get
4300 a lock granted. When the glimpse lock is cancelled on the OST
4301 the freed lock is left on the waiting list and corrupted the list.
4304 Frequency : all core dumps
4306 Description: Broke core dumps to lustre
4307 Details : Negative dentry may be unhashed if parent does not have UPDATE
4308 lock, but some callers, e.g. do_coredump, expect dentry to be
4309 hashed after successful create, hash it in ll_create_it.
4311 ------------------------------------------------------------------------------
4313 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4315 * Support for kernels:
4316 2.6.9-42.0.2.EL (RHEL 4)
4317 2.6.5-7.276 (SLES 9)
4318 2.4.21-40.EL (RHEL 3)
4319 2.6.12.6 vanilla (kernel.org)
4323 Frequency : always on RHEL 3
4325 Description: Number of open files grows over time
4326 Details : The number of open files grows over time, whether or not
4327 Lustre is started. This was due to a filp leak introduced
4328 by one of our kernel patches.
4330 ------------------------------------------------------------------------------
4332 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4334 * Support for kernels:
4335 2.6.9-42.EL (RHEL 4)
4336 2.6.5-7.276 (SLES 9)
4337 2.4.21-40.EL (RHEL 3)
4338 2.6.12.6 vanilla (kernel.org)
4343 Bugzilla : 5719, 9635, 9792, 9684
4344 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4345 Details : If a server is under heavy load and cannot reply to new
4346 connection requests before the client resends the (re)connect,
4347 the connection handling code can behave badly if two service
4348 threads are concurrently handing separate (re)connections from
4349 the same client. Add better locking to the connection handling
4350 code, and ensure that only a single connection will be processed
4351 for a given client UUID, even if the lock is dropped.
4353 Severity : enhancement
4355 Description: add TCP zero-copy support to kernel
4356 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4357 sends if the hardware supports scatter-gather and checksumming.
4358 This allows socklnd to do client-write and server-read more
4359 efficiently and reduce CPU utilization from skbuf copying.
4362 Frequency : only if NFS exporting from client
4364 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4365 Details : If setting attributes on a file created under NFS that had
4366 never been opened it would be possible to oops the client
4367 if the file had no objects.
4370 Frequency : always for liblustre
4372 Description: liblustre client does MDS+OSTs setattr RPC for each write
4373 Details : When doing a write from a liblustre client, the client
4374 incorrectly issued an RPC to the MDS and each OST the file was
4375 striped over in order to update the timestamps. When writing
4376 with small chunks and many clients this could overwhelm the MDS
4377 with RPCs. In all cases it would slow down the write because
4378 these RPCs are unnecessary.
4380 Severity : enhancement
4382 Description: allow number of MDS service threads to be changed at module load
4383 Details : It is now possible to change the number of MDS service threads
4384 running. Adding "options mds mds_num_threads={N}" to the MDS's
4385 /etc/modprobe.conf will set the number of threads for the next
4386 time Lustre is restarted (assuming the "mds" module is also
4387 reloaded at that time). The default number of threads will
4388 stay the same, 32 for most systems.
4393 Description: OST crash if filesystem is unformatted or corrupt
4394 Details : If an OST is started on a device that has never been formatted
4395 or if the filesystem is corrupt and cannot even mount then the
4396 error handling cleanup routines would dereference a NULL pointer.
4401 Description: NULL pointer deref in llap_from_page.
4402 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4403 incorrect) mapping if the page was truncated/reclaimed while it was
4404 searched for. Check for this condition and skip such pages when
4405 doing readahead. Introduce extra check to llap_from_page() to
4406 verify page->mapping->host is non-NULL (so page is not anonymous).
4409 Frequency : Sometimes when using sys_sendfile
4411 Description: "page not covered by a lock" warnings from ll_readpage
4412 Details : sendfile called ll_readpage without right page locks present.
4413 Now we introduced ll_file_sendfile that does necessary locking
4414 around call to generic_file_sendfile() much like we do in
4418 Frequency : with certain MDS communication failures at client mount time
4420 Description: NULL pointer deref after failed client mount
4421 Details : a client connection request may delayed by the network layer
4422 and not be sent until after the PTLRPC layer has timed out the
4423 request. If the client fails the mount immediately it will try
4424 to clean up before the network times out the request. Add a
4425 reference from the request import to the obd device and delay
4426 the cleanup until the network drops the request.
4429 Frequency : occasionally during client (re)connect
4431 Description: assertion failure during client (re)connect
4432 Details : processing a client connection request may be delayed by the
4433 client or server longer than the client connect timeout. This
4434 causes the client to resend the connection request. If the
4435 original connection request is replied in this interval, the
4436 client may trip an assertion failure in ptlrpc_connect_interpret()
4437 which thought it would be the only running connect process.
4440 Frequency : only with obd_echo servers and clients that are rebooted
4442 Description: kernel BUG accessing uninitialized data structure
4443 Details : When running an obd_echo server it did not start the ping_evictor
4444 thread, and when a client was evicted an uninitialized data
4445 structure was accessed. Start the ping_evictor in the RPC
4446 service startup instead of the OBD startup.
4448 Severity : enhancement
4449 Bugzilla : 10193 (patchless)
4450 Description: Remove dependency on various unexported kernel interfaces.
4451 Details : No longer need reparent_to_init, exit_mm, exit_files,
4452 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4455 Frequency : rare (only users of deprecated and unsupported LDAP config)
4457 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4458 Details : LDAP apparently contains 'lustreName' attributes instead of
4459 'name'. A simple remapping of the name is sufficient.
4462 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4464 Description: waiting_locks_callback trips kernel BUG if client is evicted
4465 Details : Running with the dump_on_timeout debug flag turned on makes
4466 it possible that the waiting_locks_callback() can try to dump
4467 the Lustre kernel debug logs from an interrupt handler. Defer
4468 this log dumping to the expired_lock_main() thread.
4470 Severity : enhancement
4472 Description: Support NFS exporting on 2.6 kernels.
4473 Details : Implement non-rawops metadata methods for NFS server to use without
4474 changing NFS server code.
4477 Frequency : very rare (synthetic metadata workload only)
4479 Description: two racing renames might cause an MDS thread to deadlock
4480 Details : Running the "racer" program may cause one MDS thread to rename
4481 a file from being the source of a rename to being the target of
4482 a rename at exactly the same time that another thread is doing
4483 so, and the second thread has already enqueued these locks after
4484 doing a lookup of the target and is trying to relock them in
4485 order. Ensure that we don't try to re-lock the same resource.
4488 Frequency : only very large systems with liblustre clients
4490 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4491 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4492 all OSTs in parallel. This allows the eviction of stale liblustre
4493 clients to proceed much faster than if they were done in series,
4494 and also offers similar improvements for other set_info RPCs.
4499 Description: excessive CPU usage during initial read phase on client
4500 Details : During the initial read phase on a client, it would agressively
4501 retry readahead on the file, consuming too much CPU and impacting
4502 performance (since 1.4.5.8). Improve the readahead algorithm
4503 to avoid this, and also improve some other common cases (read
4504 of small files in particular, where "small" is files smaller than
4505 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4510 Description: MDS crash when receiving packet with unknown intent.
4511 Details : Do not LBUG in unknown intent case, just return -EFAULT
4513 Severity : enhancement
4514 Bugzilla : 9293, 9385
4515 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4516 Details : Do not serialize getattr (non-intent version) and statfs.
4519 Frequency : occasional, when OST network is overloaded/intermittent
4521 Description: client evicted by OST after bulk IO timeout
4522 Details : If a client sends a bulk IO request (read or write) the OST
4523 may evict the client if it is unresposive to its data GET/PUT
4524 request. This is incorrect if the network is overloaded (takes
4525 too long to transfer the RPC data) or dropped the OST GET/PUT
4526 request. There is no need to evict the client at all, since
4527 the pinger and/or lock callbacks will handle this, and the
4528 client can restart the bulk request.
4531 Frequency : Always when mmapping file with no objects
4533 Description: client crashes when mmapping file with no objects
4534 Details : Check that we actually have objects in a file before doing any
4535 operations on objects in ll_vm_open, ll_vm_close and
4541 Description: Request leak when working with deleted CWD
4542 Details : Introduce advanced request refcount tracking for requests
4543 referenced from lustre intent.
4545 Severity : Enhancement
4547 Description: Cache open file handles on client.
4548 Details : MDS now will return special lock along with openhandle, if
4549 requested and client is allowed to hold openhandle, even if unused,
4550 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4551 closing files for every read/write openration.
4553 Severity : Enhancement
4555 Description: Cache open negative dentries on client when possible.
4556 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4557 negative dentries on lock revocation.
4562 Description: Remounting a client read-only wasn't possible with a zconf mount
4563 Details : It wasn't possible to remount a client read-only with llmount.
4565 Severity : enhancement
4566 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4567 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4568 MPICH is included to add Lustre-specific ADIO interfaces.
4569 This is based closely on the UFS ADIO layer and only differs
4570 in file creation, in order to allow the OST striping to be set.
4571 This is user-contributed code and not supported by CFS.
4576 Description: extended inode attributes (immutable, append-only) work improperly
4577 when 2.4 and 2.6 kernels are used on client/server or vice versa
4578 Details : Introduce kernel-independent values for these flags.
4580 Severity : enhancement
4583 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4584 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4585 granularity. Now, Fractional MB granularity is be supported,
4586 this is very useful for low memory system.
4588 Severity : enhancement
4590 Description: Getattr by fid
4591 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4592 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4593 have locks and avoids getattr rpc requests for it on every stat.
4596 Frequency : Always, for filesystems larger than 2TB
4598 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4599 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4600 filesystems larger than 2TB to be mounted with mballoc enabled.
4603 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4605 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4606 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4607 filesystems larger than 2TB is mounted then the kernel will
4608 silently corrupt the start of the filesystem. CONFIG_LBD is
4609 enabled for all CFS-supported kernels, but the possibility of
4610 this happening with a modified kernel config exists.
4612 Severity : enhancement
4614 Description: add client O_DIRECT support for 2.6 kernels
4615 Details : It is now possible to do O_DIRECT reads and writes to files
4616 in the Lustre client mountpoint on 2.6 kernel clients.
4618 Severity : enhancement
4620 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4621 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4622 parallel, not waiting for response from every OST before sending
4623 a rpc to the next OST.
4628 Description: setattr vs write race when updating file timestamps
4629 Details : Client processes that update a file timestamp into the past
4630 right after writing to the file (e.g. tar) it is possible that
4631 the updated file modification time can be reset to the current
4632 time due to a race between processing the setattr and write RPC.
4634 Severity : enhancement
4636 Description: Bring 'lfs find' closer in line with regular Linux find.
4637 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4638 -print0 options and obtains all the needed info through the lustre
4641 Severity : enhancement
4643 Description: support up to 1024 configured devices on one node
4644 Details : change obd_dev array from statically allocated to dynamically
4645 allocated structs as they are first used to reduce memory usage
4650 Description: Flush dirty partially truncated pages during truncate
4651 Details : Immediatelly flush partially truncated pages in filter_setattr,
4652 this way we completely avoid having any pages in page cache on OST
4653 and can retire ugly workarounds during writes to flush such pages.
4658 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4659 Details : protect inode from truncation within vfs_unlink() context
4660 just take a reference before calling vfs_unlink() and release it
4661 when parent's i_sem is free.
4666 Description: last_id value checked outside lock on OST caused LASSERT failure
4667 Details : If there were multiple MDS->OST object precreate requests in
4668 flight, it was possible that the OST's last object id was checked
4669 outside a lock and incorrectly tripped an assertion. Move checks
4670 inside locks, and discard old precreate requests.
4673 Frequency : always, if extents are used on OSTs
4675 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4676 Details : OSTs using the extents format would not zero the high 16 bits of
4677 the index physical block number. This is not a problem for any
4678 OST filesystems smaller than 16TB, and no kernels support ext3
4679 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4680 new/modified files) and can be fixed for existing filesystems
4681 with e2fsprogs-1.39-cfs1.
4686 Description: import connection selection may be incorrect if timer wraps
4687 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4688 import connections to be ignored if the 32-bit jiffies counter
4689 wraps. Use a 64-bit jiffies counter.
4692 Frequency : very large clusters immediately after boot
4694 Description: LNET request buffers exhausted under heavy short-term load
4695 Details : If a large number of client requests are generated on a service
4696 that has previously never seen so many requests it is possible
4697 that the request buffer growth cannot keep up with the spike in
4698 demand. Instead of dropping incoming requests, they are held in
4699 the LND until the RPC service can accept more requests.
4702 Frequency : Sometimes during replay
4704 Description: Assertion failure in ll_local_open after replay.
4705 Details : If replay happened on an open request reply before we were able
4706 to set replay handler, reply will become not swabbed tripping the
4707 assertion in ll_local_open. Now we set the handler right after
4708 recognising of open request
4711 Frequency : very rare
4713 Description: kernel reports "badness in vsnprintf"
4714 Details : Reading from the "recovery_status" /proc file in small chunks
4715 may cause a negative length in lprocfs_obd_rd_recovery_status()
4716 call to vsnprintf() (which is otherwise harmless). Exit early
4717 if there is no more space in the output buffer.
4719 Severity : enhancement
4721 Description: clear OBD RPC statistics by writing to them
4722 Details : It is now possible to clear the OBD RPC statistics by writing
4723 to the "stats" file.
4728 Description: Client mtime is not the same on different clients after utimes
4729 Details : In some cases, the client was using the utimes() syscall on
4730 a file cached on another node. The clients now validate the
4731 ctime from the MDS + OSTs to determine which one is right.
4736 Description: Inability to activate failout mode
4737 Details : lconf script incorrectly assumed that in pythong string's numeric
4738 value is used in comparisons.
4741 Frequency : always with multiple stripes per file
4743 Description: Inefficient object allocation for mutli-stripe files
4744 Details : When selecting which OSTs to stripe files over, for files with
4745 a stripe count that divides evenly into the number of OSTs,
4746 the MDS is always picking the same starting OST for each file.
4747 Return the OST selection heuristic to the original design.
4752 Description: mount failures may take full timeout to return an error
4753 Details : Under some heavy load conditions it is possible that a
4754 failed mount can wait for the full obd_timeout interval,
4755 possibly several minutes, before reporting an error.
4756 Instead return an error as soon as the status is known.
4758 Frequency : quota enabled and large files being deleted
4760 Description: releasing more than 4GB of quota at once hangs OST
4761 Details : If a user deletes more than 4GB of files on a single OST it
4762 will cause the OST to spin in an infinite loop. Release
4763 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4768 Description: statfs data retrieved from /proc may be stale or zero
4769 Details : When reading per-device statfs data from /proc, in the
4770 {kbytes,files}_{total,free,avail} files, it may appear
4771 as zero or be out of date.
4774 Frequency : systems with MD RAID1 external journal devices
4776 Description: lconf's call to blkid is confused by RAID1 journal devices
4777 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4778 of returning all block devices that match the journal UUID.
4781 Frequency : always, for aggregate stripe size over 4GB
4783 Description: assertion fails when trying to use 4GB stripe size
4784 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4785 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4788 Frequency : always on ppc64
4790 Description: the first write on an ext3 filesystem with mballoc got stuck
4791 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4792 perform endianness conversion.
4794 ------------------------------------------------------------------------------
4796 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4798 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4799 INTEROPERATE with older versions automatically. Please read the
4800 user documentation before upgrading any part of a live system.
4801 * WARNING: Lustre networking configuration changes are required with
4802 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4805 * Support for kernels:
4806 2.6.9-22.0.2.EL (RHEL 4)
4807 2.6.5-7.244 (SLES 9)
4808 2.6.12.6 vanilla (kernel.org)
4811 Severity : enhancement
4812 Bugzilla : 7981/8208
4813 Description: Introduced Lustre Networking (LNET)
4814 Details : LNET is new networking infrastructure for Lustre, it includes
4815 a reorganized network configuration mode (see the user
4816 documentation for full details) as well as support for routing
4817 between different network fabrics. Lustre Networking Devices
4818 (LNDs) for the supported network fabrics have also been
4819 created for this new infrastructure.
4821 Severity : enhancement
4822 Description: Introduced Access control lists
4823 Details : clients can set ACLs on files and directories in order to have
4824 more fine-grained permissions than the standard Unix UGO+RWX.
4825 The MDS must be started with the "-o acl" mount option.
4827 Severity : enhancement
4828 Description: Introduced filesystem quotas
4829 Details : Administrators may now establish per-user quotas on the
4832 Severity : enhancement
4834 Description: Configuration change for the XT3
4835 The PTLLND is now used to run Lustre over Portals on the XT3
4836 The configure option(s) --with-cray-portals are no longer used.
4837 Rather --with-portals=<path-to-portals-includes> is used to
4838 enable building on the XT3. In addition to enable XT3 specific
4839 features the option --enable-cray-xt3 must be used.
4844 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4845 Details : A race between allocating a new llcd and re-getting the llcd_lock
4846 allowed another thread to grab newly-allocated llcd.
4848 Severity : enhancement
4850 Description: 2.6 OST async journal commit and locking fix to improve performance
4851 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4852 2.6 kernels are now async as they already were in 2.4 kernels so
4853 that they can commit concurrently with the network bulk transfer.
4854 For block-allocated files the filter allocation semaphore is held
4855 to avoid filesystem fragmentation during allocation. BKL lock
4856 removed for 2.6 xattr operations where it is no longer needed.
4861 Description: lconf incorrectly determined whether two IP networks could talk
4862 Details : In some more complicated routing and multiple-network
4863 configurations, lconf will avoid trying to make a network
4864 connection to a disjoint part of the IP space. It was doing the
4865 math incorrectly for one set of cases.
4870 Description: Fix for potential infinite loop processing records in an llog.
4871 Details : If an llog record is corrupted/zeroed, it is possible to loop
4872 forever in llog_process(). Validate the llog record length
4873 and skip the remainder of the block on error.
4876 Frequency : occasional (liblustre only)
4878 Description: liblustre could not open files whose last component is a symlink
4879 Details : sysio_path_walk() would incorrectly pass the open intent to
4880 intermediate path components.
4883 Frequency : rare (liblustre only with non-standard tuning)
4884 Bugzilla : 7201 (7350)
4885 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4886 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4887 after it has been released, only do this for referenced locks
4889 Severity : enhancement
4891 Description: specifying an (invalid) directory default stripe_size of -1
4892 would reset the directory default striping
4893 Details : stripe_size -1 was used internally to signal directory stripe
4894 removal, now use "all default" to signal dir stripe removal
4895 as a directory striping of "all default" is not useful
4898 Frequency : common for large clusters running liblustre clients
4900 Description: doing an ls when liblustre clients are running is slow
4901 Details : sending a glimpse AST to a liblustre client waits for every AST
4902 to time out, as liblustre clients will not respond. Since they
4903 cannot cache data we refresh the OST lock LVB from disk instead.
4905 Severity : enhancement
4907 Description: doing an ls at the same time as file IO can be slow
4908 Details : enqueue and other "small" requests can be blocked behind many
4909 large IO requests. Create a new OST IO portal for non-IO
4910 requests so they can be processed faster.
4913 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4915 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4916 created file with mode 000
4917 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4918 allow an ftruncate() as a non-root user to a file with mode 000.
4919 Lustre now respects this flag to disable mode checks when
4920 truncating a file owned by the user
4923 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4926 Description: Revoking locks from clients that went dead or catatonic might take
4928 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4929 cancellation of such locks instant on servers without waiting for
4930 any reply from clients. Clients drops these locks when cancel
4931 notification from server is received without replying.
4934 Frequency : liblustre-only, when liblustre client dies or becomes busy
4936 Description: Doing ls on Linux clients can take a long time with active
4938 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4939 granting such locks to it in the first place if possible. Locks
4940 are taken by proxy on the OST during the read or write and
4941 dropped immediately afterward. Add connect flags handling, do
4942 not grant locks to liblustre clients for glimpse ASTs.
4944 Severity : enhancement
4946 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4947 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4948 sequential reads and behaves badly with applications doing random
4949 reads. Improve it by reading ahead at least read region, and
4950 avoiding excessive large RPC for small reads.
4952 Severity : enhancement
4954 Description: Creating more than 1000 files for a single job may cause a load
4955 imbalance on the OSTs if there are also a large number of OSTs.
4956 Details : qos_prep_create() uses an OST index reseed value that is an
4957 even multiple of the number of available OSTs so that if the
4958 reseed happens in the middle of the object allocation it will
4959 still utilize the OSTs as uniformly as possible.
4964 Description: OST or MDS may oops in ping_evictor_main()
4965 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4966 but doesn't restart at beginning of obd_exports_timed list
4969 Severity : enhancement
4971 Description: improve by-nid export eviction on the MDS and OST
4972 Details : allow multiple exports with the same NID to be evicted at one
4973 time without re-searching the exports list.
4976 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4978 Description: MDS may oops in groups_free()
4979 Details : in rare race conditions a newly allocated group_info struct is
4980 freed again, and this can be NULL. The 2.4 compatibility code
4981 for groups_free() checked for a NULL pointer, but 2.6 did not.
4984 Frequency : common for liblustre clients doing little filesystem IO
4985 Bugzilla : 9352, 7313
4986 Description: server may evict liblustre clients accessing contended locks
4987 Details : if a client is granted a lock or receives a completion AST
4988 with a blocking AST already set it would not reply to the AST
4989 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4992 Frequency : lfs setstripe, only systems with more than 160 OSTs
4994 Description: unable to set striping with a starting offset beyond OST 160
4995 Details : llapi_create_file() incorrectly limited the starting stripe
4996 index to the maximum single-file stripe count.
4999 Frequency : LDAP users only
5001 Description: lconf did not handle in-kernel recovery with LDAP properly
5002 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5004 Severity : enhancement
5006 Description: bind OST threads to NUMA nodes to improve performance
5007 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5008 node and do their allocations there to localize memory access
5010 Severity : enhancement
5012 Description: llmount can determine client NID directly from Myrinet (GM)
5013 Details : the client NID code from gmnalnid was moved directly into
5014 llmount, removing the need to use this or specifying the
5015 client NID explicitly when mounting GM clients with zeroconf
5018 Frequency : if client is started with down MDS
5020 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5021 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5022 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5023 ptlrpc_import_delayed_req() because we didn't check if the
5024 request was interrupted, and we also didn't break out of the
5025 event loop if there was no timeout
5030 Description: data loss during non-page-aligned writes to a single file from
5031 both multiple nodes and multiple threads on one node at same time
5032 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5033 inconsistency led to false short-reads, that were cached and later
5034 used by ->prepare_write() to fill in partially written page,
5035 leading to data loss.
5038 Frequency : always, if lconf --abort_recovery used
5040 Description: lconf --abort_recovery fails with 'Operation not supported'
5041 Details : lconf was attempting to abort recovery on the MDT device and not
5044 Severity : enhancement
5046 Description: remove cleanup logs
5047 Details : replace lconf-generated cleanup logs with lustre internal
5048 cleanup routines. Eliminates the need for client-cleanup and
5051 Severity : enhancement
5053 Description: add support for EAs (user and system) on lustre filesystems
5054 Details : it is now possible to store extended attributes in the Lustre
5055 client filesystem, and with the user_xattr mount option it
5056 is possible to allow users to store EAs on their files also
5058 Severity : enhancement
5060 Description: Add possibility (config option) to show minimal available OST free
5062 Details : When compiled with --enable-mindf configure option, statfs(2)
5063 (and so, df) will return least minimal free space available from
5064 all OSTs as amount of free space on FS, instead of summary of
5065 free spaces of all OSTs.
5067 Severity : enhancement
5069 Description: do not expand extent locks acquired on OST-side
5070 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5071 by server: they are not cached anyway.
5074 Frequency : when mmap is used/binaries executed from Lustre
5076 Description: Unmmap pages before throwing them away from read cache.
5077 Details : llap_shrink cache now attempts to unmap pages before discarding
5078 them (if unmapping failed - do not discard). SLES9 kernel has
5079 extra checks that trigger if this unmapping is not done first.
5084 Description: lconf didn't resolve symlinks before checking to see whether a
5085 given mountpoint was already in use
5088 Frequency : when migrating failover services
5089 Bugzilla : 6395, 9514
5090 Description: When migrating a subset of services from a node (e.g. failback
5091 from a failover service node) the remaining services would
5092 time out and evict clients.
5093 Details : lconf --force (implied by --failover) sets the global obd_timeout
5094 to 5 seconds in order to quickly disconnect, but this caused
5095 other RPCs to time out too quickly. Do not change the global
5096 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5098 Severity : enhancement
5099 Frequency : if MDS is started with down OST
5100 Bugzilla : 9439,5706
5101 Description: Allow startup/shutdown of an MDS without depending on the
5102 availability of the OSTs.
5103 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5104 Add appropriate locking and lov-osc refcounts for safe
5105 cleaning. Add osc abort_inflight calls in case the
5106 synchronize never started.
5109 Frequency : occasional (Cray XT3 only)
5111 Description: root not authorized to access files in CRAY_PORTALS environment
5112 Details : The client process capabilities were not honoured on the MDS in
5113 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5114 been accessed by an authorized user then root was able to access
5115 the file on the local client also. The root user capabilities
5116 are now allowed on the MDS, as this environment has secure UID.
5119 Frequency : occasional
5121 Description: ldiskfs "too long searching" message happens too often
5122 Details : A debugging message (otherwise harmless) prints too often on
5123 the OST console. This has been reduced to only happen when
5124 there are fragmentation problems on the filesystem.
5129 Description: Division by zero in statfs when all OSCs are inactive
5130 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5131 lov_statfs divides by value returned by lov_get_stripecnt().
5135 Bugzilla : 9489, 3273
5136 Description: First write from each client to each OST was only 4kB in size,
5137 to initialize client writeback cache, which caused sub-optimal
5138 RPCs and poor layout on disk for the first written file.
5139 Details : Clients now request an initial cache grant at (re)connect time
5140 and so that they can start streaming writes to the cache right
5141 away and always do full-sized RPCs if there is enough data.
5142 If the OST is rebooted the client also re-establishes its grant
5143 so that client cached writes will be honoured under the grant.
5148 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5149 Details : Now I/O RPCs go to different portal number and (presumably) fast
5150 lock requests (and glimses) and other RPCs get their own service
5151 threads pool that should be able to service those RPCs
5154 Severity : enhancement
5156 Description: Ability to exchange lustre version between client and servers and
5157 issue warnings at client side if client is too old. Also for
5158 liblustre clients there is ability to refuse connection of too old
5160 Details : New 'version' field is added to connect data structure that is
5161 filled with version info. That info is later checked by server and
5165 Frequency : rare, liblustre only.
5166 Bugzilla : 9296, 9581
5167 Description: Two simultaneous writes from liblustre at offset within same page
5168 might proceed at the same time overwriting eachother with stale
5170 Details : I/O lock withing llu_file_prwv was released too early, before data
5171 actually was hitting the wire. Extended lock-holding time until
5172 server acknowledges receiving data.
5175 Frequency : extremely rare. Never observed in practice.
5177 Description: avoid generating lustre_handle cookie of 0.
5178 Details : class_handle_hash() generates handle cookies by incrementing
5179 global counter, and can hit 0 occasionaly (this is unlikely, but
5180 not impossible, because initial value of cookie counter is
5181 selected randonly). Value of 0 is used as a sentinel meaning
5182 "unassigned handle" --- avoid it. Also coalesce two critical
5183 sections in this function into one.
5185 Severity : enhancement
5187 Description: allow liblustre clients to delegate truncate locking to OST
5188 Details : To avoid overhead of locking, liblustre client instructs OST to
5189 take extent lock in ost_punch() on client's behalf. New connection
5190 flag is added to handle backward compatibility.
5192 Severity : enhancement
5193 Bugzilla : 4928, 7341, 9758
5194 Description: allow number of OST service threads to be specified
5195 Details : a module parameter allows the number of OST service threads
5196 to be specified via "options ost ost_num_threads={N}" in the
5197 OSS's /etc/modules.conf or /etc/modprobe.conf.
5201 Bugzilla : 6146, 9635, 9895
5202 Description: servers crash with bad pointer in target_handle_connect()
5203 Details : In rare cases when a client is reconnecting it was possible that
5204 the connection request was the last reference for that export.
5205 We would temporarily drop the export reference and get a new
5206 one, but this may have been the last reference and the export
5207 was just destroyed. Get new reference before dropping old one.
5209 Severity : enhancement
5210 Frequency : if client is started with failover MDS
5212 Description: Allow multiple MDS hostnames in the mount command
5213 Details : Try to read the configuration from all specified MDS
5214 hostnames during a client mount in case the "primary"
5217 Severity : enhancement
5219 Description: Stop sending data to evicted clients as soon as possible.
5220 Details : Check if the client we are about to send or are sending data to
5221 was evicted already. (Check is done every second of waiting,
5222 for which l_wait_event interface was extended to allow checking
5223 of exit condition at specified intervals).
5226 Frequency : rare, normally only when NFS exporting is done from client
5228 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5230 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5233 Frequency : always, if srand() is called before liblustre initialization
5235 Description: Liblustre uses system PRNG disturbing its usage by user application
5236 Details : Introduce internal to lustre fast and high-quality PRNG for
5237 lustre usage and make liblustre and some other places in generic
5238 lustre code to use it.
5240 Severity : enhancement
5241 Bugzilla : 9477, 9557, 9870
5242 Description: Verify that the MDS configuration logs are updated when xml is
5243 Details : Check if the .xml configuration logs are newer than the config
5244 logs stored on the MDS and report an error if this is the case.
5245 Request --write-conf, or allow starting with --old_conf.
5247 Severity : enhancement
5249 Description: Handle symlinks in the path when checking if Lustre is mounted.
5250 Details : Resolve intermediate symlinks when checking if a client has
5251 mounted a filesystem to avoid duplicate client mounts.
5256 Description: lconf can hit an error exception but still return success.
5257 Details : The lconf command catches the Command error exception at the top
5258 level script context and will exit with the associated exit
5259 status, but doesn't ensure that this exit status is non-zero.
5264 Description: failure of ptlrpc thread startup can cause oops
5265 Details : Starting a ptlrpc service thread can fail if there are a large
5266 number of threads or the server memory is very fragmented.
5267 Handle this without oopsing.
5270 Frequency : always, only if liblustre and non-default acceptor port was used
5272 Description: liblustre cannot connect to servers with non-default acceptor port
5273 Details : tcpnal_set_default_params() was not called and was therefore
5274 ignoring the environment varaible TCPNAL_PORT, as well as other
5275 TCPNAL_ environment variables
5280 Description: two objects could be created on the same OST for a single file
5281 Details : If an OST is down, in some cases it was possible to create two
5282 objects on a single OST for a single file. No problems other
5283 than potential performance impact and spurious error messages.
5287 Bugzilla : 5681, 9562
5288 Description: Client may oops in ll_unhash_aliases
5289 Details : Client dcache may become inconsistent in race condition.
5290 In some cases "getcwd" can fail if the current directory is
5296 Description: Inode refcounting problems in NFS export code
5297 Details : link_raw functions used to call d_instantiate without obtaining
5298 extra inode reference first.
5302 Bugzilla : 9942, 9903
5303 Description: Referencing freed requests leading to crash, memleaks with NFS.
5304 Details : We used to require that call to ll_revalidate_it was always
5305 followed by ll_lookup_it. Also with revalidate_special() it is
5306 possible to call ll_revalidate_it() twice for the same dentry
5307 even if first occurence returned success. This fix changes semantic
5308 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5309 reference on a request referred from the intent.
5310 ll_intent_release() then releases such a request.
5313 Frequency : rare, normally benchmark loads only
5315 Description: unlinked inodes were kept in memory on the client
5316 Details : If a client is repeatedly creating and unlinking files it
5317 can accumulate a lot of stale inodes in the inode slab cache.
5318 If there is no other client load running this can cause the
5319 client node to run out of memory. Instead flush old inodes
5320 from client cache that have the same inode number as a new inode.
5323 Frequency : SLES9 2.6.5 kernel and long filenames only
5324 Bugzilla : 9969, 10379
5325 Description: utime reports stale NFS file handle
5326 Details : SLES9 uses out-of-dentry names in some cases, which confused
5327 the lustre dentry revalidation. Change it to always use the
5331 Frequency : rare, unless heavy write-truncate concurrency is continuous
5332 Bugzilla : 4180, 6984, 7171, 9963, 9331
5333 Description: OST becomes very slow and/or deadlocked during object unlink
5334 Details : filter_destroy() was holding onto the parent directory lock
5335 while truncating+unlinking objects. For very large objects this
5336 may block other threads for a long time and slow overall OST
5337 responsiveness. It may also be possible to get a lock ordering
5338 deadlock in this case, or run out of journal credits because of
5339 the combined truncate+unlink. Solution is to do object truncate
5340 first in one transaction without parent lock, and then do the
5341 final unlink in a new transaction with the parent lock. This
5342 reduces the lock hold time dramatically.
5345 Frequency : rare, 2.4 kernels only
5347 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5348 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5349 before unmounting their filesystem in order to avoid deadlock.
5350 The kernel_locked() function in 2.4 kernels only checks whether
5351 the kernel lock is held, not whether it is this process that is
5352 holding it as 2.6 kernels do.
5357 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5358 Details : The client ptlrpc code may be trying to reconnect to a down
5359 server before a previous connection attempt has timed out.
5360 Increase the reconnect interval to be longer than the connection
5361 timeout interval to avoid sending duplicate connections to
5365 Frequency : echo_client brw_test command
5367 Description: fix echo_client to work with OST preallocated code
5368 Details : OST preallocation code (5137) didn't take echo_client IO path
5369 into account: echo_client calls filter methods outside of any
5370 OST thread and, hence, there is no per-thread preallocated
5371 pages and buffers to use. Solution: hijack pga pages for IO. As
5372 a byproduct, this avoids unnecessary data copying.
5376 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5377 Description: Client can oops in mdc_commit_close() after open replay
5378 Details : It was possible for the MDS to return an open request with no
5379 transaction number in mds_finish_transno() if the client was
5380 evicted, but without actually returning an error. Clients
5381 would later try to replay that open and may trip an assertion
5382 Simplify the client close codepath, and always return an error
5383 from the MDS in case the open is not successful.
5386 Frequency : rare, 2.6 OSTs only
5388 Description: OST may deadlock under high load on fragmented files
5389 Details : If there was a heavy load and highly-fragmented OST filesystems
5390 it was possible to have all the OST threads deadlock waiting on
5391 allocation of biovecs, because the biovecs were not released
5392 until the entire RPC IO was completed. Instead, release biovecs
5393 as soon as they are complete to ensure forward IO progress.
5395 Severity : enhancement
5397 Description: Support for specifying external journal device at mount
5398 Details : If an OST or MDS device is formatted with an external journal
5399 device, this device major/minor is stored in the ext3 superblock
5400 and may not be valid for failover. Allow detecting and
5401 specifying the external journal at mount time.
5406 Description: Mounting an MDS with pending unlinked files may cause oops
5407 Details : target_finish_recovery() calls mds_postrecov() which returned
5408 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5409 considers this an error and immediately begins cleaning up the
5410 lov, just after starting the mds_lov process
5412 Severity : enhancement
5414 Description: Implement 'lfs df' to report actual free space on per-OST basis
5415 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5416 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5417 usage of objects; '-h' to report in human readable format.
5419 ------------------------------------------------------------------------------
5421 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5428 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5429 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5430 at startup during orphan unlinking. Instead, skip buddy update
5431 before setup, as it will be regenerated after recovery is complete.
5436 Description: If an OST is inactive, its locks might reference stale inodes.
5437 Details : lov_change_cbdata() must iterate over all namespaces, even if
5438 they are inactive to clear inode references from the lock.
5440 Severity : enhancement
5441 Frequency : occasional, if non-standard max_dirty_mb used
5443 Description: Client will block write RPCs if not enough grant
5444 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5445 then the client will block writes while waiting for another RPC
5446 to complete instead of consuming its dirty limit. With change
5447 we get improved performance when max_dirty_mb is small.
5449 Severity : enhancement
5450 Bugzilla : 3389, 6253
5451 Description: Add support for supplementary groups on the MDS.
5452 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5453 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5454 lookups for user supplementary groups into a cache.
5458 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5459 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5460 for liblustre O_EXCL is enforced by MDS.
5464 Description: readdir never returns NULL in liblustre.
5465 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5466 dentry in struct dirent.
5470 Frequency : liblustre only, depends on application IO pattern
5471 Description: liblustre clients evicted if not contacting servers
5472 Details : Don't put liblustre clients into the ping_evictor list, so
5473 they will not be evicted by the pinger ever.
5475 Severity : enhancement
5477 Description: Add ability to evict clients by NID from MDS.
5478 Details : By echoing "nid:$NID" string into
5479 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5480 $NID would be instantly evicted from this MDS and from all active
5481 OSTs connected to it.
5485 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5486 Details : lookup_it_finish() used to query file size from OSTs that was not
5491 Description: service threads change working directory to that of init
5492 Details : Starting lustre service threads may pin the working directory
5493 of the parent thread, making that filesystem busy. Threads
5494 now change to the working directory of init to avoid this.
5498 Frequency : during shutdown only
5499 Description: shutdown with a failed MDS or OST can cause unmount to hang
5500 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5505 Frequency : 2.6 only
5506 Description: chmod/chown may include an extra supplementary group
5507 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5508 supplementary group and if none is specified this is used.
5511 Bugzilla : 5479 (6816)
5513 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5514 Details : If lookup is in progress on a file that is unlinked we might try
5515 to revalidate the inode and fail in revalidate after lookup is
5516 complete and ll_file_open() enqueues the open again but
5517 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5520 Frequency : always, if lconf --abort_recovery used
5522 Description: lconf --abort_recovery fails with 'Operation not supported'
5523 Details : lconf was attempting to abort recovery on the MDT device and not
5526 ------------------------------------------------------------------------------
5528 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5533 Frequency : rare (only unsupported configurations with a node running as an
5535 Bugzilla : 6514, 5137
5536 Description: Mounting a Lustre file system on a node running as an OST could
5538 Details : OSTs now preallocates memory needed to write out data at
5539 startup, instead of when needed, to avoid having to
5540 allocate memory in possibly low memory situations.
5541 Specifically, if the file system is mounted on on OST,
5542 memory pressure could force it to try to write out data,
5543 which it needed to allocate memory to do. Due to the low
5544 memory, it would be unable to do so and the node would
5545 become unresponsive.
5547 Severity : enhancement
5549 Description: Addition of lconf --service command line option
5550 Details : lconf now accepts a '--service <arg>' option, which is
5551 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5553 Severity : enhancement
5555 Description: Failover mode is now the default for OSTs.
5556 Details : By default, OSTs will now run in failover mode. To return to
5557 the old behaviour, add '--failout' to the lmc line for OSTs.
5559 Severity : enhancement
5561 Description: Health checks are now provided for MDS and OSTs
5562 Details : Additional detailed health check information on MSD and OSTs
5563 is now provided through the procfs health_check value.
5566 Frequency : occasional, depends on IO load
5568 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5569 after numerous create/delete cycles
5570 Details : The ext3 inode allocation policy would not allocate new inodes
5571 very well on the OSTs because there are no new directories
5572 being created. Instead we look for groups with free space if
5573 the parent directories are nearly full.
5578 Description: Network or server problems during mount may cause partially
5579 mounted clients instead of returning an error.
5580 Details : The config llog parsing code may overwrite the error return
5581 code during mount error handling, returning success instead
5587 Description: MDS can fail to allocate large reply buffers
5588 Details : After long uptimes the MDS can fail to allocate large reply
5589 buffers (e.g. zconf client mount config records) due to memory
5590 fragmentation or consumption by the buffer cache. Preallocate
5591 some large reply buffers so that these replies can be sent even
5592 under memory pressure.
5596 Frequency : rare (liblustre)
5597 Description: fsx running with liblustre complained that using truncate() to
5598 extend the file doesn't work. This patch corrects that issue.
5599 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5600 ATTR_SIZE and lsm use in llu_setattr_raw.
5604 Frequency : rare, only 2.6 kernels
5605 Description: Unusual file access patterns on the MDS may result in inode
5606 data being lost in very rare circumstances.
5607 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5608 code in the 2.6 kernel caused the inode and/or EA data not to
5609 be read from disk, causing single-file corruption.
5613 Frequency : rare, only 2.6 filesystems using extents
5614 Description: Heavy concurrent write and delete load may cause data corruption.
5615 Details : It was possible under high-load situations to have an extent
5616 metadata block in the block device cache from a just-unlinked
5617 file overwrite a newly-allocated data block. We now unmap any
5618 metadata buffers that alias just-allocated data blocks.
5622 Frequency : filesystems with default stripe_count larger than 77
5623 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5624 Details : lconf specifies an inode size of 4096 bytes when the default
5625 stripe_count is larger than 77. This conflicts with the default
5626 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5627 this case to avoid pinning too much memory for large EAs.
5629 ------------------------------------------------------------------------------
5631 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5636 Frequency : rare (extremely heavy IO load with hundreds of clients)
5638 Description: Client is evicted, gets IO error writing to file
5639 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5640 caused clients to be evicted for AST timeouts. The fixes in
5641 bug 5192 mean we no longer need to have such short AST timeouts
5642 so ldlm_timeout has been increased.
5645 Frequency : occasional during --force or --failover shutdown under load
5646 Bugzilla : 5949, 4834
5647 Description: Server oops/LBUG if stopped with --force or --failover under load
5648 Details : a collection of import/export refcount and cleanup ordering
5649 issues fixed for safer force cleanup
5652 Frequency : only filesystems larger than 120 OSTs
5653 Bugzilla : 5990, 6223
5654 Description: lfs getstripe would oops on a very large filesystem
5655 Details : lov_getconfig used kfree on vmalloc'd memory
5658 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5659 Bugzilla : 6242, 6243
5660 Description: reading from files that had been truncated to a non-zero size
5661 but never opened returned no data
5662 Details : ll_file_read() reads zeros from no-object files to EOF
5667 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5668 Details : There's little harm in aborting MDS/OSS recovery and letting it
5669 try again, so I removed the LASSERT and return an error instead.
5671 Severity : enhancement
5673 Description: New debugging infrastructure for tracking down data corruption
5674 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5675 (b) cover more of the client-side code path, and (c) try to narrow
5676 down where problems occurred
5680 Bugzilla : 3819, 4364, 4397, 6313
5681 Description: Racing close and eviction MDS could cause assertion in mds_close
5682 Details : It was possible to get multiple mfd references during close and
5683 client eviction, leading to one thread referencing a freed mfd.
5685 Severity: : enhancement
5686 Bugzilla : 3262, 6359
5687 Description: Attempts to reconnect to servers are now more aggressive.
5688 Details : This builds on the enhanced upcall-less recovery that was added
5689 in 1.4.2. When trying to reconnect to servers, clients will
5690 now try each server in the failover group every 10 seconds. By
5691 default, clients would previously try one server every 25 seconds.
5696 Description: After recovery, certain operations trigger a failed
5697 assertion on a client.
5698 Details : Failing over an mds, using lconf -d --failover, while a
5699 client was doing a readdir() call would cause the client to
5700 LBUG after recovery completed and the readdir() was resent.
5702 Severity : enhancement
5704 Description: Default groups are now added by lconf
5705 Details : You can now run lconf --group <servicename> without having to
5706 manually add groups with lmc.
5709 Frequency : occasional
5711 Description: Nodes with an elan id of 0 trigger a failed assertion
5714 Frequency : always when accessing e.g. tty/console device nodes
5716 Description: tty and some other devices nodes cannot be used on lustre
5717 Details : file's private_data field is used by device data and lustre
5718 values in there got lost. New field was added to struct file to
5719 store fs-specific private data.
5722 Frequency : when exporting Lustre via NFS
5724 Description: NFSD failed occasionally when looking up a path component
5725 Details : NFSD is looking up ".." which was broken in ext3 directories
5726 that had grown large enough to become hashed.
5729 Frequency : Clusters with multiple interfaces not on the same subnet
5731 Description: Nodes will repeatedly try to reconnect to an interface which it
5732 cannot reach and report an error to the log.
5733 Details : Extra peer list entries will be created by lconf with some peers
5734 unreachable. lconf now validates the peer before adding it.
5737 Frequency : Only if a default stripe is set on the filesystem root.
5739 Description: Setting a default stripe on the filesystem root prevented the
5740 filesystem from being remounted.
5741 Details : The client was sending extra request flags in the root getattr
5742 request and did not allocate a reply buffer for the dir EA.
5745 Frequency : occasional, higher if lots of files are accessed by one client
5746 Bugzilla : 6159, 6097
5747 Description: Client trips assertion regarding lsm mismatch/magic
5748 Details : While revalidating inodes the VFS looks up inodes with ifind()
5749 and in rare cases can find an inode that is being freed.
5750 The ll_test_inode() code will free the lsm during ifind()
5751 when it finds an existing inode and then the VFS later attaches
5752 this free lsm to a new inode.
5756 Bugzilla : 6422, 7030
5757 Description: MDS deadlock between mkdir and client eviction
5758 Details : Creating a new file via mkdir or mknod (starting a transaction
5759 and getting the ns lock) can deadlock with client eviction
5760 (gets ns lock and trying to finish a synchronous transaction).
5763 Frequency : occasional
5764 Description: While starting a server, the fsfilt_ext3 module could not be
5766 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5767 kernels. Previously, lconf would still use the ext3 name
5768 when trying to load modules. Now, it will correctly use
5769 ext3 on 2.4 and ldiskfs on 2.6.
5771 Severity : enhancement
5772 Description: The default stripe count has been changed to 1
5773 Details : The interpretation of the default stripe count (0, to lfs
5774 or lmc) has been changed to mean striping across a single
5775 OST, rather than all available. For general usage we have
5776 found a stripe count of 1 or 2 works best.
5778 Severity : enhancement
5779 Description: Add support for compiling against Cray portals.
5780 Details : Conditional compiling for some areas that are different
5784 Frequency : occasional
5785 Bugzilla : 6409, 6834
5786 Description: Creating files with an explicit stripe count may lead to
5787 a failed assertion on the MDS
5788 Details : If some OSTs are full or unavailable, creating files may
5789 trigger a failed assertion on the MDS. Now, Lustre will
5790 try to use other servers or return an error to the
5794 Frequency : occasional
5796 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5797 caused lock timeout during readahead (since 1.4.2).
5798 Details : Processes doing readahead might match a lock that hasn't been
5799 granted yet if there are overlapping and conflicting lock
5800 requests. The readahead process waits on ungranted lock
5801 (original lock is CBPENDING), while OST waits for that process
5802 to cancel CBPENDING read lock and eventually evicts client.
5804 Severity : enhancement
5806 Description: Initial enabling of flock support for clients
5807 Details : Implements fcntl advisory locking and file status functions.
5808 This feature is provided as an optional mount flag (default
5809 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5810 locking are implemented yet, and those that are are not guaranteed
5811 to be completely correct in production environments.
5812 mount -t lustre -o [flock|noflock] ...
5815 Frequency : occasional
5817 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5818 assertion in the ext3 JBD (journaling) layer.
5819 Details : The b_committed_data struct is protected by the big kernel lock
5820 in 2.4 kernels, serializing journal_commit_transaction() and
5821 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5822 access to this struct. In 2.6 kernels there is finer grained
5823 locking to improve SMP performance of the JBD layer.
5827 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5832 Description: merge and cleanup kernel patches.
5835 -----------------------------------------------------------------------------
5837 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5839 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5840 versions of Lustre. Please update all servers and clients to
5841 version 1.4.2 or later at the same time. You must also run
5842 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5843 to update the configuration logs.
5845 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5846 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5847 - swab llog records of type '0' so we get proper header size/idx (5861)
5848 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5849 - fix rename of one directory over another leaking an inode (5953)
5850 - avoid SetPageDirty on 2.6 (5981)
5851 - don't re-add just-being-destroyed locks to the waiting list (5653)
5852 - when creating new directories, inherit the parent's custom
5853 striping settings if present parent (3048)
5854 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5855 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5856 - don't hold client locks on temporary worklist from l_lru (5666)
5857 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5858 - automatically evict dead clients (5921)
5859 - Update file size properly in create+truncate+fstat case (6196)
5860 - Do not unhash mountpoint dentries, do not allow removal of
5862 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5863 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5864 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5865 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5866 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5867 - Correctly handle reads of files with no objects (6243)
5868 - lctl recover will also mark a device active if deactivate used (5933)
5870 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5871 - allow --write-conf on an MDS with different nettype than client (5619)
5872 - don't write config llogs to MDS for mounts not from that MDS (5617)
5873 - lconf should create multiple TCP connections from a client (5201)
5874 - init scripts are now turned off by default; run chkconfig --on
5875 lustre and chkconfig --on lustrefs to use them
5876 - upcalls are no longer needed for clients to recover to failover
5878 - add --abort-recovery option to lconf to abort recovery on device
5880 - add support for an arbitrary number of OSTs (3026)
5881 - Quota support protocol changes.
5882 - forward compatibility changes to wire structs (6007)
5883 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5884 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5885 - improved "lustre status" script
5886 - initialize blocksize for non-regular files (6062)
5887 - added --disable-server and --disable-client configure options (5782)
5888 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5889 - Vanilla 2.4.29 support
5890 - increase maximum number of obd devices to 520 (6242)
5891 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5892 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5893 SLES 9 2.6.5 kernels
5894 - limit stripes per file to 160 (the maximum EA size) (6093)
5896 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5899 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5900 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5901 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5902 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5903 - small ext3 extents cleanups and fixes (5733)
5904 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5905 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5906 - avoid needless client->OST connect, fix handle mismatch (5317)
5907 - fix DLM error path that led to out-of-sync client, long delays (5779)
5908 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5909 - fix several locking issues related to i_size (5492,5624,5654,5672)
5910 - don't move pending lock onto export if it is already evicted (5683)
5911 - fix kernel oops when creating .foo in unlinked directory (5548)
5912 - fix deadlock in obdfilter statistics vs. object create (5811)
5913 - use time_{before,after} to avoid timer jiffies wrap (5882)
5914 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5915 - Do not leak request if server was not able to process it (5154)
5916 - If mds_open unable to find parent dir, make that negative lookup(5154)
5917 - don't create new directories with extent-mapping (5909, 5936)
5919 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5920 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5921 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5922 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5923 - support for RHEL4 kernel on i686 (5773)
5924 - provide error messages when incompatible logs are encountered (5898)
5926 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5927 * version 1.4.0.10 (1.4.1 release candidate 1)
5929 - don't keep a lock reference when lock is not granted (4238)
5930 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5931 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5932 - drop import inflight refcount on signal_completed_replay error (5255)
5933 - unlock page after async write error during send (3677)
5934 - handle missing objects in filter_preprw_read properly (5265)
5935 - no transno return for symlink open, don't save no-trasno open (3440)
5936 - don't try to complete elan receive that already failed (4012)
5937 - free RPC server reply state on error (5406)
5938 - clean up thread from ptlrpc_start_thread() on error (5160)
5939 - readahead could read extra page into cache that wasn't ejected (5388)
5940 - prevent races in class_attach/setup/cleanup/detach (5260)
5941 - don't dereference de->d_inode after l_dput of de (5458)
5942 - use "int" for stripe value returned from lock_to_stripe (5544)
5943 - mballoc allocation and error-checking fixes in 2.6 (5504)
5944 - block device patches to fix I/O request sizes in 2.6 (5482)
5945 - look up hostnames for IB nals (5602)
5946 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5947 - don't start multiple acceptors for the same port (5277)
5948 - fix incorrect LASSERT in mds_getattr_name (5635)
5949 - export a proc file for general "ping" checking (5628)
5950 - fix "lfs check" to not block when the MDS is down (5628)
5952 - service request history (4965)
5953 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5954 - create an "evict_client" /proc entry on OSTs, like the MDS has
5955 - fix mount usage message, return errors per mount(8) (5168)
5956 - change grep [] to grep "[]" in tests so they work in more UMLs
5957 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5958 - remove ancient LOV_MAGIC_V0 EA support (5047)
5959 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5960 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5961 - allow readahead allocations to fail when low on memory (5383)
5962 - mmap locking landed again, after considerable improvement (2828)
5963 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5965 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5968 - send OST transaction number in read/write reply to free req (4966)
5969 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5970 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5971 - the watchdog thread now runs as interruptible (5246)
5972 - flock/lockf fixes (but it's still disabled, pending 5135)
5973 - don't use EXT3 constants in llite code (5094)
5974 - memory shortage at startup could cause assertion (5176)
5976 - reorganization of lov code
5977 - single portals codebase
5979 - add extents/mballoc support (5025)
5980 - direct I/O reads in the obdfilter (4048)
5981 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5983 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5986 - send OST transaction number in read/write reply to free req (4966)
5987 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5988 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5989 - the watchdog thread now runs as interruptible (5246)
5990 - handle missing objects in filter_preprw_read properly (5265)
5991 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5992 - drop import inflight refcount on signal_completed_replay error (5255)
5993 - unlock page after async write error during send (3677)
5994 - return original error code on reconstructed replies (3761)
5995 - no transno return for symlink open, don't save no-trasno open (3440)
5997 - add pid to ldlm debugging output (4922)
5998 - bump the watchdog timeouts -- we can't handle 30sec yet
5999 - extra debugging for orphan dentry/inode bug (5259)
6001 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6004 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6005 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6006 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6007 - fix NULL dereference in /proc/sys/portals/routes (4827)
6008 - allow failed mdc_close() operations to be interrupted (4561)
6009 - stop precreate on OST before MDS would time out on it (4778)
6010 - don't send partial-page writes before EOF from client (4410)
6011 - discard client grant for sub-page writes on large-page clients (4520)
6012 - don't free dentries not owned by NFS code, check generation (4806)
6013 - fix lsm leak if mds_create_objects() fails (4801)
6014 - limit debug_daemon file size, always print CERROR messages (4789)
6015 - use transno after validating reply (3892)
6016 - process timed out requests if import state changes (3754)
6017 - update mtime on OST during writes, return in glimpse (4829)
6018 - add mkfsoptions to LDAP (4679)
6019 - use ->max_readahead method instead of zapping global ra (5039)
6020 - don't interrupt __l_wait_event() during strace
6022 - add software watchdogs to catch hung threads quickly (4941)
6023 - make lustrefs init script start after nfs is mounted
6024 - fix CWARN/ERROR duplication (4930)
6025 - return async write errors to application if possible (2248)
6026 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6027 - print NAL number in %x format (4645)
6028 - update barely-supported suse-2.4.21-171 series (4842)
6029 - support for sles 9 %post scripts
6030 - support for building 2.6 kernel-source packages
6031 - support for sles km_* packages
6033 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6036 - ignore -ENOENT errors in osc_destroy (3639)
6037 - notify osc create thread that OSC is being cleaned up (4600)
6038 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6039 - reconstruct ost_handle() like mds_handle() (4657)
6040 - create a new thread to do import eviction to avoid deadlock (3969)
6041 - let lconf resolve symlinked-to devices (4629)
6042 - don't unlink "objects" from directory with default EA (4554)
6043 - hold socknal file ref over connect in case target is down (4394)
6044 - allow more than 32000 subdirectories in a single directory (3244)
6045 - fix blocks count for O_DIRECT writes (3751)
6046 - OST returns ENOSPC from object create when no space left (4539)
6047 - don't send truncate RPC if file size isn't changing (4410)
6048 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6049 - bind to privileged port in socknal and tcpnal (3689)
6051 - rate limit CERROR/CWARN console message to avoid overload (4519)
6052 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6053 - basic mmap support (3918)
6054 - kernel patch series update from b1_4 (4711)
6056 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6059 - avoid crash during MDS cleanup with OST shut down (2775)
6060 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6061 - don't use bad inodes on the MDS (3744)
6062 - dynamic object preallocation to improve recovery speed (4236)
6063 - don't hold spinlock over lock dumping or change debug flags (4401)
6064 - don't zero obd_dev when it is force cleaned (3651)
6065 - print grants to console if they go negative (4431)
6066 - "lctl deactivate" will stop automatic recovery attempts (3406)
6067 - look for existing locks in ldlm_handle_enqueue() (3764)
6068 - don't resolve lock handle twice in recovery avoiding race (4401)
6069 - revalidate should check working dir is a directory (4134)
6071 - don't always mark "slow" obdfilter messages as errors (4418)
6073 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6076 - don't close LustreDB during write_conf until it is done (3860)
6077 - fix typo in lconf for_each_profile (3821)
6078 - allow dumping logs from multiple threads at one time (3820)
6079 - don't allow multiple threads in OSC recovery (3812)
6080 - fix debug_size parameters (3864)
6081 - fix mds_postrecov to initialize import for llog ctxt (3121)
6082 - replace config semaphore with spinlock (3306)
6083 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6084 - don't allow enqueue to complete on a destroyed export (3822)
6085 - down write_lock before checking llog header bitmap (3825)
6086 - recover from lock replay timeout (3764)
6087 - up llog sem before sending rpc (3652)
6088 - reduce ns lock hold times when setting kms (3267)
6089 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6090 - fix NULL deref and obd_dev leak on setup error (3312)
6091 - replace some LBUG about llog ops with error handling (3841)
6092 - don't match INVALID dentries from d_lookup and spin (3784)
6093 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6094 - fix invalid assertion in ptlrpc_set_wait (3880)
6096 - add libwrap support for the TCP acceptor (3996)
6097 - add /proc/sys/portals/routes for non-root route listing (3994)
6098 - allow setting MDS UUID in .xml (2580)
6099 - print the stack of a process that LBUGs (4228)
6101 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6104 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6105 - make sure to unset replay flag from failed open requests (3440)
6106 - if default stripe count is 0, use OST count for inode size (3636)
6107 - update parent mtime/ctime on client for create/unlink (2611)
6108 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6109 - free recovery state on server during a forced cleanup (3571)
6110 - unregister_reply for resent reqs (3063)
6111 - loop back devices mounting and status check on 2.6 (3563)
6112 - fix resource-creation race that can provoke i_size == 0 (3513)
6113 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6114 - more debugging for page-accounting assertion (3746)
6115 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6116 - avoid deadlock after precreation failure (3758)
6117 - fix race and lock order deadlock in orphan handling (3450, 3750)
6118 - add validity checks when grabbing inodes from l_ast_data (3599)
6120 - add /proc/.../recovery_status to obdfilter (3428)
6121 - lightweight CDEBUG infrastructure, debug daemon (3668)
6122 - change default OSC RPC parameters to be better on small clusters
6123 - turn off OST read cache for files smaller than 32MB
6124 - install man pages and include them in rpms (3100)
6125 - add new init script for (un)mounting lustre filesystems (2593)
6126 - run chkconfig in %post for init scripts (3701)
6127 - drop scimac NAL (unmaintained)
6129 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6132 - clean kiobufs before and after use (3485)
6133 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6134 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6135 - change init script to fail more clearly if not run as root (1528)
6136 - allow clients to reconnect during replay (1742)
6137 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6138 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6139 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6140 - fix cond_resched (3554)
6141 - only evict unfinished clients after recovery (3515)
6142 - allow bulk resend, prevent data loss (3570)
6143 - dynamic ptlrpc request buffer allocation (2102)
6144 - don't allow unlinking open directory if it isn't empty (2904)
6145 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6146 - remove extraneous obd dereference causing LASSERT failure (3334)
6147 - don't use get_cycles() when creating temp. files on the mds (3156)
6148 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6149 - handle EEXIST for set-stripe, set proper directory name (3336)
6151 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6152 - fix ksocknal_fmb_callback() error messages (2918)
6154 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6157 - don't copy lvb into (possibly NULL) reply on error (2983)
6158 - don't deref dentry after dput, don't free lvb on error (2922)
6159 - use the kms to determine writeback rpc length (2947)
6160 - increment oti_logcookies when osc is inactive (2948)
6161 - update client's i_blocks count via lvb messages (2543)
6162 - handle intent open/close of special files properly (1557)
6163 - mount MDS with errors=remount-ro, like obdfilter (2009)
6164 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6165 - don't use cancelling-locks' kms values (2947)
6166 - use highest lock extent for kms, not last one (2925)
6167 - don't dereference ERR_PTR() dentry in error handling path (3107)
6168 - fix thread race in portals_debug_dumplog() (3122)
6169 - create lprocfs device entries at setup instead of at attach (1519)
6170 - common AST error handler, don't evict client on completion race (3145)
6171 - zero nameidata in detach_mnt in 2.6 (3118)
6172 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6173 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6174 - initialize RPC timeout timer earlier for 2.6 (3219)
6175 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6176 - print nal/nid for unknown nid (3258)
6177 - additional checks for oscc recovery before doing precreate (3284)
6178 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6179 - don't crash in mdc_close for bad permissions on open (3285)
6180 - zero i_rdev for non-device files (3147)
6181 - clear page->private before handing to FS, better assertion (3119)
6182 - tune the read pipeline (3236)
6183 - fix incorrect decref of invalidated dentry (2350)
6184 - provide read-ahead stats and refine rpc in flight stats (3328)
6185 - don't hold journal transaction open across create RPC (3313)
6186 - update atime on MDS at close time (3265)
6187 - close LDAP connection when recovering to avoid server load (3315)
6188 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6189 - don't leak open file on MDS after open resend (3325)
6190 - serialize filter_precreate and filter_destroy_precreated (3329)
6191 - loop device shouldn't call sync_dev() for nul device (3092)
6192 - clear page cache after eviction (2766)
6193 - resynchronize MDS->OST in background (2824)
6194 - refuse to mount the same filesystem twice on same mountpoint (3394)
6195 - allow llmount to create routes for mounting behind routers (3320)
6196 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6197 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6198 - fix rare problem with rename on htree directories (3417)
6200 - allow default OST striping configuration per directory (1414)
6201 - fix compilation for qswnal for 2.6 kernels (3125)
6202 - increase maximum number of MDS request buffers for large systems
6203 - change liblustreapi to be useful for external progs like lfsck (3098)
6204 - increase local configuration timeout for slow disks (3353)
6205 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6207 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6210 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6211 - don't overwrite extent policy data in reply if lock was blocked (2901)
6212 - drop filter export grants atomically with removal from device (2663)
6213 - del obd_self_export from work_list in class_disconnect_exports (2908)
6214 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6215 - swab reply message in mdc_close, other PPC fixes (2464)
6216 - fix destroying of named logs (2325)
6217 - overwrite old logs when running lconf --write_conf (2264)
6218 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6219 - fix race in target_handle_connect (2898)
6220 - mds_reint_create() should take same inode create lock (2926)
6221 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6222 - don't close files for self_export to avoid uninitialized obd (2936)
6223 - allow MDS with the same name as client node (2939)
6224 - hold dentry reference for closed log files for unlink (2325)
6225 - reserve space for all logs during transactions (2059)
6226 - don't evict page beyond end of stripe extent (2925)
6227 - don't oops on a deleted current working directory (2399)
6228 - handle hard links to targets without a parent properly (2517)
6229 - don't dereference NULL lock when racing during eviction (2867)
6230 - don't grow lock extents when lots of conflicting locks (2919)
6232 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6235 - account for cache space usage on clients to avoid data loss (974)
6236 - lfsck support in lustre kernel code (2349)
6237 - reduce journal credits needed for BRW writes (2370)
6238 - orphan handling to avoid losing space on client/server crashes
6239 - ptlrpcd can be blocked, stopping ALL progress (2477)
6240 - use lock value blocks to assist in proper KMS, faster stat (1021)
6241 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6242 - recovery for initial connections (2355)
6243 - fixes for mds_cleanup_orphans (1934)
6244 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6245 - block all file creations until orphan recovery completes (1901)
6246 - client remove rq_connection from request struct (2423)
6247 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6248 - recovery timer race (2670)
6249 - mdc_close recovey bug (2532)
6250 - ptlrpc cleanup bug (2710)
6251 - mds timeout on local locks (2588)
6252 - namespace lock held during RPCs (2431)
6253 - handle interrupted sync write properly (2503)
6254 - don't try to handle a message that hasn't been replied to (2699)
6255 - client assert failure during cleanup after abort recovery (2701)
6256 - leak mdc device after failed mount (2712)
6257 - ptlrpc_check_set allows timedout requests to complete (2714)
6258 - wait for inflight reqs when ptlrpcd finishes (2710)
6259 - make sure unregistered services are removed from the srv_list
6260 - reset bulk XID's when resending them (caught by 1138 test)
6261 - unregister_bulk after timeout
6262 - fix lconf error (2694)
6263 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6264 - readahead locks pages, leaves pending causing memory pressure (2673)
6265 - increase OST request buffers to 4096 on large machines (2729)
6266 - fix up permission of existing directories in simple_mkdir (2661)
6267 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6268 - don't assign transno to errored transactions (2742)
6269 - don't delete objects on OST if given a bogus objid from MDS (2751)
6270 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6271 - if rq_no_resend, then timeout request after recovery (2432)
6272 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6273 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6274 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6275 - MDS recovery shouldn't delete objects that it has given out (2730)
6276 - if enqueue arrives after completion, don't clobber LVB (2819)
6277 - don't unlock pages twice when trigger_group_io returns error (2814)
6278 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6279 - don't write pages to disk if there was an error (1450)
6280 - don't ping imports that have recovery disabled (2676)
6281 - take buffered bytes into account when balancing socknal conn (2817)
6282 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6283 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6284 - remove little-endian swabbing of llog records (1987)
6285 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6286 - retry reposting request buffers if they fail (1191)
6287 - grow extent at grant time to avoid granting a revoked lock (2809)
6288 - lock revoke doesn't evict page if covered by a second lock (2765)
6289 - disable VM readahead to avoid reading outside lock extents (2805)
6291 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6292 - updated kernel patches for hp-2.4.20 kernel (2681)
6294 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6297 - fix truncated write corruption (2366)
6298 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6300 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6301 - ensure len(uuid) < 37 in lmc (1171)
6302 - fix ia64 OOPS in llog_test (2255)
6303 - zero end of page at obdfilter for partial page writes (2648)
6304 - don't leave stale dentries around after renames (bug 2428)
6305 - fix timeouts when evicting a client with a single lock held (2642)
6306 - set deadline for the initial HELLO message to drain (2634)
6307 - print out dotted-quad IP addresses in the socknal (2302)
6309 - additional debugging for MDS client eviction problem (2443)
6310 - fix mkfsoptions support for osts (2603, 2604)
6312 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6315 - add series for the vanilla 2.6.0 kernel
6316 - add series for the vanilla 2.4.24 kernel
6317 - add series for a cray x86/64 UL kernel drop
6318 - fix xattr patches for the vanilla 2.4.19 series
6320 - generate true UUIDs in lmc (1171)
6321 - have portals stack dumping break in UML (2466)
6322 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6323 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6324 - wake callback waiting threads on client eviction (2460)
6325 - Add --ptldebug and --subsystem to lmc (1719)
6326 - update assertion to allow safe interrupt allocation
6327 - set rq_no_resend for cancel requests (2432)
6328 - recalculate ptlrpcd timeout after resend (2494)
6329 - call vfs_rmdir when removing pending directories (2368)
6330 - fix renaming a file to itself (2429)
6331 - lmc creates a default one-stripe lov (2454)
6332 - expand procfs space to handle large clusters (2326)
6333 - increase UML stack to avoid overflow
6334 - update lconf's list of debug and subsystem masks
6335 - fix lfs find --obd (2510)
6336 - /proc tunable for disabling filter read caching (2591)
6337 - stop rpm packages from altering slapd.conf (2301)
6338 - disable nagle in the socknal under 0conf (2578)
6339 - choose mds inode size based on stripe count (2572)
6340 - fix kernel-source rpm problems (2516)
6342 - add --disable-doc to avoid pdf generation (2421)
6343 - update documentation, tests, type-os, comments
6344 - avoid format warnings on ia64
6345 - remove the TOE NAL
6346 - tiny code cleanups by removing unused fields
6348 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6351 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6352 - remove the most blatant lies from BUILDING file (2371)
6353 - change default debug level to reasonable production setting
6354 - reduce client side cache size to reduce cache flush time
6355 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6356 - make TCP zerocopy and pinger support enabled by default (2476)
6357 - sync writes completed after process exits caused crashes (2319)
6358 - maintain correct mount count on the MDS (2356)
6359 - backout 1557, because 2316 wasn't really fixed
6360 - better file I/O statistics gathering in /proc
6361 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6362 - another kernel patch to fix zero-copy TCP function export
6363 - don't take duplicate lock when processing re-sent getattr (2420)
6364 - lctl uses obd_self_export instead of creating new conn (2353)
6365 - MDS/OST recovery case which requires object creation asserted (2425)
6366 - move lfs from /usr/sbin to /usr/bin in packages
6367 - fix race between mds_client_add and mds_client_free (2417)
6368 - use kmalloc instead of slabs in portals (2430)
6369 - don't create duplicate records when a failover MDS is present (2442)
6370 - remove unnecessary mount age check (2332)
6371 - don't remove directory inodes from locks prematurely (2451)
6372 - don't break if MDS service name is the same as hostname (2103)
6373 - fix races in client write RPC generation when cache full (2482)
6375 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6378 - remove now-unused request->rq_obd (278)
6379 - if an allocation fails, print out how much memory we've used (1933)
6380 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6381 - add the "configurable stack size" patch to most series files (1256)
6382 - ability to write large log records, for 100+ OST configs (2306)
6383 - fix NULL deref when filter_prep fails (2314)
6384 - fix operator precedence error in filter_sync
6385 - dynamic allocation of socknal TX descriptors (2315)
6386 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6387 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6388 - ensure that CWARN messages in Portals always get to the syslog
6389 - __init/__exit are not for prototype decls (ldlm_init/exit)
6390 - x86-64 compile warning fixes
6391 - fix gateway LMC keyword conflict (2318)
6392 - fix MDS lock inversions in getattr/reint paths (1844)
6393 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6394 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6395 - DLM race condition prevented some lock evictions (2328)
6396 - ENOMEM detection and retry on socknal sends (2230)
6397 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6398 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6399 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6400 - better error messages when a client is rejected during recovery (1505)
6401 - avoid cancelling locks which were never granted, after failure (2330)
6402 - fix i_sem/journal inversion in mds_client_add (2333)
6403 - fix truncate/getattr lock cycle deadlock (2334)
6404 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6405 - fix two rare exit paths which could leak an l_lock() ref (2321)
6406 - fencepost error in MDS/OST orphan recovery (2226)
6407 - make log record alignment 8 bytes (1988)
6408 - lstripe now fails when requested offset > ost_count (2237)
6409 - ensure that all kernel series have a complete list.h (1607)
6410 - fix crashes in special-file operations (2316)
6411 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6413 - allow configurable automake binary, for testing new versions
6414 - small update to the lfs documentation
6416 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6418 * fix negative export reference count in fsfilt_sync (2312)
6420 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6421 * release candidate 0.9.1
6423 - orphans are moved into the PENDING directory for possible recovery
6424 - replayed opens now open by fid for orphan/rename safety (1042)
6425 - last close of an orphan inode generates a transno (683)
6426 - chdir() and mount() now pin the directory entry (1020)
6427 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6428 - discard very old requests without processing them (1502)
6429 - remove some common, well-understood CERRORs (1505)
6430 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6431 - clear "grant" flags in OST replies until OST grant code lands (1644)
6432 - fix read performance by not clobbering i_blksize on client (1598)
6433 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6434 - make LRU size a /proc tunable, clears locks when reduced (707)
6435 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6436 - server locks take references on exports now (1558)
6437 - build fixes for 2.4.20-rh trees (1663)
6438 - return an error from lov_create if all OSCs are inactive (1751)
6439 - fix import levels when a reconnect happens without a timeout (1597)
6440 - exit early from mds_open if we get a lookup error (1749)
6441 - partial page read at EOF wouldn't wait for disk before sending (1642)
6442 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6443 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6444 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6445 - write/truncate lock inversion (1639)
6446 - Don't auto-load obdclass, portals modules during cleanup (1495)
6447 - fix timestamps from jumping to "now" (1763)
6448 - extra journal assertions (1648)
6449 - add an extra multiunlink test (1771)
6450 - fix read_record/write_record API (1776)
6451 - fix leak of offset_extent, possible incorrect i_size later (1772)
6452 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6453 - Debugging for the kqswnal_get_idle_tx problems (1820)
6454 - Allow recovery to be attempted multiple times (1536)
6455 - Write out MDS last_rcvd file after it is first created (1600)
6456 - Fix tx_descriptor leak in failed transmit situations (1827)
6457 - ext3 journaling fixes for assertion failure after IO error (1871)
6458 - class_export_put() on freed export after completion AST error (1896)
6459 - Fix revalidate looping in VFS (1322)
6460 - Don't access a freed export during MDS_REINT timeout (1521)
6461 - Add open-unlink recovery support on the MDS (1673,1764)
6462 - Return an error if no MDS data was read from last_rcvd (1946)
6463 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6464 - Don't setup a disk device that doesn't match exported UUID (317)
6465 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6466 - avoid committing NULL handle in force close
6467 - local.sh is now a one-stripe LOV configuration
6468 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6469 - fix ext3 htree duplicate directory entry corruption (1516)
6470 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6471 - update to new LOV EA format (2097)
6472 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6474 2003-06-15 Phil Schwan <phil@clusterfs.com>
6477 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6478 - per-import/export recovery handling (958, 931, 959)
6479 - multiple last-rcvd slots, for serving multiple FSes (949)
6480 - connections are again shared between multiple imp/exports (963, 964)
6481 - "umount -f" would hang if any requests needed to be sent (393, 978)
6482 - avoid pinning large req buffer by copying for queued messages (989)
6483 - add "uuid" to "lctl device" command to help upcalls (991)
6484 - "open" RPCs with transnos would confuse recovery counters (1037)
6485 - do proper endian conversion of all wire messages (288, 340, 891)
6486 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6487 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6488 - fix ptlrpc_connection leak in target_handle_connect (1174)
6489 - fix import refcounting bug in OST and MDS cleanup (1134)
6490 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6491 - fix very unlikely obd_types race condition (501)
6492 - remove osc_open hack for echo_client (1187)
6493 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6494 - a failure in read_inode2 leads to deadlock (1139)
6495 - cancel ack-locks as soon as transaction is committed (1072)
6496 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6497 - make sure to commitrw after any preprw to avoid deadlock (1162)
6498 - failing to execute a file in a lustre FS would lock inode (1203)
6499 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6500 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6501 - obd_uuid2tgt would walk off the end of the list (1255)
6502 - on IA64 the osc would give portals incorrect bulk size (1258)
6503 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6504 - fix lock inversion caused by new llite matching code (1282)
6505 - limit the number of dirty pages on a client to 10MB (1286)
6506 - timed out locks were not being corrected cancelled (1289)
6507 - fix O_DIRECT above 4GB on IA-32 (1292)
6508 * major user-visible changes
6509 - fail out/fail over policy now controlled by the upcall (993)
6511 - add OBD_PING to check server availability and failure (954)
6512 - lustre messages are now sent in sending host order (288, 340, 891)
6513 - add eadatalen to MDS getattr reply (340)
6514 - OST read replies may contain second buffer, with per-page status (593)
6516 2003-03-11 Phil Schwan <phil@clusterfs.com>
6519 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6520 - failed open()s could cause deadlock; fixed (867, 869)
6521 - stop cancelling OST locks when files are closed (481)
6522 - overlapping XID spaces caused network corruption (851, 853)
6523 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6524 - setattr_raw added, to do single-RPC, server-side setattrs
6525 - lmc/lconf syntax change for OST UUIDs
6526 - fix crashy race condition between ptlrpc_free_req and osc_close
6527 - don't use request in mdc_enqueue if we hit a timeout (889)
6528 - don't set the inode i_size for regular files from the MDS (896)
6529 - handle out of order completion AST (842)
6530 - don't LBUG if a lock request times out after receiving AST (913)
6531 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6532 - if a bad lock AST arrives, send an error instead of dropping entirely
6533 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6534 - fix leak in bulk IO when only partially completed (899, 900, 926)
6535 - fix O_DIRECT for ia64 (55)
6536 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6537 - C-z after timeout could hang a process forever; fixed (977)
6539 - client-side I/O cache (678, 924, 929, 941, 970)
6541 - READPAGE and SETATTRs which don't take server-side locks get
6544 2003-02-11 Phil Schwan <phil@clusterfs.com>
6547 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6548 - Fix fsfilt_extN_readpage() to read a full page of directory
6549 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6550 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6551 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6552 - Most intent operations were removed, in favour of a new RPC mode
6553 that does a single RPC to the server and bypasses most of the VFS
6554 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6555 - Aggressively cancel local locks on DLM servers
6556 - mds_reint_unlink sends EA to the client if it's the last nlink.
6557 client uses that EA to unlink OST objects.
6558 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6559 - recursive symlinks were fixed (439)
6560 - fixed NULL deref in DEBUG_REQ
6561 - filter_update_lastobjid no longer calls sync, which annoyed extN
6562 - fixed multi-client small-writes to a single file problem (445)
6563 - fixed mtime updates during file writes (607)
6564 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6565 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6566 - fixed Linux OST size reporting problem (444, 656)
6567 - OST now updates object mtime with writes or setattr (607, 619)
6568 - client verifies file size before zeroing page past EOF (445)
6569 - OST now writes last allocated objid to disk with allocation (108)
6570 - LOV on echo now works (409)
6572 - mds_reint_unlink sends a new buffer, with the EA included. this
6573 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6574 set if a regular file was being unlinked, and it was the last link
6575 - use PtlGet from the target for bulk writes (315)
6576 - OST now updates object mtime with writes or setattr (607, 619)
6577 - LDLM now has a grant-time callback to revalidate locked items, if
6579 - Many MDS operations were reorganized to combat race conditions
6581 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6582 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6584 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6587 - Fully reactivate OST imports after reconnection (512, others)
6588 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6589 - More graceful error handling for truncating on dead OST (515)
6590 - Don't error out unless we're actually accessing dead stripes (474)
6591 - Fix garbage sizes when stripes are missing (410)
6592 - LRU counters were broken, causing constant lock purge (433, 432)
6593 - garbage on read from stripes with failed OSTs (441)
6594 - mark OSCs as active before reconnecting during recovery (438)
6595 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6596 - lfind did not preserve OST order in output (443)
6597 - symlinks cause hung clients, incorrect data (439)
6598 - stop dereferencing request after dropping refcount (457)
6599 - don't LASSERT(spin_is_locked) on non-SMP (455)
6600 - fixes for many rename() bugs
6601 - fstat didn't correctly synchronize attributes (399)
6602 - server must handle lock cancellation during blocking AST prep (487)
6603 - bulk descriptors were free()d too soon (511)
6604 - fix paths in lconf, which would load incorrect modules (451, 507)
6605 - fix confusing lconf 'host not found' error message (386)
6606 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6607 - fix race condition in mdc_blocking_ast() for inode access (526)
6608 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6609 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6610 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6611 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6612 - fix lstripe in several ways (kernel side) (527)
6613 - fix request leak in ldlm_cli_enqueue (262)
6614 - incorrect OSC was marked inactive after OST failure
6615 - call mds_fs_cleanup before unmounting filesystem (524)
6616 - fix races between taking ns_lock and ldlm_lock_change_resource
6617 - fix races updating LOV export open file list
6618 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6619 - fix recovery NULL deref in ldlm_cli_cancel_unused
6620 - fix some DLM races by using new hash table for lock handles (419)
6621 - permit the client to specify desired inodes, at replay
6622 - duplicate requests when we queue them for replay reintegration
6623 - fix last_rcvd offset calculation
6624 - sync after each recovered transaction, so we always make progress
6625 - never, not always, ERESTART requests without transnos
6626 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6628 - skip replay if the MDS says that the client is already connected
6629 - don't check for a recovery-enabled export to match lctl's UUID
6630 - don't INC_USE_COUNT for phantom exports
6631 - don't crash when cleaning up phantom exports (567)
6632 - don't double-finish or set replay data for errored mdc_open requests
6633 - abort requests when they time out, so we don't get old replies
6634 - send/receive replies for AST messages again
6635 - if the client says that it doesn't have the lock, cancel it on the
6637 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6638 mark it as destroyed, it will all work out when decref is called
6639 - fix module use counts (22, 581)
6641 - ASTs now expect a reply (server cancels lock on error reply)
6643 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6646 - fix many simultaneous client startup (392)
6647 - fix dentry->d_it clobbering
6648 - credentials weren't being shipped for readdir/getattr operations
6649 - remove invalid assertions triggered during some concurrent MD
6651 - proper Lustre versions added (336, 389)
6652 - fix memory leak for create error case (398)
6653 - fix LOV locking bug that would get cli/srv out of sync
6654 - fix echo client over LOV (409)
6655 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6656 - fix double-O_EXCL intent crash (424)
6657 - avoid sending multiple lock CANCELs (352)
6659 - MDS can do multi-client recovery (modulo bugs in new code)
6661 - many updates, edits, cleanups
6663 2002-11-18 Phil Schwan <phil@clusterfs.com>
6666 - fix null d_it dereference (346)
6667 - fix full OST/dbench hang (333)
6668 - fix permission problem with file removal (286)
6669 - fix removal of OSCs from LOV when they fail
6670 - fix NULL deref during bulk timeout (214)
6671 - fix problems related to multiple filesystems on one MDS (241)
6672 - fixed serious subtle metadata locking bugs
6673 - free locks on clients when inodes are removed due to memory
6675 - fix inode pointer in lock data (285)
6676 - partial support for multiple MDS on a single host (241)
6677 - data locks weren't cancelled at clear_inode time (290, 311)
6678 - intent locks could lead to unbounded lock growth (205)
6679 - added a maximum lock count, an LRU list, and a flusher
6680 - fix multiple rename (365)
6681 - properly abstracted the echo client
6682 - OSC locked 1 byte too many; fixed
6683 - rewrote brw callback code:
6684 - fixed recovery bugs related to LOVs (306)
6685 - fixed too-many-pages-in-one-write crash (191)
6686 - fixed (again) crash in sync_io_timeout (214)
6687 - probably fixed callback-related race (385)
6689 - Add capability to MDS protocol
6690 - LDLM cancellations and callbacks on different portals
6692 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6695 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6696 - defer open object destruction to close time (601981, 138)
6697 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6698 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6699 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6700 - don't call obd_setattr in ll_file_release for destroyed objects
6701 * protocol change to lustre_msg: move |version| and add |flags|
6702 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6703 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6704 * added replay of create, unlink, link and rename operations during
6705 MDS failover; recovery should be much more robust now
6706 * remove failed OSCs from LOVs (only lov_create uses this so far)
6707 * the lustre-HOWTO was brought (more) up to date (582544)
6709 2002-10-23 Phil Schwan <phil@clusterfs.com>
6712 - in-use dentries weren't being reused properly (617851)
6713 - prevent multiple LDLM setup (599178)
6714 - fix LOV size calculations for truncate (617853)
6715 - fix client handling of MDS intent errors (POSIX)
6716 - fix permission bug in lovstripe.c test (624321)
6717 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6718 - truncate past end of file could corrupt data
6719 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6720 - a race in recovery could return ETIMEDOUT to apps (623947)
6721 - building outside the source directory was fixed
6722 * the lustre-HOWTO was brought (more) up to date (582544)
6723 * major progress was made on recovery functionality
6725 2002-10-10 Phil Schwan <phil@clusterfs.com>
6728 - recovery deadlock fix
6729 - rm -rf causes LBUG fix (617817)
6730 - file open by multiple tasks fix (618962)
6731 - directory permissions bugs (602707 and 620007)
6732 - journal_stop fixed with locking (611313)
6733 - O_APPEND failures resolved (618273, perhaps 614459)
6734 - lconf PATH fix (619770)
6735 - IA64 build fix (621450)
6736 - RPC buffer sizes scale with amount of memory
6738 2002-10-01 Phil Schwan <phil@clusterfs.com>
6741 - locks would be cancelled without throwing away data pages,
6742 resulting in inconsistent data (605627)
6743 - inode attributes were not always being refreshed (605627, 612449)
6744 - lconf now continues to cleanup after lctl reports an error
6745 - MDS now enforces user permissions (602707)
6746 - lprocfs cleanup fixed, but not yet enabled (614157)
6747 - fixed infinite server hang, should a client not respond to an AST
6748 - avoid going into recovery if user calls readlink() with a buffer
6749 that's too small (613941)
6750 - AST RPCs no longer require replies (614867) -- this may be changed
6751 - don't crash server if client sends an IOV that's too big (611336)
6752 - fixed lock conversion deadlock (611892)
6753 - fixed the following of symlinks (614622)
6754 * recovery: the server can remove locks from a client that dies, other
6755 clients can make progress
6756 * more extN patch fixes
6757 * compile-time configurable ptlrpc buffer allocations
6759 - collaborative read cache document
6760 - Lustre Lite Performance CDR document-in-progress
6762 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6765 - fix typo in patch-2.4.18
6767 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6770 - clear ptlrpc request each time in handle_incoming_request()
6771 - unlink of files now destroys the object on the OST
6773 2002-09-19 Peter Braam <braam@clusterfs.com>
6775 * add hard link support
6776 * change obdfile creation method
6777 * kernel patch changed
6779 2002-09-19 Peter Braam <braam@clusterfs.com>
6782 - stack overflow bug in extN fixed
6784 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6786 * documentation updates
6787 - add man pages for config tools
6788 - update tests/README to describe testing with new config tools
6789 - finish metadata API descriptions
6790 * bug fixes and cleanups
6791 - statfs workaround for 16TB limit
6792 - LOV stripe allocation improved, can stripe on subset of OSTs
6793 - LOV file size/IO offset was wrong for files > 4GB in size
6794 - object EA data was being dropped, caused files to be unreadable
6795 - memory overflow with non-LOV OST caused memory corruption
6796 - fixed regression tests to work with new config tools, obdfilter
6797 - fixed bug when directory size became larger than 1 block
6798 - fixed bug (for single client case) when PWD was deleted
6799 - invalidate local directory pages when doing intent-based ops
6800 - avoid LDLM oops when lock callback contained bad data
6802 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6804 * documentation updates
6805 * bug fixes and cleanups
6806 - configuration tools
6809 - 64-bit compile warnings
6810 - 64-bit internal statfs data
6812 * test_brw on persistent OST devices
6814 * lprocfs (disabled)
6816 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6818 * documentation updates
6819 * bug fixes and cleanups
6820 * configuration tools
6822 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6825 * many small fixes to 0_5_4
6826 * io/network handling
6827 * thinkos in MDS operations
6829 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6832 * crucial basic fixes to 0.5.3
6833 * IOR, Iozone work over Elan
6836 2002-08-07 Phil Schwan <phil@clusterfs.com>
6837 * version 0_5_3, our first alpha
6838 * we use the new Portals iovs
6839 * documentation updates
6840 * bug fixes and cleanups
6841 * small changes in the DLM wire protocol
6843 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6844 * version 0_5_1 with some initial stability,
6845 * locking on MD and file I/O.
6846 * documentation updates
6847 * several bug fixes since 0.5.0
6848 * small changes in wire protocol
6850 2002-07-18 Phil Schwan <phil@clusterfs.com>
6852 * delivered as Lustre Light Alpha
6853 * fixed a crash after handling invalid MDS requests
6854 * fixed directory pages for architectures with non-4k pages sizes
6856 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6857 * release version v0_4_4
6858 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6860 * Fixes a number of interruption problems with OST operations.
6861 * Update documentation for portals header changes
6862 * Move all wire protocol structs/defines to lustre_idl.h
6863 * Fixes symlink length bug.
6864 * Add tcpdump to repository.
6866 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6867 * release version v0_4_3
6868 * Fixes statfs for inodes on extN.
6869 * Fixes bug in runtests which would delete /etc/hosts.
6870 * Use 64-bit object IDs wherever possible (not into VFS though)
6871 Remove ost_get_info, which is unused by lustre, and out of date.
6873 2002-07-03 Peter Braam <braam@clusterfs.com>
6874 * release version v0_4_2 Fixes a lookup error (type not passed)
6875 * move forward to head of Portals
6876 * move forward to latest Lustre kernel
6878 2002-06-25 Peter Braam <braam@clusterfs.com>
6879 * release version v0_4_1. Hopefully stable on single node use.