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 4.18.0-348.23.1.el8 (RHEL8.5)
7 * Other server kernels known to build and work at some point (others may also work):
8 3.10.0-862.14.4.el7 (RHEL7.5)
9 3.10.0-957.27.2.el7 (RHEL7.6)
10 3.10.0-1062.18.1.el7 (RHEL7.7)
11 3.10.0-1127.19.1.el7 (RHEL7.8)
12 3.10.0-1160.66.1.el7 (RHEL7.9)
13 4.18.0-80.11.2.el8 (RHEL8.0)
14 4.18.0-147.8.1.el8 (RHEL8.1)
15 4.18.0-193.28.1.el8 (RHEL8.2)
16 4.18.0-240.22.1.el8 (RHEL8.3)
17 4.18.0-305.25.1.el8 (RHEL8.4)
18 4.4.120-92.70 (SLES12 SP2)
19 4.4.180-94.100 (SLES12 SP3)
20 4.4.0-131 (Ubuntu 16.04)
21 4.15.0-32 (Ubuntu 18.04)
22 5.4.0-48 (Ubuntu 20.04)
23 vanilla linux 5.4.0 (ZFS + ldiskfs)
24 vanilla linux 5.4.21 (ZFS + ldiskfs)
25 vanilla linux 5.4.136 (ZFS + ldiskfs)
26 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
27 * Client primary kernels built and tested during release cycle:
28 4.18.0-372.9.1.el8 (RHEL8.6)
29 5.4.0-37 (Ubuntu 20.04)
30 5.3.18-59.27 (SLES15 SP3)
31 * Other clients known to build on these kernels at some point (others may also work):
32 3.10.0-862.14.4.el7 (RHEL7.5)
33 3.10.0-957.27.2.el7 (RHEL7.6)
34 4.14.0-49.13.1.el7a (RHEL7.5)
35 4.14.0-115.2.2.el7a (RHEL7.6)
36 3.10.0-1062.18.1.el7 (RHEL7.7)
37 3.10.0-1127.19.1.el7 (RHEL7.8)
38 3.10.0-1160.66.1.el7 (RHEL7.9)
39 4.18.0-80.11.2.el8 (RHEL8.0)
40 4.18.0-147.8.1.el8 (RHEL8.1)
41 4.18.0-193.28.1.el8 (RHEL8.2)
42 4.18.0-240.22.1.el8 (RHEL8.3)
43 4.18.0-305.25.1.el8 (RHEL8.4)
44 4.18.0-348.23.1.el8 (RHEL8.5)
45 4.4.120-92.70 (SLES12 SP2)
46 4.4.180-94.100 (SLES12 SP3)
47 4.12.14-95.48 (SLES12 SP4)
48 4.12.14-122.116 (SLES12 SP5)
49 4.12.14-197.75 (SLES15 SP1)
50 5.3.18-24.96 (SLES15 SP2)
51 4.4.0-131 (Ubuntu 16.04)
52 4.15.0-48 (Ubuntu 18.04)
53 5.8.0-53 (Ubuntu 20.04.2 HWE)
54 5.11.0-31 (Ubuntu 20.04.3 HWE)
55 5.11.0 (vanilla kernel.org)
56 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
57 * Recommended ZFS version: 2.1.2
58 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
59 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
60 stack overflow. For more information, please refer to bugzilla 17630.
61 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
62 the re-exporting nodes to be mounted with "32bitapi" mount option
64 --------------------------------------------------------------------------------
69 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
70 for currently supported client and server kernel versions.
71 * Server primary kernels built and tested during release cycle:
72 4.18.0-348.2.1.el8 (RHEL8.5)
73 * Other server kernels known to build and work at some point (others may also work):
74 3.10.0-862.14.4.el7 (RHEL7.5)
75 3.10.0-957.27.2.el7 (RHEL7.6)
76 3.10.0-1062.18.1.el7 (RHEL7.7)
77 3.10.0-1127.19.1.el7 (RHEL7.8)
78 3.10.0-1160.49.1.el7 (RHEL7.9)
79 4.18.0-80.11.2.el8 (RHEL8.0)
80 4.18.0-147.8.1.el8 (RHEL8.1)
81 4.18.0-193.28.1.el8 (RHEL8.2)
82 4.18.0-240.22.1.el8 (RHEL8.3)
83 4.18.0-305.25.1.el8 (RHEL8.4)
84 4.4.120-92.70 (SLES12 SP2)
85 4.4.180-94.100 (SLES12 SP3)
86 4.4.0-131 (Ubuntu 16.04)
87 4.15.0-32 (Ubuntu 18.04)
88 5.4.0-48 (Ubuntu 20.04)
89 vanilla linux 5.4.0 (ZFS + ldiskfs)
90 vanilla linux 5.4.21 (ZFS + ldiskfs)
91 vanilla linux 5.4.136 (ZFS + ldiskfs)
92 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
93 * Client primary kernels built and tested during release cycle:
94 4.18.0-348.2.1.el8 (RHEL8.5)
95 5.4.0-37 (Ubuntu 20.04)
96 5.3.18-59.27 (SLES15 SP3)
97 * Other clients known to build on these kernels at some point (others may also work):
98 3.10.0-862.14.4.el7 (RHEL7.5)
99 3.10.0-957.27.2.el7 (RHEL7.6)
100 4.14.0-49.13.1.el7a (RHEL7.5)
101 4.14.0-115.2.2.el7a (RHEL7.6)
102 3.10.0-1062.18.1.el7 (RHEL7.7)
103 3.10.0-1127.19.1.el7 (RHEL7.8)
104 3.10.0-1160.49.1.el7 (RHEL7.9)
105 4.18.0-80.11.2.el8 (RHEL8.0)
106 4.18.0-147.8.1.el8 (RHEL8.1)
107 4.18.0-193.28.1.el8 (RHEL8.2)
108 4.18.0-240.22.1.el8 (RHEL8.3)
109 4.18.0-305.25.1.el8 (RHEL8.4)
110 4.4.120-92.70 (SLES12 SP2)
111 4.4.180-94.100 (SLES12 SP3)
112 4.12.14-95.48 (SLES12 SP4)
113 4.12.14-122.91 (SLES12 SP5)
114 4.12.14-197.75 (SLES15 SP1)
115 5.3.18-24.96 (SLES15 SP2)
116 4.4.0-131 (Ubuntu 16.04)
117 4.15.0-48 (Ubuntu 18.04)
118 5.8.0-53 (Ubuntu 20.04.2 HWE)
119 5.11.0-31 (Ubuntu 20.04.3 HWE)
120 5.11.0 (vanilla kernel.org)
121 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
122 * Recommended ZFS version: 2.0.7
123 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
124 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
125 stack overflow. For more information, please refer to bugzilla 17630.
126 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
127 the re-exporting nodes to be mounted with "32bitapi" mount option
129 --------------------------------------------------------------------------------
133 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
134 for currently supported client and server kernel versions.
135 * Server primary kernels built and tested during release cycle:
136 3.10.0-1062.18.1.el7 (RHEL7.7)
137 3.10.0-1127.19.1.el7 (RHEL7.8)
138 3.10.0-1160.6.1.el7 (RHEL7.9)
139 * Other server kernels known to build and work at some point (others may also work):
140 3.10.0-862.14.4.el7 (RHEL7.5)
141 3.10.0-957.27.2.el7 (RHEL7.6)
142 4.18.0-80.11.2.el8 (RHEL8.0)
143 4.18.0-147.8.1.el8 (RHEL8.1)
144 4.18.0-193.28.1.el8 (RHEL8.2)
145 4.18.0-240.1.1.el8 (RHEL8.3)
146 4.4.120-92.70 (SLES12 SP2)
147 4.4.180-94.100 (SLES12 SP3)
148 4.4.0-131 (Ubuntu 16.04)
149 4.15.0-32 (Ubuntu 18.04)
150 5.4.0-48 (Ubuntu 20.04)
151 vanilla linux 5.4.0 (ZFS + ldiskfs)
152 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
153 * Client primary kernels built and tested during release cycle:
154 3.10.0-1062.18.1.el7 (RHEL7.7)
155 3.10.0-1127.19.1.el7 (RHEL7.8)
156 3.10.0-1160.6.1.el7 (RHEL7.9)
157 4.12.14-95.48 (SLES12 SP4)
158 4.12.14-122.20 (SLES12 SP5)
159 4.12.14-197.75 (SLES15 SP1)
160 5.3.18-24.24 (SLES15 SP2)
161 4.15.0-48 (Ubuntu 18.04)
162 * Other clients known to build on these kernels at some point (others may also work):
163 3.10.0-862.14.4.el7 (RHEL7.5)
164 3.10.0-957.27.2.el7 (RHEL7.6)
165 4.14.0-49.13.1.el7a (RHEL7.5)
166 4.14.0-115.2.2.el7a (RHEL7.6)
167 4.18.0-80.11.2.el8 (RHEL8.0)
168 4.18.0-147.8.1.el8 (RHEL8.1)
169 4.18.0-193.28.1.el8 (RHEL8.2)
170 4.18.0-240.1.1.el8 (RHEL8.3)
171 4.4.120-92.70 (SLES12 SP2)
172 4.4.180-94.100 (SLES12 SP3)
173 4.4.0-131 (Ubuntu 16.04)
174 5.4.0-37 (Ubuntu 20.04)
175 5.4.0 (vanilla kernel.org)
176 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
177 * Recommended ZFS version: 2.0.0
178 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
179 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
180 stack overflow. For more information, please refer to bugzilla 17630.
181 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
182 the re-exporting nodes to be mounted with "32bitapi" mount option
184 --------------------------------------------------------------------------------
188 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
189 for currently supported client and server kernel versions.
190 * Server primary kernels built and tested during release cycle:
191 3.10.0-1062.1.1.el7 (RHEL7.7)
192 * Other server kernels known to build and work at some point (others may also work):
193 3.10.0-862.14.4.el7 (RHEL7.5)
194 3.10.0-957.27.2.el7 (RHEL7.6)
195 4.18.0-80.7.1.el8 (RHEL8.0)
196 4.4.120-92.70 (SLES12 SP2)
197 4.4.180-94.100 (SLES12 SP3)
198 4.4.0-131 (Ubuntu 16.04)
199 4.15.0-32 (Ubuntu 18.04)
200 vanilla linux 4.6.7 (ZFS only)
201 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
202 * Client primary unpatched kernels built and tested during release cycle:
203 3.10.0-1062.1.1.el7 (RHEL7.7)
204 4.12.14-95.29 (SLES12 SP4)
205 4.15.0-48 (Ubuntu 18.04)
206 * Other clients known to build on unpatched kernels at some point (others may also work):
207 3.10.0-862.14.4.el7 (RHEL7.5)
208 3.10.0-957.27.2.el7 (RHEL7.6)
209 4.14.0-49.13.1.el7a (RHEL7.5)
210 4.14.0-115.2.2.el7a (RHEL7.6)
211 4.18.0-80.7.1.el8 (RHEL8.0)
212 4.4.120-92.70 (SLES12 SP2)
213 4.4.180-94.100 (SLES12 SP3)
214 4.4.0-131 (Ubuntu 16.04)
215 vanilla linux 4.15.0 (ZFS only)
216 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
217 * Recommended ZFS / SPL version: 0.7.13
218 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
219 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
220 stack overflow. For more information, please refer to bugzilla 17630.
221 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
222 the re-exporting nodes to be mounted with "32bitapi" mount option
223 * Note for Ubuntu 14.04.5:
224 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
225 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
227 --------------------------------------------------------------------------------
231 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
232 for currently supported client and server kernel versions.
233 * Server known to build on the following kernels (others may also work):
234 2.6.32-696.18.7.el6 (RHEL6.9)
235 3.10.0-862.14.4.el7 (RHEL7.5)
236 3.10.0-957.el7 (RHEL7.6)
237 4.4.120-92.70 (SLES12 SP2)
238 4.4.162-94.69 (SLES12 SP3)
239 4.4.0-131 (Ubuntu 16.04)
240 4.15.0-32 (Ubuntu 18.04)
241 vanilla linux 4.6.7 (ZFS only)
242 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
243 * Client known to build on unpatched kernels (others may also work):
244 2.6.32-696.18.7.el6 (RHEL6.9)
245 2.6.32-754.6.3.el6 (RHEL6.10)
246 3.10.0-862.14.4.el7 (RHEL7.5)
247 3.10.0-957.el7 (RHEL7.6)
248 4.14.0-49.13.1.el7a (RHEL7.5)
249 4.14.0-115.2.2.el7a (RHEL7.6)
250 4.4.120-92.70 (SLES12 SP2)
251 4.4.162-94.69 (SLES12 SP3)
252 4.4.0-131 (Ubuntu 16.04)
253 4.15.0-32 (Ubuntu 18.04)
254 vanilla linux 4.15.0 (ZFS only)
255 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
256 * Recommended ZFS / SPL version: 0.7.9
257 * Tested with ZFS / SPL version: 0.7.9
258 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
259 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
260 stack overflow. For more information, please refer to bugzilla 17630.
261 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
262 the re-exporting nodes to be mounted with "32bitapi" mount option
263 * Note for Ubuntu 14.04.5:
264 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
265 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
267 --------------------------------------------------------------------------------
269 2018-04-03 Intel Corporation
271 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
272 for currently supported client and server kernel versions.
273 * Server known to build on patched kernels:
274 2.6.32-431.29.2.el6 (RHEL6.5)
275 2.6.32-504.30.3.el6 (RHEL6.6)
276 2.6.32-573.26.1.el6 (RHEL6.7)
277 2.6.32-642.15.1.el6 (RHEL6.8)
278 2.6.32-696.18.7.el6 (RHEL6.9)
279 3.10.0-693.21.1.el7 (RHEL7.4)
280 3.0.101-0.47.71 (SLES11 SP3)
281 3.0.101-107 (SLES11 SP4)
282 3.12.74-60.64.40 (SLES12 SP1)
283 4.4.114-92.64 (SLES12 SP2)
284 4.4.114-94.11 (SLES12 SP3)
285 3.13.0-101 (Ubuntu 14.04, ZFS only)
286 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
287 4.4.0-116 (Ubuntu 16.04)
288 vanilla linux 4.6.7 (ZFS only)
289 * Client known to build on unpatched kernels:
290 2.6.32-431.29.2.el6 (RHEL6.5)
291 2.6.32-504.30.3.el6 (RHEL6.6)
292 2.6.32-573.26.1.el6 (RHEL6.7)
293 2.6.32-642.15.1.el6 (RHEL6.8)
294 2.6.32-696.18.7.el6 (RHEL6.9)
295 3.10.0-693.21.1.el7 (RHEL7.4)
296 3.0.101-0.47.71 (SLES11 SP3)
297 3.0.101-107 (SLES11 SP4)
298 3.12.74-60.64.40 (SLES12 SP1)
299 4.4.114-92.64 (SLES12 SP2)
300 4.4.114-94.11 (SLES12 SP3)
301 3.13.0-101 (Ubuntu 14.04)
302 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
303 4.4.0-116 (Ubuntu 16.04)
305 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
306 * Recommended ZFS / SPL version: 0.7.6
307 * Tested with ZFS / SPL version: 0.7.6
308 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
309 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
310 stack overflow. For more information, please refer to bugzilla 17630.
311 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
312 the re-exporting nodes to be mounted with "32bitapi" mount option
313 * Note for Ubuntu 14.04.5:
314 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
315 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
317 --------------------------------------------------------------------------------
319 2017-07-05 Intel Corporation
321 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
322 for currently supported client and server kernel versions.
323 * Server known to build on patched kernels:
324 2.6.32-431.29.2.el6 (RHEL6.5)
325 2.6.32-504.30.3.el6 (RHEL6.6)
326 2.6.32-573.26.1.el6 (RHEL6.7)
327 2.6.32-642.15.1.el6 (RHEL6.8)
328 2.6.32-696.3.1.el6 (RHEL6.9)
329 3.10.0-514.21.1.el7 (RHEL7.3)
330 3.0.101-0.47.71 (SLES11 SP3)
331 3.0.101-100 (SLES11 SP4)
332 3.12.74-60.64.40 (SLES12 SP1)
333 4.4.59-92.17 (SLES12 SP2)
334 vanilla linux 4.6.7 (ZFS only)
335 * Client known to build on unpatched kernels:
336 2.6.32-431.29.2.el6 (RHEL6.5)
337 2.6.32-504.30.3.el6 (RHEL6.6)
338 2.6.32-573.26.1.el6 (RHEL6.7)
339 2.6.32-642.15.1.el6 (RHEL6.8)
340 2.6.32-696.3.1.el6 (RHEL6.9)
341 3.10.0-514.21.1.el7 (RHEL7.3)
342 3.0.101-0.47.71 (SLES11 SP3)
343 3.0.101-100 (SLES11 SP4)
344 3.12.74-60.64.40 (SLES12 SP1)
345 4.4.59-92.17 (SLES12 SP2)
347 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
348 * Recommended ZFS / SPL version: 0.6.5.9
349 * Tested with ZFS / SPL version: 0.6.5.9
350 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
351 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
352 stack overflow. For more information, please refer to bugzilla 17630.
353 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
354 the re-exporting nodes to be mounted with "32bitapi" mount option
356 --------------------------------------------------------------------------------
358 12-07-2016 Intel Corporation
360 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
361 for currently supported client and server kernel versions.
362 * Server known to build on patched kernels:
363 2.6.32-431.29.2.el6 (RHEL6.5)
364 2.6.32-504.30.3.el6 (RHEL6.6)
365 2.6.32-573.26.1.el6 (RHEL6.7)
366 2.6.32-642.11.1.el6 (RHEL6.8)
367 3.10.0-514.el7 (RHEL7.3)
368 3.0.101-0.47.71 (SLES11 SP3)
369 3.0.101-80 (SLES11 SP4)
370 3.12.62-60.64.8 (SLES12 SP1)
371 vanilla linux 4.5.7 (ZFS only)
372 * Client known to build on unpatched kernels:
373 2.6.32-431.29.2.el6 (RHEL6.5)
374 2.6.32-504.30.3.el6 (RHEL6.6)
375 2.6.32-573.26.1.el6 (RHEL6.7)
376 2.6.32-642.11.1.el6 (RHEL6.8)
377 3.10.0-514.el7 (RHEL7.3)
378 3.0.101-0.47.71 (SLES11 SP3)
379 3.0.101-80 (SLES11 SP4)
380 3.12.62-60.64.8 (SLES12 SP1)
382 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
383 * Recommended ZFS / SPL version: 0.6.5.8
384 * Tested with ZFS / SPL version: 0.6.5.8
385 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
386 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
387 stack overflow. For more information, please refer to bugzilla 17630.
388 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
389 the re-exporting nodes to be mounted with "32bitapi" mount option
391 --------------------------------------------------------------------------------
393 02-29-2016 Intel Corporation
395 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
396 for currently supported client and server kernel versions.
397 * Server known to build on patched kernels:
398 2.6.32-431.29.2.el6 (RHEL6.5)
399 2.6.32-504.30.3.el6 (RHEL6.6)
400 2.6.32-573.12.1.el6 (RHEL6.7)
401 3.10.0-327.3.1.el7 (RHEL7.2)
402 3.0.101-0.47.71 (SLES11 SP3)
403 3.0.101-68 (SLES11 SP4)
404 vanilla linux 4.2.1 (ZFS only)
405 * Client known to build on unpatched kernels:
406 2.6.32-431.29.2.el6 (RHEL6.5)
407 2.6.32-504.30.3.el6 (RHEL6.6)
408 2.6.32-573.12.1.el6 (RHEL6.7)
409 3.10.0-327.3.1.el7 (RHEL7.2)
410 3.0.101-0.47.71 (SLES11 SP3)
411 3.0.101-68 (SLES11 SP4)
414 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
415 * Recommended ZFS / SPL version: 0.6.4.2
416 * Tested with ZFS / SPL version: 0.6.4.2
417 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
418 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
419 stack overflow. For more information, please refer to bugzilla 17630.
420 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
421 the re-exporting nodes to be mounted with "32bitapi" mount option
423 --------------------------------------------------------------------------------
425 03-10-2015 Intel Corporation
427 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
428 for currently supported client and server kernel versions.
429 * Server known to build on patched kernels:
430 2.6.32-431.29.2.el6 (RHEL6.5)
431 2.6.32-504.8.1.el6 (RHEL6.6)
432 3.0.101-0.46 (SLES11 SP3)
433 * Client known to build on unpatched kernels:
434 2.6.32-431.29.2.el6 (RHEL6.5)
435 2.6.32-504.8.1.el6 (RHEL6.6)
436 3.10.0-123.20.1.el7 (RHEL7)
437 3.0.101-0.46 (SLES11 SP3)
438 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
439 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
440 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
441 stack overflow. For more information, please refer to bugzilla 17630.
442 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
443 the re-exporting nodes to be mounted with "32bitapi" mount option
445 Severity : enhancement
447 Description: control OST-index in IDIF via ROCOMPAT flag.
448 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
449 last_rcvd file. For new formatted OST device, it will be auto set;
450 for the case of upgrading from old OST device, you can enable it
451 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
452 enabled, for new created OST-object, its IDIF-in-LMA will contain
453 the OST-index; for the existing OST-object, the OSD will convert
454 old format IDIF as new format IDIF with OST-index stored in the
455 LMA EA when accessing such OST-object or via OI scrub. Once such
456 flag is enabled, it cannot be reverted back, so the system cannot
457 be downgraded to the orignal incompatible version.
459 --------------------------------------------------------------------------------
461 07-30-2014 Intel Corporation
463 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
464 for currently supported client and server kernel versions.
465 * Server known to build on patched kernels:
466 2.6.32-431.20.3.el6 (RHEL6.5)
467 3.0.101-0.31 (SLES11 SP3)
468 * Client known to build on unpatched kernels:
469 2.6.32-431.20.3.el6 (RHEL6.5)
470 3.0.101-0.31 (SLES11 SP3)
473 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
474 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
475 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
476 stack overflow. For more information, please refer to bugzilla 17630.
477 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
478 the re-exporting nodes to be mounted with "32bitapi" mount option
480 --------------------------------------------------------------------------------
482 10-31-2013 Intel Corporation
484 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
485 * Server builds for kernels:
486 2.6.32-358.18.1.el6 (RHEL6.4)
487 * Client builds for unpatched kernels:
488 2.6.32-358.18.1.el6 (RHEL6.4)
489 3.0.92-0.8 (SLES11 SP3)
491 * Recommended e2fsprogs version: 1.42.7.wc2 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 05-30-2013 Intel Corporation
502 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
503 * Server builds for kernels:
504 2.6.32-358.6.2.el6 (RHEL6.4)
505 * Client builds for unpatched kernels:
506 2.6.32-358.6.2.el6 (RHEL6.4)
507 3.0.74-0.6.6 (SLES11 SP2)
509 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
510 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
511 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
512 stack overflow. For more information, please refer to bugzilla 17630.
513 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
514 the re-exporting nodes to be mounted with "32bitapi" mount option
516 --------------------------------------------------------------------------------
518 9-30-2012 Whamcloud, Inc.
520 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
521 * Server support for kernels:
522 2.6.32-279.5.1.el6 (RHEL6)
523 * Client support for unpatched kernels:
524 2.6.32-279.5.1.el6 (RHEL6)
526 2.6.18-238.19.1.el5 (RHEL5)
527 2.6.16.60-0.69.1 (SLES10)
528 * Recommended e2fsprogs version: 1.41.90.wc4
529 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
530 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
531 stack overflow. For more information, please refer to bugzilla 17630.
532 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
533 the re-exporting nodes to be mounted with "32bitapi" mount option
535 Severity : enhancement
537 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
539 Severity : enhancement
541 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
543 Severity : enhancement
545 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
547 Severity : enhancement
549 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
551 Severity : enhancement
553 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
555 --------------------------------------------------------------------------------
557 04-30-2012 Whamcloud, Inc.
559 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
560 * Server support for kernels:
561 2.6.32-220.4.2.el6 (RHEL6)
562 * Client support for unpatched kernels:
563 2.6.32-220.4.2.el6 (RHEL6)
565 2.6.18-238.19.1.el5 (RHEL5)
566 2.6.16.60-0.69.1 (SLES10)
567 * Recommended e2fsprogs version: 1.41.90.wc4
568 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
569 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
570 stack overflow. For more information, please refer to bugzilla 17630.
571 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
572 the re-exporting nodes to be mounted with "32bitapi" mount option
574 Severity : enhancement
576 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
578 Severity : enhancement
580 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
582 Severity : enhancement
584 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
586 Severity : enhancement
588 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
590 --------------------------------------------------------------------------------
592 09-30-2011 Whamcloud, Inc.
594 * Server support for kernels:
595 2.6.18-238.19.1.el5 (RHEL5)
596 2.6.32-131.6.1.el6 (RHEL6)
597 * Client support for unpatched kernels:
598 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
599 2.6.18-238.19.1.el5 (RHEL5)
600 2.6.32-131.6.1.el6 (RHEL6)
601 2.6.29.4-167.fc11 (FC11)
602 2.6.16.60-0.69.1 (SLES10)
604 * Recommended e2fsprogs version: 1.41.90.wc3
605 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
606 Lustre filesystem with 4K stack may cause a stack overflow. For more
607 information, please refer to bugzilla 17630.
608 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
609 mounted with "32bitapi" mount option
611 Severity : enhancement
612 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
614 Severity : enhancement
615 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
617 Severity : enhancement
618 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
621 Frequency : only with SLES10
623 Description: Use OFED "KMP" provided by Novell
624 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
625 Lustre is now built against this package. That means you need to
626 install the ofed-kmp package from Novell for the patchless client.
627 Note that the ofed-kmp that Novell ships may not exactly match the
628 kernel version but should still be compatible.
630 Severity : enhancement
632 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
633 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
635 --------------------------------------------------------------------------------
637 2010-07-15 Oracle, Inc.
639 * Server support for kernels:
640 2.6.18-164.11.1.el5 (RHEL 5)
641 2.6.18-164.11.1.0.1.el5 (OEL 5)
642 * Client support for unpatched kernels:
643 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
644 2.6.18-164.11.1.el5 (RHEL 5),
645 2.6.18-164.11.1.0.1.el5 (OEL 5)
646 2.6.16.60-0.42.8 (SLES 10),
648 2.6.29.4-167.fc11 (FC11)
649 * Recommended e2fsprogs version: 1.41.10-sun2
651 Severity : enhancement
653 Description: Distribute lfs_migrate script with lustre
654 Details : The lfs_migrate mentioned in the manual is now included in the
655 lustre tree in lustre/scripts and will distributed in lustre RPMs.
657 Severity : enhancement
659 Description: Failover nid list added to import proc file.
661 Severity : enhancement
663 Description: "som_preview" mount option required to use SOM
664 Details : The clients will need to use som_preview mount option
668 Description: Lustre does not do 1MB IOs to HW RAID
669 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
671 Severity : enhancement
673 Description: lfs quota output is not very convenient for awk/sed-parsing
674 Details : Some positions in lfs quota output table could be empty or
675 non-empty which made it hard to parse it with scripts, now
676 a dash is put instead of space where there is not supposed
681 Description: Update to OFED 1.5.1
682 Details : OFED 1.5.1 will be used in environments that don't use
683 vendor supplied OFED.
687 Description: Remove lnet. prefix from lctl params display
688 Details : Also renamed lnet.memused to lnet_memused to
693 Description: Use conf_param -d to remove permanant settings.
697 Description: Add list_param -R
698 Details : lctl list_param now supports recursive mode.
701 Bugzilla : 21829, 22312
702 Description: Fix llobdstat.
703 Details : 1. Now updating the write_bytes counter
704 2. Port script fixes from 1.8.
708 Description: decrease the usage of memory on clients.
709 Details : 1. On clients, recycle dentries and inodes unused.
710 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
715 Description: llobdstat fix and enhancement
716 Details : add a counter to set a limit to how many samples will be returned
717 fix a wildcard in the path to limit to obdfilter stats only
721 Description: lfs check only allowed for root
722 Details : Code cleanups and added a regresion test of non-root use of
727 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
728 Details : Fix hr->hr_index overload.
732 Description: Fix FIEMAP functionality.
736 Description: FC11 patchless client support added
740 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
741 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
745 Description: New acc-sm script and various test-framework enhancments.
749 Description: Previously landed patch was reverted.
753 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
757 Description: Add ldiskfs maxdirsize mount option.
761 Description: Fix mkfs.lustre support for 16TB LUNs.
765 Description: Print mmp_check_interval
766 Details : print mmp_check_interval and make it possible to abort
767 mount operation in case it takes too long.
771 Description: Add cascading_rw to lustre mpi tests.
775 Description: Store SOM data in EA.
776 Details : Also several SOM related cleanups.
780 Description: Add refcount to md_open_data
784 Description: Use refcount on lov_request_set to prevent memory corruption.
788 Description: Statahead: use parent dir's i_mutex for synchronization.
792 Description: loadgen improvements.
796 Description: Enable peer health detection in ksocklnd and ko2iblnd.
800 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
804 Description: Fix race on obd_devcie:md_stats
808 Description: Fix SLES 11 client builds.
809 Details : Resolved uninitalized pointer gcc warnings.
813 Description: Allow non-root access for "lfs check".
814 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
818 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
820 Severity : enhancement
822 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
823 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
827 Description: Sanity test_27p fails.
828 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
829 Sanity tests 27 were cleaned up and improved.
833 Description: Unify req format on client/servers
834 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
835 buffers and swabbers approach..
840 Description: drop join file stuff.
842 Severity : enhancement
844 Description: Add 2.6.27 and SLES11 (patchless client) support
848 Description: Fix in ptlrpc_expire_one_request() to print the signed time
851 Severity : enhancement
853 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
856 Severity : enhancement
858 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
860 Severity : enhancement
862 Description: correctly handle big reply message.
863 Details : send LNet event if reply is bigger than buffer and adjust this buffer
868 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
870 Severity : enhancement
872 Description: Add support for OEL5.
876 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
878 Severity : enhancement
880 Description: "weak-modules" support
881 Details : Implement "weak-modules" support which enables kernel modules
882 to be used with any kernel that implements the same kABI. In
883 order to achieve this modules are now installed in
884 /lib/modules/$(uname -r)/updates/kernel on all distributions.
888 Description: client could not reconnect to OST because of an active request.
889 Details : The client now retries to reconnect to the same server, if a
890 connect request failed with EBUSY or -EAGAIN. Test case 35b
891 conf-sanity.sh has been created.
895 Description: Don't allow make backward step on assignin osc next id.
896 Details : race between allocation next id and ll_sync thread can be cause
897 of set wrong osc next id and can be kill valid ost objects.
901 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
905 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
909 Description: Error handling in osc_statfs_interpret() has been improved.
910 Details : Check in osc_statfs_interpret() for EBADR.
914 Description: Do not send statfs() requests to OSTs disabled by administrator.
915 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
917 Severity : enhancement
919 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
921 Severity : enhancement
923 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
925 Severity : enhancement
927 Description: Enhancement for lfs(1) command to use numeric uid/gid.
931 Description: Improved cleanup procedure for conf-sanity test_46a.
935 Description: Do not start bulk transfer, if client tried to reconnect.
936 Details : Do not start bulk transfer, if client tried to reconnect,
937 but failed due to active rpcs. Check for exp_abort_active_req
938 has been implemented.
942 Description: Prevent inconsistences between linux and lustre mount structures.
943 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
944 Make the sleep interruptible.
946 Severity : enhancement
948 Description: provide server to client comms path
949 Details : server to client communications path via new LDLM_SET_INFO rpc
953 Description: truncate starts GFP_FS allocation under transaction causing deadlock
954 Details : ldiskfs_truncate calls grab_cache_page which may start page
955 allocation under an open transaction. This may lead to
956 calling prune_icache with consequent lustre reentrance.
960 Description: send correctly lsm on open replay
961 Details : MDS is trust to LSM size on replay open, but client can set wrong size
966 Description: enable client interface failover
967 Details : When a child reconnects from another NID, properly update export
968 nid hash position and ldlm reverse import.
970 Severity : enhancement
972 Description: Communicate OST degraded/readonly state via statfs to MDS
973 Details : Flags in the statfs returned from OSTs indicate whether the
974 OST is in a degraded RAID state, or if the filesystem has
975 turned read-only after a filesystem error is detected.
980 Description: don't panic if EPROTO was hit when reading symlink
981 Details : correctly handling request reference in error cases.
983 Severity : enhancement
985 Description: Add LustreNetLink, a kernel-userspace communcation path.
987 Severity : enhancement
989 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
994 Description: access to llog context before init.
995 Details : move handling CATALOGS file at osc layer and forbid access to llog
999 Frequency : rare, only when client is writing to OST on local node
1001 Description: Avoid deadlock for local client writes
1002 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1003 memory freeing context. This allows OST threads to set the
1004 PF_MEMALLOC flag on task structures in order to allocate memory
1005 from reserved pools and complete IO.
1006 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1007 so that the OST threads generate memory pressure and allow
1008 inactive pages to be reclaimed.
1010 Severity : enhancement
1012 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1017 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1018 Details : move ldlm namespace creation in setup phase to avoid grab
1019 _lprocfs_lock with cli_sem held.
1023 Description: Temporarily disable grant shrink.
1024 Details : Disable the feature for debugging.
1027 Frequency : only when formatting test filesystems
1029 Description: Unable to run several mkfs.lustre on loop devices at the same time
1030 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1031 formatting. The solution is to proper handle the error.
1033 Severity : enhancement
1035 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1037 Severity : enhancement
1039 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1044 Description: resolve race between obd_disconnect and class_disconnect_exports
1045 Details : if obd_disconnect will be called to already disconnected export he
1046 forget release one reference and osc module can't unloaded.
1048 Severity : enhancement
1050 Description: Allow tuning service thread via /proc
1051 Details : For each service a new
1052 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1053 created that can be used to set min/max thread counts, and get the
1054 current number of running threads.
1059 Descriptoin: don't return error if have partially created objects for file.
1060 Details : lov_update_create_set uses set->set_success as index for created
1061 objects, so if some requests will be failed, they will have hole
1062 at end of array and qos_shrink_lsm for allocate correct lsm.
1064 Severity : enhancement
1066 Description: Update OFED support to 1.4.1
1071 Description: Reduce small size read RPC
1072 Details : Set read-ahead limit for every file and only do read-ahead when
1073 available read-ahead pages are bigger than 1M to avoid small size
1076 Severity : enhancement
1078 Description: move AT tunable parameters for more consistent usage
1079 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1081 Severity : enhancement
1083 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1084 Details : allow skip disconnected ost for send statfs request and hide error
1090 Description: fix racy locking of mballoc block bitmaps causing BUG
1091 Details : The locking of the mballoc buddy bitmap and the in-memory
1092 block bitmap was using two different spin locks in some
1093 cases. This made it possible to incorrectly access the
1094 mballoc bitmap while another process was modifying it,
1095 causing a sanity assertion to fail. While no on-disk corruption
1096 was reported, there was some risk of this happening.
1099 Frequency : rare, on llog test 6
1101 Descriptoin: don't allow connect to already connected import
1102 Details : allowing connect to already connected import is hide connecting problem.
1107 Descriptoin: allow kill process which wait statahead result
1108 Details : for some reasons 'ls' can stick in waiting result from statahead,
1109 in this case need way for kill this process.
1111 Severity : enhancement
1113 Description: Add state history info file, enhance import info file
1114 Details : Track import connection state changes in a new osc/mdc proc file;
1115 add overview-type data to the osc/mdc import proc file.
1117 Severity : enhancement
1119 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1121 Severity : enhancement
1123 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1128 Descriptoin: OST grant too much space to client even there are not enough space.
1129 Details : Client will shrink its grant cache to OST if there are no write
1130 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1131 this grant cache if there are already not enough avaible space
1132 (left_space < total_clients * 32M).
1135 Frequency : start MDS on uncleanly shutdowned MDS device
1137 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1138 Details : stay in waiting mds<>ost recovery finished produce random bugs
1139 due race between two ll_sync thread for one lov target. send
1140 ACTIVATE event only if connect realy finished and import have
1144 Frequency : rare, connect and disconnect target at same time
1146 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1147 Details : don't call obd_disconnect under lov_lock. this long time
1148 operation and can block ptlrpcd which answer to connect request.
1153 Descriptoin: don't lose wakeup for imp_recovery_waitq
1154 Details : recover_import_no_retry or invalidate_import and import_close can
1155 both sleep on imp_recovery_waitq, but we was send only one wakeup
1159 Frequency : always with long access acl
1161 Descriptoin: mds can't pack reply with long acl.
1162 Details : mds don't control size of acl but they limited by reint/getattr
1165 Severity : enhancement
1167 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1169 Severity : enhancement
1171 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1174 Frequency : start MDS on uncleanly shutdowned MDS device
1176 Descriptoin: aborting recovery hang on MDS
1177 Details : don't throttle destroy RPCs for the MDT.
1180 Frequency : on remount
1182 Description: external journal device not working after the remount
1183 Details : clear dev_rdonly flag for external journal devices in
1189 Description: shutdown vs evict race
1190 Details : client_disconnect_export vs connect request race.
1191 if client will evicted at this time - we start invalidate
1192 thread without referece to import and import can be freed
1196 Frequency : rare, need acl's on inode.
1198 Description: client can't handle ost additional correctly
1199 Details : if ost was added after client connected to mds client can have
1200 hit lnet_try_match_md ... to big messages to wide striped files.
1201 in this case need teach client to handle config events about add
1202 lov target and update client max ea size at that event.
1204 Severity : enhancement
1206 Description: Changelogs
1207 Details : Changelogs are a lightweight mechanism to track filesystem
1208 metadata and namespace changes. The changelog is recorded
1209 permanently on the MDTs, and is periodically "consumed" / purged
1210 when records are no longer needed.
1212 Severity : enhancement
1214 Description: compact fld format with extents
1215 Details : Store range of seq rather than every seq in FLD. Seq
1216 controller update FLD rather than clients. In Case of CMD, mdt0
1217 has FLD, all other metadata server act as non persistent proxy
1218 for FLD queries and cache fld entries in fld cache.
1223 Description: don't skip ost target if they assigned to file
1224 Details : Drop slow OSCs if we can, but not for requested start idx.
1225 This means "if OSC is slow and it is not the requested
1226 start OST, then it can be skipped, otherwise skip it only
1227 if it is inactive/recovering/out-of-space.
1231 Description: more cleanup in mds_lov
1232 Details : not send LOV EA under replay, we can't know about they size at this
1233 time. Don't allow client connect to mds before any ost connected,
1234 for avoid problems with LOV EA size and returning EIO to client.
1236 Severity : enhancement
1238 Description: Interoperability at server side (Disk interoperability)
1240 Severity : enhancement
1242 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1244 Severity : enhancement
1246 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1248 Severity : enhancement
1250 Description: New client IO stack (CLIO).
1252 Severity : enhancement
1254 Description: Commit on sharing. Eliminate inter-client dependencies between
1255 uncommitted transactions by doing transaction commits.
1256 Thereby clients may recovery independently.
1259 Frequency : Create a symlink file with a very long name
1261 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1262 Details : If there is no extra space in the request for early cancels,
1263 ldlm_req_handles_avail() returns 0 instead of a negative value.
1265 Severity : enhancement
1267 Description: Add /proc entry for import status
1268 Details : The mdc, osc, and mgc import directories now have
1269 an import directory that contains useful import data for debugging
1270 connection problems.
1272 Severity : enhancement
1274 Description: Re-disable certain /proc logging
1275 Details : Enable and disable client's offset_stats, extents_stats and
1276 extents_stats_per_process stats logging on the fly.
1279 Frequency : Only on FC kernels 2.6.22+
1281 Description: oops in statahead
1282 Details : Do not drop reference count for the dentry from VFS when lookup,
1283 VFS will do that by itself.
1285 Severity : enhancement
1287 Description: Generic /proc file permissions
1288 Details : Set /Proc file permissions in a more generic way to enable non-
1289 root users operate on some /proc files.
1293 Description: Hitting mdc_commit_close() ASSERTION
1294 Details : Properly handle request reference release in
1295 ll_release_openhandle().
1299 Description: quota recovery deadlock during mds failover
1300 Details : This patch includes att18982, att18236, att18237 in bz14840.
1302 1. fix osts hang when mds does failover with quotaon
1303 2. prevent watchdog storm when osts threads wait for the
1308 Frequency : only patchless client
1309 Description: add workaround for race between add/remove dentry from hash
1311 Severity : enhancement
1313 Description: Allow OST glimpses to return PW locks
1317 Description: LBUG when llog conf file is full
1318 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1322 Description: Prevent import from entering FULL state when server in recovery
1326 Description: service mount cannot take device name with ":"
1327 Details : Only when device name contains ":/" will mount treat it as
1333 Description: replace ptlrpcd with the statahead thread to interpret the async
1334 statahead RPC callback
1338 Frequency : on recovery
1339 Description: I/O failures after umount during fail back
1340 Details : if client reconnected to restarted server we need join to recovery
1341 instead of find server handler is changed and process self
1342 eviction with cancel all locks.
1344 Severity : enhancement
1346 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1348 Severity : enhancement
1350 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1352 Severity : enhancement
1354 Description: Upcall on Lustre log has been dumped
1355 Details : Allow for a user mode script to be called once a Lustre log has
1356 been dumped. It passes the filename of the dumped log to the
1357 script, the location of the script can be specified via
1358 /proc/sys/lnet/debug_log_upcall.
1363 Description: avoid idr_remove called for id which is not allocated.
1364 Details : Move assigment s_dev for clustered nfs to end of initialization,
1365 to avoid problem with error handling.
1370 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1371 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1372 into CONN_USED_HASH and this prodice warning when put connection
1373 again in unused hash.
1376 Severity : enhancement
1378 Description: File striping can now be set to use an arbitrary pool of OSTs.
1380 Severity : enhancement
1382 Description: Export bytes_read/bytes_write count on OSC/OST.
1386 Description: Early reply size mismatch, MGC loses connection
1387 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1388 the connect flags are properly negotiated.
1393 Description: quotas are not honored with O_DIRECT
1394 Details : all writes with the flag O_DIRECT will use grants which leads to
1395 this problem. Now using OBD_BRW_SYNC to guard this.
1399 Description: add quota statistics
1400 Details : 1. sort out quota proc entries and proc code.
1401 2. add quota statistics
1403 Severity : enhancement
1405 Description: enable quota support for HEAD.
1409 Description: Properly propagate oinfo flags from lov to osc for statfs
1410 Details : restore missing copy oi_flags to lov requests.
1412 Severity : enhancement
1414 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1415 : l_getgroups(8), lst(8), routerstat(8)
1416 Details : included man pages for llobdstat(8), llstat(8),
1417 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1419 Severity : enhancement
1421 Description: configure's --enable-quota should check the
1422 : kernel .config for CONFIG_QUOTA
1423 Details : configure is terminated if --enable-quota is passed but
1424 : no quota support is in kernel
1428 Description: Remove portals compatibility
1429 Details : Remove portals compatibility, not interoperable with releases
1434 Description: Resolve device initialization race
1435 Details : Prevent proc handler from accessing devices added to the
1436 obd_devs array but yet be intialized.
1438 Severity : enhancement
1440 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1442 Severity : enhancement
1444 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1449 Description: Using wrong pointer in osc_brw_prep_request
1450 Details : Access to array[-1] can produce panic if kernel compiled with
1451 CONFIG_PAGE_ALLOC enabled
1455 Description: Client runs out of low memory
1456 Details : Consider only lowmem when counting initial number of llap pages
1460 Description: *optional* service tags registration
1461 Details : if the "service tags" package is installed on a Lustre node
1462 When the filesystem is mounted, a local-node service tag will
1463 be created. See http://inventory.sun.com/ for more information
1464 about the Service Tags asset management system.
1468 Description: Kernel BUG tries to release flock
1469 Details : Lustre does not destroy flock lock before last reference goes
1470 away. So always drop flock locks when client is evicted and
1471 perform unlock regardless of successfulness of speaking to MDS.
1475 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1479 Description: sanity test 65a fails if stripecount of -1 is set
1480 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1485 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1486 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1489 Severity : enhancement
1491 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1495 Description: do not process already freed flock
1496 Details : flock can possibly be freed by another thread before it reaches
1497 to ldlm_flock_completion_ast.
1501 Description: LBUG during stress test
1502 Details : Need properly lock accesses the flock deadlock detection list.
1506 Description: oops in page fault handler
1507 Details : kernel page fault handler can return two special 'pages' in error
1508 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1512 Description: timeout with invalidate import.
1513 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1514 should be handled by ptlrpcd. This produce long age waiting and
1515 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1517 Severity : enhancement
1519 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1523 Description: enable MGS and MDT services start separately
1524 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1525 a co-located MGS without starting the MGS, which is a complement
1526 to 'nosvc' mount option.
1530 Frequency : after recovery
1531 Description: precreate to many object's after del orphan.
1532 Details : del orphan st in oscc last_id == next_id and this triger growing
1533 count of precreated objects. Set flag LOW to skip increase count
1534 of precreated objects.
1538 Frequency : rare, on clear nid stats
1539 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1540 Details : when clean nid stats sometimes try destroy live entry,
1541 and this produce panic in free.
1545 Description: Stack overflow during MDS log replay
1546 ease stack pressure by using a thread dealing llog_process.
1550 Description: wait until IO finished before start new when do lock cancel.
1551 Details : VM protocol want old IO finished before start new, in this case
1552 need wait until PG_writeback is cleared until check dirty flag and
1553 call writepages in lock cancel callback.
1555 Severity : enhancement
1557 Description: using special macro for print time and cleanup in includes.
1561 Description: mds_mfd_close() ASSERTION(rc == 0)
1562 Details : In mds_mfd_close(), we need protect inode's writecount change
1563 within its orphan write semaphore to prevent possible races.
1567 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1571 Frequency : rare, on shutdown ost
1572 Description: don't hit live lock with umount ost.
1573 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1574 use shrink_dcache_sb instead.
1578 Description: don't panic with use echo client
1579 Details : echo client pass NULL as client nid pointer and this produce null
1580 pointer dereference.
1584 Description: fix build on ppc32
1585 Details : compile code with -m64 flag produce wrong object file for ppc32.
1589 Description: add message levels for liblustreapi
1593 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1594 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1595 will result in an -EAGAIN and in the exit call path we call
1596 alloc_rr() which will with increasing aggressiveness attempt to
1597 aquire precreated objects on the minimum number of required OSCs.
1601 Description: Use old size assignment to avoid deadlock
1602 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1603 the scheduling while holding a spinlock. We do not need locking
1604 for size in ll_update_inode() because size is only updated from
1605 the MDS for directories or files without objects, so there is no
1606 other place to do the update, and concurrent access to such inodes
1607 are protected by the inode lock.
1611 Description: resolve "_IOWR redefined" build error on SLES10
1615 Description: dump the memory debugging after all modules are unloaded to
1616 suppress false negative in conf_sanity test 39
1618 Severity : enhancement
1620 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1625 Description: fix SLES kernel versioning
1626 Details : the kernel version for our SLES 10 kernel did not include a "-"
1627 before the "smp" at the end. while this was not a problem in
1628 general, it did mean that software trying to use the kernel
1629 version to try to detect a vendor specific kernel would fail.
1630 this was most evident by the OFED build scripts.
1634 Description: Don't update lov_desc members until making sure they are valid
1635 Details : When updating lov_desc members via proc fs, need fix their
1636 validities before doing the real update.
1640 Description: don't put request into delay list while invalidate in flight.
1641 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1642 import in flight. this produce timeout for invalidate and sometimes
1643 can cause stale data.
1647 Frequency : on ppc only
1648 Description: not convert ost objects for directory because it's not exist.
1649 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1653 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1655 Severity : enhancement
1657 Description: Update to support 2.6.22.14 vanilla kernel.
1661 Frequency : rare, on recovery
1662 Description: read procfs can produce deadlock in some situation
1663 Details : Holding lprocfs lock which send rpc can produce block for destroy
1664 obd objects and this also block reconnect with -EALREADY. This
1665 isn't fix all lprocfs bugs - but make it rare.
1667 Severity : enhancement
1669 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1672 Frequency : frequent on X2 node
1674 Description: mdc_set_open_replay_data LBUG
1675 Details : Set replay data for requests that are eligible for replay.
1679 Description: lustre_mgs: operation 101 on unconnected MGS
1680 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1681 MGC, and late on MGC cannot successfully connect to MGS and a lot
1682 of the error messages complaining that MGS is not connected.
1686 Frequency : on network error
1687 Description: panic with double free request if network error
1688 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1689 but it's true only for synchronus enqueue, for async enqueue
1690 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1693 Severity : enhancement
1695 Description: client-side metadata stat-ahead during readdir(directory readahead)
1696 Details : perform client-side metadata stat-ahead when the client detects
1697 readdir and sequential stat of dir entries therein
1700 Frequency : on start mds
1702 Description: Implement get_info(last_id) in obdfilter.
1705 Frequency : occasional
1707 Description: Correctly check stale fid, not start epoch if ost not support SOM
1708 Details : open with flag O_CREATE need set old fid in op_fid3 because
1709 op_fid2 was overwritten with new generated fid, but mds can answer
1710 with one of these two fids and both is not stale. Setattr
1711 incorrectly started an epoch and assume will be called
1712 done_writing, but without SOM done_writing ever being called.
1715 Frequency : rare, depends on device drivers and load
1717 Description: MDS or OSS nodes crash due to stack overflow
1718 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1719 In some cases, in conjunction with device drivers that use a lot
1720 of stack the MDS (or possibly OSS) service threads could overflow
1721 the stack. One change which was identified to consume additional
1722 stack has been reworked to avoid the extra stack usage.
1725 Frequency : occasional
1727 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1728 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1729 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1730 oscc_has_objects and see OSC has no objests and call
1731 oscc_internal_create to resend create request.
1733 Severity : enhancement
1735 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1737 Severity : enhancement
1739 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1742 Frequency : very rare
1744 Description: Fix replay if there is an un-replied request and open
1745 Details : In some cases, older replay request will revert the
1746 mcd->mcd_last_xid on MDS which is used to record the client's
1747 latest sent request.
1749 Severity : enhancement
1751 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1753 Severity : enhancement
1755 Description: Add rhel5 support to HEAD.
1757 Serverity : enhancement
1759 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1764 Description: Don't allow skipping OSTs if index has been specified.
1765 Details : Don't allow skipping OSTs if index has been specified, make
1766 locking in internal create lots better.
1770 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1771 Details : access to bitfield in structure is always rounded to long
1772 and this produce problem with not atomic change any bit.
1776 Description: Lustre make rpms failed.
1777 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1778 builds Lustre rpms from tarball.
1781 Frequency : rare on shutdown ost
1783 Description: If llog cancel was not send before clean_exports phase, this can
1784 produce deadlock in llog code.
1785 Details : If llog thread has last reference to obd and call class_import_put
1786 this produce deadlock because llog_cleanup_commit_master wait when
1787 last llog_commit_thread exited, but this never success because was
1788 called from llog_commit_thread.
1792 Description: allow userland application know is lost one of stripes.
1793 Details : fill lvb_blocks with error code on ost and return it to
1794 application if error flag found.
1798 Description: NULL lov_tgts causing MDS oops
1799 Details : more safe checks for NULL lov_tgts for avoid oops.
1801 Severity : enhancement
1803 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1807 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1808 Details : If an OST is down the MDS will hang indefinitely in
1809 obd_statfs_rqset() waiting for the statfs data. While for
1810 MDS QOS usage of statfs, it should not stuck in waiting.
1812 Severity : enhancement
1814 Description: remote_acl support
1815 Details : Support ACL-based permission check for remote user.
1816 Support setfacl/getfacl for remote user with the utils
1817 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1818 the system "{s,g}etfacl" utils.
1820 Severity : enhancement
1822 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1824 Severity : enhancement
1826 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1830 Description: make mgs_setparam() handle fsname containing dash
1831 Details : fsname containing a dash does not work with lctl conf_param
1833 Severity : enhancement
1835 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1837 Severity : enhancement
1839 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1843 Frequency : rare, at shutdown
1844 Description: access already free / zero obd_namespace.
1845 Details : if client_disconnect_export was called without force flag set,
1846 and exist connect request in flight, this can produce access to
1847 NULL pointer (or already free pointer) when connect_interpret
1848 store ocd flags in obd_namespace.
1852 Frequency : only at startup
1853 Description: not alloc memory with spinlock held.
1854 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1855 if any spinlock held.
1857 Severity : enhancement
1859 Description: make lustre randomly fail allocating memory
1860 Details : Make lustre randomly failed allocating memory for testing purpose.
1862 Severity : enhancement
1864 Description: lost problems with lov objid file
1865 Details : Fixes some scability and access to not inited memory problems
1866 in work with lov objdid file.
1871 Description: lfs find does not continue on file error
1872 Details : Continue other files processing when a file/dir is absent.
1876 Description: Inconsistent usage of lustre_pack_reply()
1877 Details : Standardize the usage of lustre_pack_reply() such that it
1878 always generate a CERROR on failure.
1881 Frequency : occasional
1883 Description: MDS hang or stay in waiting lock
1884 Details : If client receive lock with CBPENDING flag ldlm need send lock
1885 cancel as separate rpc, to avoid situation when cancel request
1886 can't processed due all i/o threads stay in wait lock.
1890 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1894 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1898 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1902 Description: Update to RHEL4 latest kernel.
1904 Severity : enhancement
1906 Description: Build SLES10 patchless client fails
1907 Details : The configure was broken by run ./configure with
1908 --with-linux-obj=.... argument for patchless client. When the
1909 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1910 can't search header adequately. Use absolute path such as
1911 -I($LINUX)/include instead.
1915 Description: interrupt oig_wait produce painc on resend.
1916 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1917 private set, and this produce situation when rq_ptlrpcd_data not
1918 copyed to new allocated request and triggered LBUG on assert
1919 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1920 ptlrpcd set if request is changed and can be safety changed to use
1925 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1926 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1927 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1928 these macros and replace them with their difinitions instead.
1932 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1933 Details : Update which_patch & target file for SLES10 latest kernel.
1936 Description: Remove client patches from SLES 10 kernel.
1937 Details : This causes SLES 10 clients to behave as patchless clients
1938 even on a Lustre-patched (server) kernel.
1940 Severity : enhancement
1942 Description: self-adjustable client's lru lists
1943 Details : use adaptive algorithm for managing client cached locks lru
1944 lists according to current server load, other client's work
1945 pattern, memory activities, etc. Both, server and client
1946 side namespaces provide number of proc tunables for controlling
1949 Severity : enhancement
1951 Description: light-weight GSS support
1952 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1953 and reduce performance overhead.
1955 Severity : enhancement
1957 Description: Linux keyring support
1958 Details : Support using service of Linux keyring for Lustre GSS internal
1959 context refresh/cache mechanism.
1963 Description: Fix errors in lfs documentation
1964 Details : Fixes man pages
1968 Description: don't use GFP_* in generic Lustre code.
1969 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1973 Description: obdclass is limited by single OBD_ALLOC(idarray)
1974 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1978 Description: warnings with build patchless client with vanila 2.6.19 and up
1979 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1980 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1984 Frequency : Only for SLES
1985 Description: sanity_quota fail test_1
1986 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1987 file, which makes TSTID[2] inunique.
1991 Frequency : Only for SLES9
1992 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1993 Details : Update target/ChangeLog/which_patch .
1997 Frequency : Only for test_10 in sanity-quota.sh
1998 Description: a bug in quota test code
1999 Details : checking a test flag in a wrong place
2003 Frequency : Only for test_13 in sanity-quota.sh
2004 Description: a bug in quota test script
2005 Details : shouldn't check overfull allocated quota
2009 Description: jbd statistics
2010 Details : Port older jbd statistics patch for sles10
2012 Severity : enhancement
2014 Description: Add printing inode info into message about error in writepage.
2019 Description: reply_lock_interpret crash due to race with it and lock cancel.
2020 Details : Do not replay locks that are being cancelled. Do not reference
2021 locks by their address during replay, just by their handle.
2025 Frequency : When flocks are used.
2026 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2027 Details : Flock locks might destroy just granted lock if it could be merged
2028 with another existing flock, this is done in completion handler,
2029 so teach ldlm_cli_enquque_fini that this is a valid case for
2035 Description: Oops in read and write path when failing to allocate lock.
2036 Details : Check if lock allocation failed and return error back.
2040 Description: lstripe command fails for valid OST index
2041 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2042 instead of lov->desc.ld_active_tgt_count.
2046 Description: sanity.sh failed test 103
2047 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2052 Description: df doesn't work properly if diskfs blocksize != 4K
2053 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2057 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2058 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2060 Severity : enhancement
2062 Description: Adaptive timeouts
2063 Details : RPC timeouts adapt to changing server load and network
2064 conditions to reduce resend attempts and improve recovery time.
2068 Description: llapi_file_create() does not allow some changes
2069 Details : add llapi_file_open() that allows specifying the mode and
2070 open flags, and also returns an open file handle.
2074 Description: merge and cleanup kernel patches.
2075 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2079 Description: Add journal checksum support.(Kernel part)
2080 Details : The journal checksum feature adds two new flags i.e
2081 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2082 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2083 indicates that the commit block contains the checksum for
2084 the blocks described by the descriptor blocks. Now commit
2085 record can be sent to disk without waiting for descriptor
2086 blocks to be written to disk. This behavior is controlled
2087 using JBD2_FEATURE_ASYNC_COMMIT flag.
2091 Description: OSS needs mutliple precreate threads
2092 Details : Add ability to start more than one create thread per OSS.
2096 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2097 Details : execute lfs setstripe on client
2101 Description: mds_obd_create error creating tmp object
2102 Details : When the user sets quota on root, llog will be affected and can't
2103 create files and write files.
2106 Frequency : Always on ia64 patchless client, and possibly others.
2108 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2109 Details : This allows the patchless client to be loaded on architectures
2110 without this export.
2114 Description: RedHat Update kernel for RHEL5
2115 Details : Add the kernel config file for RHEL5.
2119 Description: RedHat Update kernel for RHEL5
2120 Details : Modify the kernel config file more closer RHEL5.
2124 Description: Build failure against Centos5 (RHEL5)
2125 Details : Define PAGE_SIZE when it isn't present.
2129 Description: test_7 on sanity-quota.sh can't be used on separate machines
2130 Details : add facet to handle it
2134 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2135 Details : with NFS, the anon dentry's parent was set to itself in
2136 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2137 corresponding dentry other than use rec->ur_name.
2139 Severity : enhancement
2141 Description: lfs setstripe enhancement
2142 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2146 Description: enable data checksumming by default
2147 Details : enable checksum by default, allow --disable-checksum
2148 configure option and "-o nochecksum" mount option. Checksums
2149 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2153 Description: lustre support for RHEL5
2154 Details : Add support for RHEL5.
2158 Description: Client eviction due to failover config
2159 Details : after a connection loss, the lustre client should attempt to
2160 reconnect to the last active server first before trying the
2161 other potential connections.
2165 Description: when mds and osts use different quota unit(32bit and 64bit),
2166 quota will be released repeatly.
2167 Details : void sending multiple quota reqs to mds, which will keep the
2168 status between the reqs.
2172 Description: osts not allocated evenly to files
2173 Details : change the condition to increase offset_idx
2176 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2178 Description: Data corruption for OSTs that are formatted larger than 2TB
2180 Details : When generating the bio request for lustre file writes the
2181 sector number would overflow a temporary variable before being
2182 used for the IO. The data reads correctly from Lustre (which
2183 will overflow in a similar manner) but other file data or
2184 filesystem metadata may be corrupted in some cases.
2188 Description: Tune the kernel for good SCSI performance.
2189 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2190 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2195 Description: rewrite ext2-derived code in obdclass/uuid.c
2196 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2197 from scratch preserving functionality.
2201 Description: "lfs find -obd UUID" prints directories
2202 Details : "lfs find -obd UUID" will return all directory names instead
2203 of just file names. It is incorrect because the directories
2204 do not reside on the OSTs.
2208 Description: use i_size_read and i_size_write in 2.6 port
2209 Details : replace inode->i_size access with i_size_read/write()
2212 Frequency : when removing large files
2214 Description: scheduling issue during removal of large Lustre files
2215 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2216 It causes scheduling issues when removing large files (17TB in the
2220 Frequency : only with liblustre clients on XT3
2222 Description: evictions taking too long
2223 Details : allow llrd to evict clients directly on OSTs
2226 Frequency : only on ppc
2228 Description: /proc/fs/lustre/devices broken on ppc
2229 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2230 We should make sure the type of 'index' is loff_t and then cast
2231 explicitly as needed below. Do not assign an explicitly cast
2235 Frequency : only for rhel5
2237 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2238 Details : Modify the target file & which_kernel.
2240 Severity : enhancement
2242 Description: omit set fsid for export NFS
2243 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2244 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2248 Frequency : Always, for kernels after 2.6.16
2249 Description: Fix warning idr_remove called for id=.. which is not allocated.
2250 Details : Last kernels save old s_dev before kill super and not allow
2251 to restore from callback - restore it before call kill_anon_super.
2255 Description: buffer overruns could theoretically occur
2256 Details : llapi_semantic_traverse() modifies the "path" argument by
2257 appending values to the end of the origin string, and a
2258 overrun may occur. Adding buffer overrun check in liblustreapi.
2262 Description: Fix error on 'ls .' at the top of the Lustre mount.
2263 Details : Don't revalidate dentry if it is a root dentry.
2267 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2268 Details : Modify vm-tunables-rhel4.patch.
2272 Description: Kernel config for 2.6.18-vanilla.
2273 Details : Modify targets/2.6-vanilla.target.in.
2274 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2275 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2276 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2277 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2281 Description: improve handling recoverable errors
2282 Details : if request processig with error which can be recoverable on server
2283 request should be resend, otherwise page released from cache and
2288 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2289 Details : Modify target file & which_patch.
2290 A flaw was found in the IA32 system call emulation provided
2291 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2292 value could be stored in the %RAX register, which could trigger an
2293 out-of-bounds system call table access. An untrusted local user
2294 could exploit this flaw to run code in the kernel
2295 (ie a root privilege escalation). (CVE-2007-4573).
2299 Description: O_DIRECT bypasses client statistics.
2300 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2301 read_ahead_stats were updated. copied the stats section from
2302 osc_send_oap_rpc() into async_internal().
2306 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2310 Description: change order of libsysio includes
2311 Details : '#include sysio.h' should always come before '#include xtio.h'
2315 Description: handle "serious error: objid * already exists" more gracefully
2316 Details : If LAST_ID value on disk is smaller than the objects existing in
2317 the O/0/d* directories, it indicates disk corruption and causes an
2318 LBUG(). If the object is 0-length, then we should use the existing
2319 object. This will help to avoid a full fsck in most cases.
2321 Severity : enhancement
2323 Description: adapt the lustre_config script to support the upgrade case
2324 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2325 server targets to 1.6 in parallel.
2329 Description: To avoid grant space > avaible space when the disk is almost
2330 full. Without this patch you might see the error "grant XXXX >
2331 available" or some LBUG about grant, when the disk is almost
2333 Details : In filter_check_grant, for non_grant cache write, we should
2334 check the left space by if (*left > ungranted + bytes), instead
2335 of (*left > ungranted), because only we are sure the left space
2336 is enough for another "bytes", then the ungrant space should be
2337 increase. In client, we should update cl_avail_grant only there
2338 is OBD_MD_FLGRANT in the reply.
2342 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2343 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2344 described in CVE-2007-4573. This problem could allow untrusted
2345 local users to gain root access.
2348 Frequency : when using O_DIRECT and quotas
2350 Description: Incorrect file ownership on O_DIRECT output files
2351 Details : block usage reported by 'lfs quota' does not take into account
2352 files that have been written with O_DIRECT.
2357 Description: touch file failed when fs is not full
2358 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2359 otherwise we can get ENOSP while fs is not full.
2363 Description: parallel lock callbacks
2364 Details : Instead of sending blocking and completion callbacks as separated
2365 requests, adding them to a set and sending in parallel.
2368 Frequency : only for Cray XT3
2369 Bugzilla : 12829/13455
2370 Description: Changing primary group doesn't change the group lustre assigns to
2372 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2373 overridden with the primary group provided by the group upcall,
2374 whereas the supplied fsgid can be trusted if it is in the list of
2375 supplementary groups returned by the group upcall.
2377 Severity : enhancement
2379 Description: Allow masking D_WARNING, D_ERROR messages from console
2380 Details : Console messages can now be disabled via lnet.printk.
2384 Description: User code with malformed file open parameter crashes client node
2385 Details : Before packing join_file req, all the related reference should be
2386 checked carefully in case some malformed flags cause fake
2387 join_file req on client.
2391 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2392 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2393 mode is equal to the granted mode. However, it should also check
2394 that LDLM_FL_CP_REQD is not set.
2398 Description: Heavy nfs access might result in deadlocks
2399 Details : After ELC code landed, it is now improper to enqueue any mds
2400 locks under och_sem, because enqueue might want to decide to
2401 cancel open locks for same inode we are holding och_sem for.
2405 Description: Client eviction while running blogbench
2406 Details : A lot of unlink operations with concurrent I/O can lead to a
2407 deadlock causing evictions. To address the problem, the number of
2408 oustanding OST_DESTROY requests is now throttled to
2409 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2410 callbacks are processed in priority.
2414 Description: enable ACLs on MDS by default
2415 Details : ACLs must be enabled on MDS by default.
2418 Frequency : PPC/PPC64 only
2420 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2421 Details : fix duplicated definitions between asm-ppc64/types.h and
2422 lustre_types.h on PPC.
2425 Frequency : PPC/PPC64 only
2427 Description: asm-ppc/segment.h does not exist
2428 Details : fix compile issue on PPC.
2432 Description: better handle error messages in extents code
2435 Frequency : RHEL4 only
2437 Description: mkfs is very slow on IA64/RHEL4
2438 Details : A performance regression has been discovered in the MPT Fusion
2439 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2440 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2441 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2443 Severity : enhancement
2445 Description: SNMP support enhancement
2446 Details : Adding total number of sampled request for an MDS node in snmp
2449 Severity : enhancement
2451 Description: Optimize ldlm waiting list processing for PR extent locks
2452 Details : When processing waiting list for read extent lock and meeting read
2453 lock that is same or wider to it that is not contended, skip
2454 processing rest of the list and immediatelly return current
2455 status of conflictness, since we are guaranteed there are no
2456 conflicting locks in the rest of the list.
2460 Description: Time out and refuse to reconnect
2461 Details : When the failover node is the primary node, it is possible
2462 to have two identical connections in imp_conn_list. We must
2463 compare not conn's pointers but NIDs, otherwise we can defeat
2464 connection throttling.
2468 Description: port llog fixes from b1_6 into HEAD
2469 Details : Port llog reference couting and some llog cleanups from b1_6
2470 (bug 10800) into HEAD, for protect from panic and access to already
2471 free llog structures.
2475 Description: Detect stride IO mode in read-ahead
2476 Details : When a client does stride read, read-ahead should detect that and
2477 read-ahead pages according to the detected stride pattern.
2481 Description: data checksumming impacts single node performance
2482 Details : add support for several checksum algorithm. Currently, only CRC32
2483 and Adler-32 are supported. The checksum type can be changed on
2484 the fly via /proc/fs/lustre/osc/*/checksum_type.
2488 Description: use adler32 for page checksums
2489 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2494 Description: build for x2 fails
2495 Details : fix compile issue on Cray systems.
2499 Description: Properly match for duplicate locks
2500 Details : Due to different lock order from skiplists code, we need to
2501 traverse entire list for now
2504 Frequency : only on PPC/SLES10
2506 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2507 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2508 be defined. Add a hack in mkfs_lustre.c to work around this compile
2513 Description: LASSERT on MDS when client holding flock lock dies
2514 Details : ldlm pool logic depends on number of granted locks equal to
2515 number of released locks which is not true for flock locks, so
2516 just exclude such locks from consideration.
2520 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2521 Details : Use fsfilt_sync() for both the whole filesystem sync and
2522 individual file sync to eliminate dangerous inode locking
2523 with I_LOCK that can lead to a deadlock.
2527 Description: performance in 1.6.3
2528 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2529 sure that 1MB requests are not fragmented by the block layer.
2531 Severity : enhancement
2533 Description: organize the server-side client stats on per-nid basis
2534 Details : Change the structure of stats under obdfilter and mds to
2542 The "uuid"s file would list the uuids of _active_ exports.
2543 And the clear entry is to clear all stats and stale nids.
2545 Severity : enhancement
2547 Description: eliminate client locks in face of contention
2548 Details : file contention detection and lockless i/o implementation
2549 for contended files.
2553 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2556 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2558 Description: MDS may lose file striping (and hence file data) in some cases
2559 Details : If there are additional extended attributes stored on the MDS,
2560 in particular ACLs, SELinux, or user attributes (if user_xattr
2561 is specified for the client mount options) then there is a risk
2562 of attribute loss. Additionally, the Lustre file striping
2563 needs to be larger than default (e.g. striped over all OSTs),
2564 and an additional attribute must be stored initially in the
2565 inode and then increase in size enough to be moved to the
2566 external attribute block (e.g. ACL growing in size) for file
2571 Description: skiplist implementation simplification
2572 Details : skiplists are used to group compatible locks on granted list
2573 that was implemented as tracking first and last lock of each
2574 lock group the patch changes that to using doubly linked lists
2578 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2579 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2580 because LASSERT on both the data supplied by a client, and the
2581 data on disk is dangerous and incorrect.
2583 Severity : enhancement
2585 Description: Slow truncate/writes to huge files at high offsets.
2586 Details : Directly associate cached pages to lock that protect those pages,
2587 this allows us to quickly find what pages to write and remove
2588 once lock callback is received.
2592 Description: more ldlm soft lockups
2593 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2594 starve other threads from the resource lock for a long time in
2595 case of long waiting queue, so change the debug level from
2596 D_OTHER to the less frequently used D_INFO.
2598 Severity : enhancement
2600 Description: add -gid, -group, -uid, -user options to lfs find
2604 Description: Hung threads in invalidate_inode_pages2_range
2605 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2606 once one is completed. As a result, some RPCs are stuck in the
2607 queue and are never sent.
2611 Description: filter threads hungs on waiting journal commit
2612 Details : Cleanup filter group llog code, then only filter group llog will
2613 be only created in the MDS/OST syncing process.
2617 Description: Procfs and llog threads access destoryed import sometimes.
2618 Details : Sync the import destoryed process with procfs and llog threads by
2619 the import refcount and semaphore.
2621 Severity : enhancement
2623 Description: openlock cache of b1_6 port to HEAD
2628 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2629 Details : If insertion of an extent fails, then discard the inode
2630 preallocation and free data blocks else it can lead to duplicate
2635 Description: don't always update ctime in ext3_xattr_set_handle()
2636 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2637 In some cases the ctime should not be updated, for example for
2638 2.0->1.8 compatibility it is necessary to delete an xattr and it
2639 should not update the ctime.
2643 Bugzilla : 15713/16362
2644 Description: Assertion in iopen_connect_dentry in 1.6.3
2645 Details : looking up an inode via iopen with the wrong generation number can
2646 populate the dcache with a disconneced dentry while the inode
2647 number is in the process of being reallocated. This causes an
2648 assertion failure in iopen since the inode's dentry list contains
2649 both a connected and disconnected dentry.
2653 Description: assertion failure in ldlm_handle2lock()
2654 Details : fix a race between class_handle_unhash() and class_handle2object()
2655 introduced in lustre 1.6.5 by bug 13622.
2660 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2661 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2666 Description: oops in sd_iostats_seq_show()
2667 Details : unloading/reloading the scsi low level driver triggers a kernel
2668 bug when trying to access the sd iostat file.
2673 Description: Kernel panics during QLogic driver reload
2674 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2675 patch, causing memory corruption.
2680 Description: journal_dev option does not work in b1_6
2681 Details : pass mount option during pre-mount.
2683 Severity : enhancement
2685 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2686 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2687 extent information of a file. It can be used to map logical blocks
2688 in a file to physical blocks in the block device.
2692 Description: LDLM soft lockups - improvement
2693 Details : It is be possible to send the lock handle along with each read
2694 or write request because the client is already doing a lock match
2695 itself so there isn't any reason the OST should have to re-do that
2699 Frequency : only on Cray X2
2701 Description: X2 build failures
2702 Details : fix build failures on Cray X2.
2706 Description: xid & resent requests
2707 Details : Initialize RPC XID from clock at startup (randomly if clock is
2710 Severity : enhancement
2712 Description: Add lustre_start utility to start or stop multiple Lustre servers
2717 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2718 Details : In case of memory pressure, list_del() can be called twice on
2719 req->rq_history_list, causing a kernel oops.
2723 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2724 Details : fix stack overflow in the distributed lock manager by defering
2725 export eviction after a failed AST to the elt thread instead of
2726 handling it in the dlm interpret routine.
2730 Description: Convert some comments to new format.
2731 Details : Update documenting comments to match doxygen conventions.
2735 Description: Grammar fixes.
2736 Details : A couple of trivial spelling fixes.
2740 Description: OSD_COUNTERS-mandatory
2741 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2742 to simplify the code.
2746 Description: simplify cmm_device freeing logic.
2747 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2751 Description: Add lockdep support to dt_object_operations locking interface.
2752 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2753 parameter indicating lock ordering. Update mdd code to use new
2758 Description: Introduce failloc constants for lockless IO tests.
2759 Details : Add two new failloc constants to test lockless IO. Only one of
2760 them in implemented---another is checked in yet to be landed
2765 Description: Add lockdep support for inode mutex.
2766 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2767 in the situations where Lustre has to lock more than one inode
2772 Description: Add optional invariants checking support.
2773 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2774 configure switch. This macro is to be used for consistency and
2775 sanity checks that are too expensive to be left in `production'
2780 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2781 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2786 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2787 Details : Use spin_lock_nested() in (the only) situation where more than
2788 one ldlm_lock is locked simultaneously. Also, fix possible
2789 dead-lock in ldlm_lock_change_resource() by enforcing particular
2794 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2795 Details : Instead of specifying each ldlm_lock call-back through separate
2796 parameter, wrap them into struct ldlm_callback_suite.
2800 Description: Kill join_lru obd method and its callers.
2801 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2802 in memory---a requirement implemented through ->o_join_lru() obd
2803 method in HEAD. Remove this method and its users.
2807 Description: Add asynchronous ldlm ENQUEUE completion handler.
2808 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2809 case that stock ldlm_completion_ast() cannot handle as it waits
2810 until lock is granted. Introduce new ldlm_completion_ast_async()
2811 for this. Also comment ldlm_completion_ast().
2815 Description: ldlm_error <-> errno conversion.
2816 Details : Add functions to map (rather arbitrary) between LDLM error codes
2817 and standard errno values. CLIO needs this to prevent LDLM specific
2818 constants from escaping ldlm and osc.
2822 Description: Kill unused ldlm_handle2lock_ns() function.
2823 Details : Kill unused ldlm_handle2lock_ns() function.
2827 Description: Add lu_ref support to ldlm_lock
2828 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2829 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2830 and ldlm_resource. LDLM interface has to be changed, because code that
2831 releases a reference on a lock, has to "know" what reference this is.
2832 In the most frequent case
2834 lock = ldlm_handle2lock(handle);
2836 LDLM_LOCK_PUT(lock);
2838 no changes are required. When any other reference (received _not_ from
2839 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2840 instead of LDLM_LOCK_PUT().
2842 Arguably, changes are pervasive, and interface requires some discipline
2843 for proper use. On the other hand, it was very instrumental in finding
2844 a few leaked lock references.
2848 Description: Add ldlm_lock_addref_try().
2849 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2850 attempts to addref a lock that might be being canceled
2855 Description: Add ldlm_weigh_callback().
2856 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2857 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2858 hard-coded `number of pages' logic.
2862 Description: Add lockdep annotations to llog code.
2863 Details : Use appropriately tagged _nested() locking calls in the places
2864 where llog takes more than one ->lgh_lock lock.
2868 Description: Add loi_kms_set().
2869 Details : Wrap kms updates into a helper function.
2873 Description: Constify instances of struct lsm_operations.
2874 Details : Constify instances of struct lsm_operations.
2878 Description: lu_conf support.
2879 Details : On a server, a file system object is uniquely identified
2880 by a fid, which is sufficient to locate and load all object
2881 state (inode). On a client, on the other hand, more data are
2882 necessary instantiate an object. Change lu_object_find() and
2883 friends to take additional `lu_conf' argument describing object.
2884 Typically this includes layout information.
2888 Description: lu_context fixes.
2889 Details : Introduce new lu_context functions that are needed on the client
2890 side, where some system threads (ptlrpcd) are shared by multiple
2891 modules, and so cannot be stopped during module shutdown.
2895 Description: Add start and stop methods to lu_device_type_operations.
2896 Details : Introduce two new methods in lu_device_type_operations, that are
2897 invoked when first instance of a given type is created and last one
2898 is destroyed respectively. This is need by CLIO.
2902 Description: Add lu_ref support to struct lu_device.
2903 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2904 track leaked references.
2908 Description: Introduce lu_kmem_descr.
2909 Details : lu_kmem_descr and its companion interface allow to create
2910 and destroy a number of kmem caches at once.
2914 Description: Fix lu_object finalization race.
2915 Details : Fix a race between lu_object_find() finding an object and its
2916 concurrent finalization. This race is (most likely) not possible
2917 on the server, but might happen on the client.
2921 Description: Introduce lu_ref interface.
2922 Details : lu_ref is a debugging module allowing to track references to
2923 a given object. It is quite cpu expensive, and has to be
2924 explicitly enabled with --enable-lu_ref. See usage description
2929 Description: Factor lu_site procfs stats into a separate function.
2930 Details : Separate lu_site stats printing code into a separate function
2931 that can be reused on a client.
2935 Description: Constify instances of struct {lu,dt,md}_device_operations.
2936 Details : Constify instances of struct {lu,dt,md}_device_operations.
2940 Description: Introduce struct md_site and move meta-data specific parts of
2941 struct lu_site here.
2942 Details : Move md-specific fields out of struct lu_site into special struct
2943 md_site, so that lu_site can be used on a client.
2947 Description: Kill mdd_lov_destroy().
2948 Details : Remove unused mdd code.
2952 Description: Add st_block checking to multistat.c.
2953 Details : Add st_block checking to multistat.c.
2957 Description: Add lu_ref support to struct obd_device.
2958 Details : Add lu_ref tracking to obd_device.
2962 Description: Kill obd_set_fail_loc().
2963 Details : Remove unused code.
2967 Description: Add special type for ptlrpc_request interpret functions.
2968 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2969 there. Actual usage will be in CLIO.
2973 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2974 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2975 doesn't work with lockdep.
2979 Description: Add rwv.c test program.
2980 Details : New testing program exercising readv(2) and writev(2) (Qian).
2984 Description: Add sendfile.c test program.
2985 Details : New testing program exercising sendfile(2) (Jay).
2989 Description: Ratelimit a message that can be very frequent.
2990 Details : Ratelimit a memory allocation failure message that can
2995 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2996 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3000 Description: Liblustre build fixes.
3001 Details : Liblustre build fixes.
3005 Description: libcfs: add cfs_{need,cond}_resched() interface.
3006 Details : libcfs: add cfs_{need,cond}_resched() definition and
3007 implementations for Linux, NT, and liblustre.
3009 Severity : enhancement
3011 Description: More exported tunables for mballoc
3012 Details : Add support for tunable preallocation window and new tunables for
3013 large/small requests
3017 Description: Detect corruption of block bitmap and checking for preallocations
3018 Details : Checks validity of on-disk block bitmap. Also it does better
3019 checking of number of applied preallocations. When corruption is
3020 found, it turns filesystem readonly to prevent further corruptions.
3024 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3025 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3026 window to 0. If it does hit cache miss, and read-pattern is still
3027 stride-io mode, does not reset the stride window, but also does not
3028 increase the stride window length in this case.
3032 Frequency : only for big-endian servers
3033 Description: Check if system is big-endian while mounting fs with extents feature
3034 Details : Mounting a filesystem with extents feature will fail on big-endian
3035 systems since ext3-based ldiskfs is not supported on big-endian
3036 systems. This can be over-riden with "bigendian_extents" mount option.
3038 Severity : enhancement
3040 Description: The root squash functionality
3041 Details : A security feature, which is to prevent users from being able
3042 to mount lustre on their desktop, run as root, and delete
3043 all of the files in the filesystem. The goal is accomplished by
3044 remapping user id (UID) and group id (GID) of the root user to
3045 a UID and GID specified by the system administartor via Lustre
3046 configuration management server (MGS). The functionality also
3047 allows to specify sets of clients for which the remapping does
3052 Description: Excessive recovery window
3053 Details : With AT enabled, the recovery window can be excessively long (6000+
3054 seconds). To address this problem, we no longer use
3055 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3056 timeout no longer depends on the service time, it is set to
3057 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3058 time via pb_service_time.
3062 Description: Watchdog triggered on MDS failover
3063 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3064 the OSTs know that the MDS "UUID" can be reused for the same export
3065 from a different NID, so we do not need to wait for the export to be
3069 Frequency : rare, only if using MMP with Linux RAID
3071 Description: MMP doesn't work with Linux RAID
3072 Details : While using HA for Lustre servers with Linux RAID, it is possible
3073 that MMP will not detect multiple mounts. To make this work we
3074 need to unplug the device queue in RAID when the MMP block is being
3075 written. Also while reading the MMP block, we should read it from
3076 disk and not the cached one.
3078 Severity : enhancement
3080 Description: open file using fid
3081 Details : A file can be opened using just its fid, like
3082 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3085 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3088 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3089 Details : add proper locking for creation of jbd_4k slab cache
3093 Description: MMP check in ext3_remount() fails without displaying any error
3094 Details : When multiple mount protection fails during remount, proper error
3097 Severity : enhancement
3099 Description: Allow stripe size to be up to 4G-64k
3100 Details : Fix math logic to allow large stripe sizes.
3104 Description: add check for >8TB ldiskfs filesystems
3105 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3106 Don't allow >8TB ldiskfs filesystems to be mounted without
3107 force_over_8tb mount option
3112 Descriptoin: OSS complains "too many blocks" and crashes
3113 Details : In filter_commitrw_write():
3114 1. lquota_chkquota();
3115 2. obdo_from_inode();
3116 3. lquota_pending_commit();
3117 "2." has changed uid and gid in this case, then caused
3118 lquota_chkquota() and lquota_pending_commit() to use
3119 different uid and gid. That is the root of the bug.
3123 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3124 Details : Detect and print out the 'stripe count = -1' case properly.
3128 Description: Out or order replies might be lost on replay
3129 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3130 request with tid smaller than one currently being added, add it
3131 to the start, not end of the list.
3133 --------------------------------------------------------------------------------
3135 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3137 * Support for kernels:
3138 2.6.5-7.283 (SLES 9),
3139 2.6.9-55.EL (RHEL 4),
3140 2.6.16.46-0.14 (SLES 10),
3141 2.6.18.8 vanilla (kernel.org)
3142 * Client support for unpatched kernels:
3143 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3144 2.6.16 - 2.6.22 vanilla (kernel.org)
3145 * Due to recently discovered recovery problems, we do not recommend
3146 using patchless RHEL 4 clients with this or any earlier release.
3147 * Recommended e2fsprogs version: 1.39.cfs8
3148 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3149 * Starting with this release, the ldiskfs backing filesystem required
3150 by Lustre is now in its own package, lustre-ldiskfs. This package
3151 should be installed. It is versioned separately from Lustre and
3152 may be released separately in future.
3156 Description: block reactivating mgc import until all deactivates complete
3157 Details : Fix race when failing back MDT/MGS to itself (testing)
3160 Frequency : at statup only
3162 Description: mds_lov_synchronize race leads to various problems
3163 Details : simultaneous MDT->OST connections at startup can cause the
3164 sync to abort, leaving the OSC in a bad state.
3166 Severity : enhancement
3168 Description: add optional extra BUILD_VERSION info
3169 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3170 to override the lustre version.
3173 Frequency : 2.6.18 servers only
3175 Description: ll_kern_mount() doesn't release the module reference
3176 Details : The ldiskfs module reference count never drops down to 0
3177 because ll_kern_mount() doesn't release the module reference.
3182 Description: server LBUG when using old ost_num_threads parameter
3183 Details : Accept the old ost_num_threads parameter but warn that it
3184 is deprecated, and fix an off-by-one error that caused an LBUG.
3189 Description: Transient SCSI error results in persistent IO issue
3190 Details : iobuf->dr_error is not reinitialized to 0 between two
3194 Frequency : sometimes when underlying device returns I/O errors
3196 Description: OSTs not going read-only during write failures
3197 Details : OSTs are not remounted read-only when the journal commit threads
3198 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3199 instead of the ext3 wrappers.
3203 Description: poor connect scaling with increasing client count
3204 Details : Don't run filter_grant_sanity_check for more than 100 exports
3205 to improve scaling for large numbers of clients.
3208 Frequency : SLES10 only
3210 Description: sanity-quota.sh quotacheck failed: rc = -22
3211 Details : Quotas cannot be enabled on SLES10.
3214 Frequency : liblustre clients only
3216 Description: getdirentries does not give error when run on compute nodes
3217 Details : getdirentries does not fail when the size specified as an argument
3218 is too small to contain at least one entry
3220 Severity : enhancement
3222 Description: Add LNET router traceability for debug purposes
3223 Details : If a checksum failure occurs with a router as part of the
3224 IO path, the NID of the last router that forwarded the bulk data
3225 is printed so it can be identified.
3230 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3231 Details : Due to a race condition, liblustre clients were occasionally
3232 evicted incorrectly.
3234 Severity : enhancement
3236 Description: lfs setstripe use optional parameters instead of postional
3239 Severity : enhancement
3241 Description: Nanosecond timestamp support for ldiskfs
3242 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3243 resolution timestamps. There is not yet support for this at
3244 the Lustre filesystem level.
3247 Frequency : during server recovery
3249 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3250 Details : request with rq_no_resend flag not awake l_wait_event if they get
3254 Frequency : nfs export on patchless client
3256 Description: connectathon hang when test nfs export over patchless client
3257 Details : Disconnected dentry cannot be found with lookup, so we do not need
3258 to unhash it or make it invalid
3261 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3262 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3263 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3268 Description: open req refcounting wrong on reconnect
3269 Details : If reconnect happened between getting open reply from server and
3270 call to mdc_set_replay_data in ll_file_open, we will schedule
3271 replay for unreferenced request that we are about to free.
3272 Subsequent close will crash in variety of ways.
3273 Check that request is still eligible for replay in
3274 mdc_set_replay_data().
3279 Description: disable writes to filesystem when reading health_check file
3280 Details : the default for reading the health_check proc file has changed
3281 to NOT do a journal transaction and write to disk, because this
3282 can cause reads of the /proc file to hang and block HA state
3283 checking on a healthy but otherwise heavily loaded system. It
3284 is possible to return to the previous behaviour during configure
3285 with --enable-health-write.
3287 Severity : enhancement
3289 Description: 64-bit inode version
3290 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3291 to the inode. This will be required for version-based recovery.
3296 Description: MDS fails to start if a duplicate client export is detected
3297 Details : in some rare cases it was possible for a client to connect to
3298 an MDS multiple times. Upon recovery the MDS would detect this
3299 and fail during startup. Handle this more gracefully.
3301 Severity : enhancement
3303 Description: Add -o localflock option to simulate old noflock
3305 Details : This will achieve local-only flock/fcntl locks
3311 Description: log_commit_thread vs filter_destroy race leads to crash
3312 Details : Take import reference before releasing llog record semaphore
3317 Description: Wrong request locking in request set processing
3318 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3319 imp_delayed_list, in this place should be used imp_lock.
3322 Frequency : when reconnection
3324 Description: Grant Leak when osc reconnect to OST
3325 Details : When osc reconnect ost, OST(filter) should check whether it
3326 should grant more space to client by comparing fed_grant and
3327 cl_avail_grant, and return the granted space to client instead
3328 of "new granted" space, because client will call osc_init_grant
3329 to update the client grant space info.
3332 Frequency : when client reconnect to OST
3334 Description: Grant Leak when osc do resend and replay bulk write
3335 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3336 bulk write request, because the grant info while be sync between
3337 OSC and OST when reconnect, and we should ignore the grant info
3338 these of resend/replay write req.
3343 Description: Grant space more than available left space sometimes.
3344 Details : When then OST is about to be full, if two bulk writing from
3345 different clients came to OST. Accord the avaliable space of the
3346 OST, the first req should be permitted, and the second one
3347 should be denied by ENOSPC. But if the seconde arrived before
3348 the first one is committed. The OST might wrongly permit second
3349 writing, which will cause grant space > available space.
3352 Frequency : when client is evicted
3354 Description: Grant might be wrongly erased when osc is evicted by OST
3355 Details : when the import is evicted by server, it will fork another
3356 thread ptlrpc_invalidate_import_thread to invalidate the
3357 import, where the grant will be set to 0. While the original
3358 thread will update the grant it got when connecting. So if
3359 the former happened latter, the grant will be wrongly errased
3360 because of this race.
3365 Description: Checking Stale with correct fid
3366 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3367 to check whether it is stale, because sometimes, we want the
3368 enqueue happened anyway, and op_data.fid2 will not be initialized.
3370 Severity : enhancement
3372 Description: update patchless client
3373 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3376 Frequency : only with 2.4 kernel
3378 Description: random memory corruption
3379 Details : size of struct ll_inode_info is to big for union inode.u and this
3380 can be cause of random memory corruption.
3385 Description: Memory leak in recovery
3386 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3387 It should also check obd_fail before fsfilt_start, otherwise if
3388 fsfilt_start return -EROFS,(failover mds during mds recovery).
3389 then the req will return with repmsg->transno = 0 and rc = EROFS.
3390 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3391 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3392 be freed no matter whether fsfilt_commit success or not.
3395 Frequency : only with huge count clients
3397 Description: Prevents from taking the superblock lock in llap_from_page for
3399 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3400 for llap_from_page call in ll_removepage prevents from taking the
3401 superblock lock for a soon died page.
3406 Description: Not check open intent error before release open handle
3407 Details : in some rare cases, the open intent error is not checked before
3408 release open handle, which may cause
3409 ASSERTION(open_req->rq_transno != 0), because it tries to release
3410 the failed open handle.
3415 Description: Set cat log bitmap only after create log success.
3416 Details : in some rare cases, the cat log bitmap is set too early. and it
3417 should be set only after create log success.
3421 Description: Accessing a block bevice can re-enable I/O when Lustre is
3422 tearing down a device.
3423 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3428 Description: service threads may hog cpus when there are a lot of requests
3429 Details : Insert cond_resched to give other threads a chance to use some CPU
3434 Description: the cat log was not initialized in recovery
3435 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3436 unlink log on mds will not be initialized until mds post
3437 recovery. And also in mds post recovery, the unlink log will
3438 initialization will be done asynchronausly, so there will be race
3439 between add unlink log and unlink log initialization.
3443 Description: brw_stats were being printed incorrectly
3444 Details : brw_stats were being printed as log2 but all of them were not
3445 recorded as log2. Also remove some code duplication arising from
3446 filter_tally_{read,write}.
3450 Frequency : rare, only in recovery.
3451 Description: ASSERTION(req->rq_type != LI_POISON) failed
3452 Details : imp_lock should be held while iterating over imp_sending_list for
3453 prevent destroy request after get timeout in ptlrpc_queue_wait.
3457 Description: replay-single.sh test 52 fails
3458 Details : A lock's skiplist need to be cleanup when it being unlinked
3459 from its resource list.
3463 Description: Short directio read returns full requested size rather than
3465 Details : Direct I/O operations should return actual amount of bytes
3466 transferred rather than requested size.
3468 Severity : enhancement
3470 Description: metadata RPC reduction (e.g. for rm performance)
3471 Details : decrease the amount of synchronous RPC between clients and servers
3472 by canceling conflicing lock before the operation on the client
3473 and packing thier handles into the main operation RPC to server.
3475 Severity : enhancement
3477 Description: Async OSC create to avoid the blocking unnecessarily.
3478 Details : If an OST has no remaining object, system will block on the
3479 creation when it needs to create a new object on this OST. Now,
3480 ways use pre-created objects when available, instead of blocking on
3481 an empty osc while others are not empty. If we must block, we
3482 block for the shortest possible period of time.
3486 Description: improve handling recoverable errors
3487 Details : if request processig with error which can be recoverable on server
3488 request should be resend, otherwise page released from cache and
3491 Severity : enhancement
3493 Description: refine locking for avoid write wrong info into lov_objid file
3494 Details : fix possible races with add new target and write/update data in
3497 --------------------------------------------------------------------------------
3499 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3504 Frequency : on some architectures
3506 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3507 Details : Uninitialized flags sometimes cause configuration commands to
3511 Frequency : patchless clients only
3513 Description: missing __iget() symbol export
3514 Details : The __iget() symbol export is missing. To avoid the need for
3515 this on patchless clients the deathrow inode reaper is turned
3516 off, and we depend on the VM to clean up old inodes. This
3517 dependency was during via the fix for bug 12181.
3519 --------------------------------------------------------------------------------
3521 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3523 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3524 INTEROPERATE with older versions automatically. In many cases a
3525 special upgrade step is needed. Please read the
3526 user documentation before upgrading any part of a live system.
3527 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3528 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3529 * WARNING: Lustre configuration and startup changes are required with
3530 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3534 Severity : enhancement
3536 Description: MountConf
3537 Details : Lustre configuration is now managed via mkfs and mount
3538 commands instead of lmc and lconf. New obd types (MGS, MGC)
3539 are added for dynamic configuration management. See
3540 https://mail.clusterfs.com/wikis/lustre/MountConf for
3543 Severity : enhancement
3545 Description: dynamic OST addition
3546 Details : OSTs can now be added to a live filesystem
3548 Severity : enhancement
3550 Description: startup order invariance
3551 Details : MDTs and OSTs can be started in any order. Clients only
3552 require the MDT to complete startup.
3554 Severity : enhancement
3556 Description: parallel, asynchronous orphan cleanup
3557 Details : orphan cleanup is now performed in separate threads for each
3558 OST, allowing parallel non-blocking operation.
3560 Severity : enhancement
3562 Description: optimized stripe assignment
3563 Details : stripe assignments are now made based on ost space available,
3564 ost previous usage, and OSS previous usage, in order to try
3565 to optimize storage space and networking resources.
3567 Severity : enhancement
3569 Description: Permanently set tunables
3570 Details : All writable /proc/fs/lustre tunables can now be permanently
3571 set on a per-server basis, at mkfs time or on a live system.
3573 Severity : enhancement
3575 Description: Lustre message v2
3576 Details : Add lustre message format v2.
3578 Severity : enhancement
3580 Description: client OST exclusion list
3581 Details : Clients can be started with a list of OSTs that should be
3582 declared "inactive" for known non-responsive OSTs.
3586 Description: SPEC SFS validation failure on NFS v2 over lustre.
3587 Details : Changes the blocksize for regular files to be 2x RPC size,
3588 and not depend on stripe size.
3590 Severity : enhancement
3592 Description: Multiple MD RPCs in flight.
3593 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3594 To avoid overly-overloading MDS, introduce a limit on number of
3595 MDS RPCs in flight for a single client and add /proc controls
3596 to adjust this limit.
3598 Severity : enhancement
3600 Description: client read/write statistics
3601 Details : Add client read/write call usage stats for performance
3602 analysis of user processes.
3603 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3604 file access. extents_stats shows chunk size distribution.
3605 extents_stats_per_process show chunk size distribution per
3608 Severity : enhancement
3610 Description: mds statistics
3611 Details : Add detailed mds operations statistics in
3612 /proc/fs/lustre/mds/*/stats.
3616 Description: Failure of copying files with lustre special EAs.
3617 Details : Client side always return success for setxattr call for lustre
3618 special xattr (currently only "trusted.lov").
3623 Description: Refcount LNET uuids
3624 Details : The global LNET uuid list grew linearly with every startup;
3625 refcount repeated list entries instead of always adding to
3628 Severity : enhancement
3630 Description: Dynamic service threads
3631 Details : Within a small range, start extra service threads
3632 automatically when the request queue builds up.
3635 Frequency : mixed-endian client/server environments
3637 Description: mixed-endian crashes
3638 Details : The new msg_v2 system had some failures in mixed-endian
3641 Severity : enhancement
3643 Description: Easy OST removal
3644 Details : OSTs can be permanently deactivated with e.g. 'lctl
3645 conf_param lustre-OST0001.osc.active=0'
3647 Severity : enhancement
3649 Description: MGS proc entries
3650 Details : Added basic proc entries for the MGS showing what filesystems
3653 Severity : enhancement
3655 Description: provide MGS failover
3656 Details : Added config lock reacquisition after MGS server failover.
3658 Severity : enhancement
3660 Description: add Linux 2.4 support
3661 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3665 Description: a large application tries to do I/O to the same resource and dies
3666 in the middle of it.
3667 Details : Check the req->rq_arrival time after the call to
3668 ost_brw_lock_get(), but before we do anything about
3669 processing it & sending the BULK transfer request. This
3670 should help move old stale pending locks off the queue as
3671 quickly as obd_timeout.
3674 Frequency : when an incorrect nid is specified during startup
3676 Description: ptlrpc connect to non-existant node causes kernel crash
3677 Details : LNET can't be re-entered from an event callback, which
3678 happened when we expire a message after the export has been
3679 cleaned up. Instead, hand the zombie cleanup off to another
3682 Severity : enhancement
3684 Description: plain/inodebits lock performance improvement
3685 Details : Grouping plain/inodebits in granted list by their request modes
3686 and bits policy, thus improving the performance of search through
3690 Frequency : only if OST filesystem is corrupted
3692 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3693 Details : for a short time RPCs with bulk IO are in the replay list,
3694 but replay of bulk IOs is unimplemented. If the OST filesystem
3695 is corrupted due to disk cache incoherency and then replay is
3696 started it is possible to trip an assertion. Avoid putting
3697 committed RPCs into the replay list at all to avoid this issue.
3700 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3702 Description: System hang on startup
3703 Details : This bug allowed the liblustre (e.g. catamount) client to
3704 return to the app before handling all startup RPCs. This
3705 could leave the node unresponsive to lustre network traffic
3706 and manifested as a server ptllnd timeout.
3708 Severity : enhancement
3710 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3711 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3712 boolean to control whether to print peer debug info when a
3713 client's RPC times out.
3716 Frequency : only for kernels with patches from Lustre below 1.4.3
3718 Description: Remove old rdonly API
3719 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3722 Frequency : only for devices with external journals
3724 Description: Set external device read-only also
3725 Details : During a commanded failover stop, we set the disk device
3726 read-only while the server shuts down. We now also set any
3727 external journal device read-only at the same time.
3730 Frequency : when upgrading from 1.4 while trying to change parameters
3732 Description: The wrong (new) MDC name was used when setting parameters for
3733 upgraded MDT's. Also allows changing of OSC (and MDC)
3734 parameters if --writeconf is specified at tunefs upgrade time.
3737 Frequency : when setting specific ost indicies
3739 Description: QOS code breaks on skipped indicies
3740 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3741 created with --index need not be sequential.
3743 Severity : enhancement
3745 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3746 Details : The uninit_groups feature works in conjunction with the kernel
3747 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3748 up the pass1 processing of e2fsck. This is a read-only feature
3749 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3750 mount filesystems that have had this feature enabled.
3752 Severity : enhancement
3754 Description: Improve multi-block allocation algorithm to avoid fragmentation
3755 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3756 allocation locality and avoid filesystem fragmentation.
3758 ------------------------------------------------------------------------------
3760 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3762 * Support for kernels:
3763 2.4.21-47.0.1.EL (RHEL 3)
3764 2.6.5-7.283 (SLES 9)
3765 2.6.9-42.0.10.EL (RHEL 4)
3766 2.6.12.6 vanilla (kernel.org)
3767 2.6.16.27-0.9 (SLES 10)
3768 * Recommended e2fsprogs version: 1.39.cfs5
3770 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3774 Frequency : occasional, depends on client load and configuration
3775 Bugzilla : 12181, 12203
3776 Description: data loss for recently-modified files
3778 Details : In some cases it is possible that recently written or created
3779 files may not be written to disk in a timely manner (this should
3780 normally be within 30s unless client IO load is very high).
3781 The problem appears as zero-length files or files that are a
3782 multiple of 1MB in size after a client crash or client eviction
3783 that are missing data at the end of the file.
3785 This problem is more likely to be hit on clients where files are
3786 repeatedly created and unlinked in the same directory, clients
3787 have a large amount of RAM, have many CPUs, the filesystem has
3788 many OSTs, the clients are rebooted frequently, and/or the files
3789 are not accessed by other nodes after being written.
3791 The presence of the problem can be detected by looking at
3792 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3793 smaller than the second (nr_unused) then dirty files will not
3794 be flushed automatically to disk. "sync; sleep 10" should be
3795 run several times on the node before unmounting it to update
3796 Lustre (this is also safe to run on nodes without this problem).
3798 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3799 that can cause this same problem, so customers using that kernel
3800 also need to update the kernel in addition to Lustre. In order
3801 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3803 It is normal that files written just before a client crash (less
3804 than 30s) may not yet have been flushed to disk, even for local
3808 Frequency : frequent on thin XT3 nodes
3810 Description: UUID collision on thin XT3 Linux nodes
3811 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3812 randomly, since we relied on an insufficiently-seeded PRNG.
3817 Description: OSS hangs after "All ost request buffers busy"
3818 Details : A deadlock between quota and journal operations caused OSS
3819 hangs after printing "All ost request buffers busy."
3822 Frequency : always on liblustre builds
3824 Description: Cleanup compiler warnings on liblustre
3827 Frequency : always on liblustre builds on XT3
3829 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3833 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3834 > 32000 subdirectories
3835 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3836 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3837 subdirectory count crosses 32000. This will aid e2fsck to
3838 correctly handle more than 32000 subdirectories.
3841 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3843 Description: System hang on startup
3844 Details : This bug allowed the liblustre (e.g. catamount) client to
3845 return to the app before handling all startup RPCs. This
3846 could leave the node unresponsive to lustre network traffic
3847 and manifested as a server ptllnd timeout.
3849 Severity : enhancement
3851 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3852 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3853 boolean to control whether to print peer debug info when a
3854 client's RPC times out.
3859 Description: make O_SYNC working on 2.6 kernels
3860 Details : 2.6 kernels use different method for mark pages for write,
3861 so need add a code to lustre for O_SYNC work.
3866 Description: Failure to close file and release space on NFS
3867 Details : Put inode details into lock acquired in ll_intent_file_open.
3868 Use mdc_intent_lock in ll_intent_open to properly
3869 detect all kind of errors unhandled by mdc_enqueue.
3874 Description: proc file read during shutdown sometimes raced obd removal,
3876 Details : Add lock to prevent obd access after proc file removal.
3879 Frequency : Only for files larger than 4GB on 32-bit clients.
3881 Description: improperly doing page alignment of locks
3882 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3883 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3888 Description: RPCs being resent when they shouldn't be
3889 Details : Some RPCs that should not be resent are being resent. This
3890 can cause inconsistencies in the RPC state machine. Do not
3891 resend such requests.
3894 Frequency : rare, only with NFS export
3896 Description: Crash on NFS re-export node
3897 Details : under very unusual load conditions an assertion is hit in
3898 ll_intent_file_open()
3901 Frequency : only if OST filesystem is corrupted
3903 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3904 Details : for a short time RPCs with bulk IO are in the replay list,
3905 but replay of bulk IOs is unimplemented. If the OST filesystem
3906 is corrupted due to disk cache incoherency and then replay is
3907 started it is possible to trip an assertion. Avoid putting
3908 committed RPCs into the replay list at all to avoid this issue.
3913 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3914 Details : Large single O_DIRECT read and write calls can fail to allocate
3915 a sufficiently large buffer to process the request. In case of
3916 allocation failure the allocation is retried with a smaller
3917 buffer and broken into smaller requests.
3919 Severity : enhancement
3921 Description: Add -o localflock option to simulate old noflock behaviour.
3922 Details : This will achieve local-only flock/fcntl locks coherentness.
3927 Description: versioning check is incomplete
3928 Details : Checking the version difference of client vs. server, report
3929 error if the gap is too big.
3934 Description: add support PG_writeback bit
3935 Details : add support for PG_writeback bit for Lustre, for more carefull
3936 work with page cache in 2.6 kernel. This also fix some deadlocks
3937 and remove hack for work O_SYNC with 2.6 kernel.
3939 Severity : enhancement
3941 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3942 Details : The uninit_groups feature works in conjunction with the kernel
3943 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3944 up the pass1 processing of e2fsck. This is a read-only feature
3945 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3946 mount filesystems that have had this feature enabled.
3948 Severity : enhancement
3950 Description: Improve multi-block allocation algorithm to avoid fragmentation
3951 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3952 allocation locality and avoid filesystem fragmentation.
3954 ------------------------------------------------------------------------------
3956 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3958 * Support for kernels:
3959 2.6.9-42.0.3.EL (RHEL 4)
3960 2.6.5-7.276 (SLES 9)
3961 2.4.21-47.0.1.EL (RHEL 3)
3962 2.6.12.6 vanilla (kernel.org)
3963 2.6.16.21-0.8 (SLES10)
3964 * Recommended e2fsprogs version: 1.39.cfs2-0
3966 * The backwards-compatible /proc/sys/portals symlink has been removed
3967 in this release. Before upgrading, please ensure that you change
3968 any configuration scripts or /etc/sysctl.conf files that access
3969 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3970 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3971 in advance of the upgrade on any system running Lustre 1.4.6 or
3972 newer, since /proc/sys/lnet was added in that version.
3973 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3977 Frequency : only when quota is used
3979 Description: avoid scanning export list for quota master
3980 Details : Change the algorithms to avoid scanning export list in order
3981 to improve the efficiency.
3984 Frequency : MDS failover only, very rarely
3986 Description: "went back in time" messages on mds failover
3987 Details : The greatest transno may be lost when the current operation
3988 finishes with an error (transno==0) and the client's last_rcvd
3989 record is over-written. Save the greatest transno in the
3990 mds_last_transno for this case.
3993 Frequency : always for specific kernels and striping counts
3995 Description: client may get "Matching packet too big" without ACL support
3996 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3997 when trying to access files in certain configurations. The
3998 clients should in fact be denied when mounting because they do
3999 not understand ACLs.
4002 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4004 Description: many clients connecting with IO in progress causes connect timeouts
4005 Details : Avoid synchronous journal commits to avoid delays caused by many
4006 clients connecting/disconnecting when bulk IO is in progress.
4007 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4008 OST_IO_PORTAL to avoid delays behind potentially many pending
4012 Frequency : occasionally with multiple writers to a single file
4014 Description: shared writes to file may result in wrong size reported by stat()
4015 Details : Allow growing of kms when extent lock is cancelled
4018 Frequency : always with random mmap IO to multi-striped file
4020 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4021 Details : Only if the hole is at the end of OST object so that kms is too
4022 small. Fix is to increase kms accordingly in ll_nopage.
4025 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4027 Description: writes to a missing object would leak memory on the OST
4028 Details : If there is an inconsistency between the MDS and OST filesystems,
4029 such that the MDS references an object that doesn't exist, writes
4030 to that object will leak memory due to incorrect cleanup in the
4031 error handling path, eventually running out of memory on the OST.
4036 Description: Creating too long symlink causes lustre errors
4037 Details : Check symlink and name lengths before sending requests to MDS.
4040 Frequency : only if flock is enabled (not on by default)
4042 Description: posix locks not released on fd closure on 2.6.9+
4043 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4044 kernels, this caused such locks not to be released on fd close and
4045 then assertions on fs unmount about still used locks.
4048 Frequency : MDS failover only, very rarely
4050 Description: clients may get ASSERTION(granted_lock != NULL)
4051 Details : When request was taking a long time, and a client was resending
4052 a getattr by name lock request. The were multiple lock requests
4053 with the same client lock handle and
4054 mds_getattr_name->fixup_handle_for_resent_request found one of the
4055 lock handles but later failed with ASSERTION(granted_lock != NULL).
4060 Description: handle->h_buffer_credits > 0, assertion failure
4061 Details : h_buffer_credits is zero after truncate, causing assertion
4062 failure. This patch extends the transaction or creates a new
4066 Frequency : NFS re-export or patchless client
4067 Bugzilla : 11179, 10796
4068 Description: Crash on NFS re-export node (__d_move)
4069 Details : We do not want to hash the dentry if we don't have a lock.
4070 But if this dentry is later used in d_move, we'd hit uninitialised
4071 list head d_hash, so we just do this to init d_hash field but
4072 leave dentry unhashed.
4075 Frequency : NFS re-export or patchless client
4077 Description: NFS exports has problem with symbolic link
4078 Details : lustre client didn't properly install dentry when re-exported
4079 to NFS or running patchless client.
4082 Frequency : NFS re-export or patchless client
4084 Description: Various nfs/patchless fixes.
4085 Details : fixes reuse disconected alias for lookup process - this fixes
4086 warning "find_exported_dentry: npd != pd",
4087 fix permission error with open files at nfs.
4088 fix apply umask when do revalidate.
4091 Frequency : occasional
4093 Description: Crash on NFS re-export node
4094 Details : calling clear_page() on the wrong pointer triggered oops in
4095 generic_mapping_read().
4098 Frequency : rarely, using O_DIRECT IO
4100 Description: unaligned directio crashes client with LASSERT
4101 Details : check for unaligned buffers before trying any requests.
4104 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4106 Description: stale data returned from RAID cache
4107 Details : If only a small amount of IO is done to the RAID device before
4108 reading it again it is possible to get stale data from the RAID
4109 cache instead of reading it from disk.
4112 Frequency : always for sles10 kernel
4114 Description: sles10 support
4115 Details : ll_follow_link: compile fixes and using of nd_set_link
4116 under newer kernels.
4119 Frequency : depends on arch, kernel and compiler version, always on sles10
4122 Description: recursive or deep enough symlinks cause stack overflow
4123 Details : getting rid of large stack-allocated variable in
4127 Frequency : depends on hardware
4129 Description: lustre write performance loss in the SLES10 kernel
4130 Details : the performance loss is caused by using of write barriers in the
4131 ext3 code. The SLES10 kernel turns barrier support on by
4132 default. The fix is to undo that change for ldiskfs.
4134 ------------------------------------------------------------------------------
4136 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4138 * Support for kernels:
4139 2.6.9-42.0.3EL (RHEL 4)
4140 2.6.5-7.276 (SLES 9)
4141 2.4.21-47.0.1.EL (RHEL 3)
4142 2.6.12.6 vanilla (kernel.org)
4146 Frequency : quota enabled and large files being deleted
4148 Description: releasing more than 4GB of quota at once hangs OST
4149 Details : If a user deletes more than 4GB of files on a single OST it
4150 will cause the OST to spin in an infinite loop. Release
4151 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4156 Description: statfs data retrieved from /proc may be stale or zero
4157 Details : When reading per-device statfs data from /proc, in the
4158 {kbytes,files}_{total,free,avail} files, it may appear
4159 as zero or be out of date.
4162 Frequency : systems with MD RAID1 external journal devices
4164 Description: lconf's call to blkid is confused by RAID1 journal devices
4165 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4166 of returning all block devices that match the journal UUID.
4169 Frequency : always, for aggregate stripe size over 4GB
4171 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4172 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4173 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4176 Frequency : always if "lfs find" used on a local file/directory
4178 Description: "lfs find" segfaults if used on a local file/directory
4179 Details : The case where a directory component was not specified wasn't
4180 handled correctly. Handle this properly.
4183 Frequency : always on ppc64
4185 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4186 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4187 perform endianness conversion.
4190 Frequency : rarely (truncate to non-zero file size after write under load)
4191 Bugzilla : 10730, 10687
4192 Description: Files padded with zeros to next 4K multiple
4193 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4194 it is possible that files that are truncated to a non-zero size
4195 immediately after being written are filled with zero bytes beyond
4196 the truncated size. No file data is lost.
4198 Severity : enhancement
4200 Description: Allow recovery/failover for liblustre clients.
4201 Details : liblustre clients were unaware of failover configurations until
4204 Severity : enhancement
4206 Description: user file locks should fail when not mounting with flock option
4207 Details : Set up an error-returning stub in ll_file_operations.lock field
4208 to prevent incorrect behaviour when client is mounted without
4209 flock option. Also, set up properly f_op->flock field for
4213 Frequency : always on ia64
4215 Description: "lfs df" loops on printing out MDS statfs information
4216 Details : The obd_ioctl_data was not initialized and in some systems
4217 this caused a failure during the ioctl that did not return
4218 an error. Initialize the struct and return an error on failure.
4221 Frequency : SLES 9 only
4223 Description: Error of copying files with lustre special EAs as root
4224 Details : Client side always return success for setxattr call for lustre
4225 special xattr (currently only "trusted.lov").
4228 Frequency : rarely on clusters with both ia64+i386 clients
4230 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4231 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4232 clients are doing concurrent writes to the same file it is
4233 possible that the ia64 clients may LASSERT because the OST
4234 extent locks are not PAGE_SIZE aligned. Ensure that grown
4235 locks are always aligned on the request boundary.
4238 Frequency : specific use, occasional
4240 Description: Overwriting in use executable truncates on-disk binary image
4241 Details : If one node attempts to overwrite an executable in use by
4242 another node, we now correctly return ETXTBSY instead of
4243 truncating the file.
4245 Severity : enhancement
4247 Description: Async OSC create to avoid the blocking unnecessarily.
4248 Details : If an OST has no remaining object, system will block on the creating
4249 when need to create a new object on this OST. Now, ways use
4250 pre-created objects when available, instead of blocking on an
4251 empty osc while others are not empty. If we must block, we block
4252 for the shortest possible period of time.
4257 Description: chmod on Lustre root is propagated to other clients
4258 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4259 invalid by the follow_mount time.
4262 Frequency : liblustre clients only
4264 Description: Race in 'instant cancel' lock handling could lead to such locks
4265 never to be granted in case of SMP MDS
4266 Details : Do not destroy not yet granted but cbpending locks in
4270 Frequency : replay/resend of open
4272 Description: non null lock assetion failure in mds_intent_policy
4273 Details : Trying to replay/resend lockless open requests resulted in
4274 mds_open() returning 0 with no lock. Now it sets a flag if
4275 a lock is going to be returned.
4277 Severity : enhancement
4279 Description: Checksum enhancements
4280 Details : New checksum enhancements allow for resending RPCs that failed
4283 Severity : enhancement
4285 Description: Tunables on number of dirty pages in cacche
4286 Details : Allow to set limit on number of dirty pages cached.
4291 Description: client crash on unmount - lock still has references
4292 Details : In some error handling cases it was possible to leak a lock
4293 reference on a client while accessing a file. This was not
4294 harmful to the client during operation, but would cause the
4295 client to crash when the filesystem is unmounted.
4298 Frequency : specific case, rare
4300 Description: ETXTBSY on mds though file not in use by client
4301 Details : ETXTBSY is no longer incorrectly returned when attempting to
4302 chmod or chown a directory that the user previously tried to
4303 execute or a currently-executing binary.
4306 Frequency : extremely rare except on liblustre-based clients
4308 Description: Lustre space not freed when files are deleted
4309 Details : Clean up open-unlinked files after client eviction. Previously
4310 the unlink was skipped and the files remained as orphans.
4315 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4316 Details : In some cases it was possible to send a blocking callback to
4317 a client doing a glimpse, even though that client didn't get
4318 a lock granted. When the glimpse lock is cancelled on the OST
4319 the freed lock is left on the waiting list and corrupted the list.
4322 Frequency : all core dumps
4324 Description: Broke core dumps to lustre
4325 Details : Negative dentry may be unhashed if parent does not have UPDATE
4326 lock, but some callers, e.g. do_coredump, expect dentry to be
4327 hashed after successful create, hash it in ll_create_it.
4329 ------------------------------------------------------------------------------
4331 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4333 * Support for kernels:
4334 2.6.9-42.0.2.EL (RHEL 4)
4335 2.6.5-7.276 (SLES 9)
4336 2.4.21-40.EL (RHEL 3)
4337 2.6.12.6 vanilla (kernel.org)
4341 Frequency : always on RHEL 3
4343 Description: Number of open files grows over time
4344 Details : The number of open files grows over time, whether or not
4345 Lustre is started. This was due to a filp leak introduced
4346 by one of our kernel patches.
4348 ------------------------------------------------------------------------------
4350 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4352 * Support for kernels:
4353 2.6.9-42.EL (RHEL 4)
4354 2.6.5-7.276 (SLES 9)
4355 2.4.21-40.EL (RHEL 3)
4356 2.6.12.6 vanilla (kernel.org)
4361 Bugzilla : 5719, 9635, 9792, 9684
4362 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4363 Details : If a server is under heavy load and cannot reply to new
4364 connection requests before the client resends the (re)connect,
4365 the connection handling code can behave badly if two service
4366 threads are concurrently handing separate (re)connections from
4367 the same client. Add better locking to the connection handling
4368 code, and ensure that only a single connection will be processed
4369 for a given client UUID, even if the lock is dropped.
4371 Severity : enhancement
4373 Description: add TCP zero-copy support to kernel
4374 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4375 sends if the hardware supports scatter-gather and checksumming.
4376 This allows socklnd to do client-write and server-read more
4377 efficiently and reduce CPU utilization from skbuf copying.
4380 Frequency : only if NFS exporting from client
4382 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4383 Details : If setting attributes on a file created under NFS that had
4384 never been opened it would be possible to oops the client
4385 if the file had no objects.
4388 Frequency : always for liblustre
4390 Description: liblustre client does MDS+OSTs setattr RPC for each write
4391 Details : When doing a write from a liblustre client, the client
4392 incorrectly issued an RPC to the MDS and each OST the file was
4393 striped over in order to update the timestamps. When writing
4394 with small chunks and many clients this could overwhelm the MDS
4395 with RPCs. In all cases it would slow down the write because
4396 these RPCs are unnecessary.
4398 Severity : enhancement
4400 Description: allow number of MDS service threads to be changed at module load
4401 Details : It is now possible to change the number of MDS service threads
4402 running. Adding "options mds mds_num_threads={N}" to the MDS's
4403 /etc/modprobe.conf will set the number of threads for the next
4404 time Lustre is restarted (assuming the "mds" module is also
4405 reloaded at that time). The default number of threads will
4406 stay the same, 32 for most systems.
4411 Description: OST crash if filesystem is unformatted or corrupt
4412 Details : If an OST is started on a device that has never been formatted
4413 or if the filesystem is corrupt and cannot even mount then the
4414 error handling cleanup routines would dereference a NULL pointer.
4419 Description: NULL pointer deref in llap_from_page.
4420 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4421 incorrect) mapping if the page was truncated/reclaimed while it was
4422 searched for. Check for this condition and skip such pages when
4423 doing readahead. Introduce extra check to llap_from_page() to
4424 verify page->mapping->host is non-NULL (so page is not anonymous).
4427 Frequency : Sometimes when using sys_sendfile
4429 Description: "page not covered by a lock" warnings from ll_readpage
4430 Details : sendfile called ll_readpage without right page locks present.
4431 Now we introduced ll_file_sendfile that does necessary locking
4432 around call to generic_file_sendfile() much like we do in
4436 Frequency : with certain MDS communication failures at client mount time
4438 Description: NULL pointer deref after failed client mount
4439 Details : a client connection request may delayed by the network layer
4440 and not be sent until after the PTLRPC layer has timed out the
4441 request. If the client fails the mount immediately it will try
4442 to clean up before the network times out the request. Add a
4443 reference from the request import to the obd device and delay
4444 the cleanup until the network drops the request.
4447 Frequency : occasionally during client (re)connect
4449 Description: assertion failure during client (re)connect
4450 Details : processing a client connection request may be delayed by the
4451 client or server longer than the client connect timeout. This
4452 causes the client to resend the connection request. If the
4453 original connection request is replied in this interval, the
4454 client may trip an assertion failure in ptlrpc_connect_interpret()
4455 which thought it would be the only running connect process.
4458 Frequency : only with obd_echo servers and clients that are rebooted
4460 Description: kernel BUG accessing uninitialized data structure
4461 Details : When running an obd_echo server it did not start the ping_evictor
4462 thread, and when a client was evicted an uninitialized data
4463 structure was accessed. Start the ping_evictor in the RPC
4464 service startup instead of the OBD startup.
4466 Severity : enhancement
4467 Bugzilla : 10193 (patchless)
4468 Description: Remove dependency on various unexported kernel interfaces.
4469 Details : No longer need reparent_to_init, exit_mm, exit_files,
4470 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4473 Frequency : rare (only users of deprecated and unsupported LDAP config)
4475 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4476 Details : LDAP apparently contains 'lustreName' attributes instead of
4477 'name'. A simple remapping of the name is sufficient.
4480 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4482 Description: waiting_locks_callback trips kernel BUG if client is evicted
4483 Details : Running with the dump_on_timeout debug flag turned on makes
4484 it possible that the waiting_locks_callback() can try to dump
4485 the Lustre kernel debug logs from an interrupt handler. Defer
4486 this log dumping to the expired_lock_main() thread.
4488 Severity : enhancement
4490 Description: Support NFS exporting on 2.6 kernels.
4491 Details : Implement non-rawops metadata methods for NFS server to use without
4492 changing NFS server code.
4495 Frequency : very rare (synthetic metadata workload only)
4497 Description: two racing renames might cause an MDS thread to deadlock
4498 Details : Running the "racer" program may cause one MDS thread to rename
4499 a file from being the source of a rename to being the target of
4500 a rename at exactly the same time that another thread is doing
4501 so, and the second thread has already enqueued these locks after
4502 doing a lookup of the target and is trying to relock them in
4503 order. Ensure that we don't try to re-lock the same resource.
4506 Frequency : only very large systems with liblustre clients
4508 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4509 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4510 all OSTs in parallel. This allows the eviction of stale liblustre
4511 clients to proceed much faster than if they were done in series,
4512 and also offers similar improvements for other set_info RPCs.
4517 Description: excessive CPU usage during initial read phase on client
4518 Details : During the initial read phase on a client, it would agressively
4519 retry readahead on the file, consuming too much CPU and impacting
4520 performance (since 1.4.5.8). Improve the readahead algorithm
4521 to avoid this, and also improve some other common cases (read
4522 of small files in particular, where "small" is files smaller than
4523 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4528 Description: MDS crash when receiving packet with unknown intent.
4529 Details : Do not LBUG in unknown intent case, just return -EFAULT
4531 Severity : enhancement
4532 Bugzilla : 9293, 9385
4533 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4534 Details : Do not serialize getattr (non-intent version) and statfs.
4537 Frequency : occasional, when OST network is overloaded/intermittent
4539 Description: client evicted by OST after bulk IO timeout
4540 Details : If a client sends a bulk IO request (read or write) the OST
4541 may evict the client if it is unresposive to its data GET/PUT
4542 request. This is incorrect if the network is overloaded (takes
4543 too long to transfer the RPC data) or dropped the OST GET/PUT
4544 request. There is no need to evict the client at all, since
4545 the pinger and/or lock callbacks will handle this, and the
4546 client can restart the bulk request.
4549 Frequency : Always when mmapping file with no objects
4551 Description: client crashes when mmapping file with no objects
4552 Details : Check that we actually have objects in a file before doing any
4553 operations on objects in ll_vm_open, ll_vm_close and
4559 Description: Request leak when working with deleted CWD
4560 Details : Introduce advanced request refcount tracking for requests
4561 referenced from lustre intent.
4563 Severity : Enhancement
4565 Description: Cache open file handles on client.
4566 Details : MDS now will return special lock along with openhandle, if
4567 requested and client is allowed to hold openhandle, even if unused,
4568 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4569 closing files for every read/write openration.
4571 Severity : Enhancement
4573 Description: Cache open negative dentries on client when possible.
4574 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4575 negative dentries on lock revocation.
4580 Description: Remounting a client read-only wasn't possible with a zconf mount
4581 Details : It wasn't possible to remount a client read-only with llmount.
4583 Severity : enhancement
4584 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4585 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4586 MPICH is included to add Lustre-specific ADIO interfaces.
4587 This is based closely on the UFS ADIO layer and only differs
4588 in file creation, in order to allow the OST striping to be set.
4589 This is user-contributed code and not supported by CFS.
4594 Description: extended inode attributes (immutable, append-only) work improperly
4595 when 2.4 and 2.6 kernels are used on client/server or vice versa
4596 Details : Introduce kernel-independent values for these flags.
4598 Severity : enhancement
4601 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4602 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4603 granularity. Now, Fractional MB granularity is be supported,
4604 this is very useful for low memory system.
4606 Severity : enhancement
4608 Description: Getattr by fid
4609 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4610 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4611 have locks and avoids getattr rpc requests for it on every stat.
4614 Frequency : Always, for filesystems larger than 2TB
4616 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4617 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4618 filesystems larger than 2TB to be mounted with mballoc enabled.
4621 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4623 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4624 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4625 filesystems larger than 2TB is mounted then the kernel will
4626 silently corrupt the start of the filesystem. CONFIG_LBD is
4627 enabled for all CFS-supported kernels, but the possibility of
4628 this happening with a modified kernel config exists.
4630 Severity : enhancement
4632 Description: add client O_DIRECT support for 2.6 kernels
4633 Details : It is now possible to do O_DIRECT reads and writes to files
4634 in the Lustre client mountpoint on 2.6 kernel clients.
4636 Severity : enhancement
4638 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4639 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4640 parallel, not waiting for response from every OST before sending
4641 a rpc to the next OST.
4646 Description: setattr vs write race when updating file timestamps
4647 Details : Client processes that update a file timestamp into the past
4648 right after writing to the file (e.g. tar) it is possible that
4649 the updated file modification time can be reset to the current
4650 time due to a race between processing the setattr and write RPC.
4652 Severity : enhancement
4654 Description: Bring 'lfs find' closer in line with regular Linux find.
4655 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4656 -print0 options and obtains all the needed info through the lustre
4659 Severity : enhancement
4661 Description: support up to 1024 configured devices on one node
4662 Details : change obd_dev array from statically allocated to dynamically
4663 allocated structs as they are first used to reduce memory usage
4668 Description: Flush dirty partially truncated pages during truncate
4669 Details : Immediatelly flush partially truncated pages in filter_setattr,
4670 this way we completely avoid having any pages in page cache on OST
4671 and can retire ugly workarounds during writes to flush such pages.
4676 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4677 Details : protect inode from truncation within vfs_unlink() context
4678 just take a reference before calling vfs_unlink() and release it
4679 when parent's i_sem is free.
4684 Description: last_id value checked outside lock on OST caused LASSERT failure
4685 Details : If there were multiple MDS->OST object precreate requests in
4686 flight, it was possible that the OST's last object id was checked
4687 outside a lock and incorrectly tripped an assertion. Move checks
4688 inside locks, and discard old precreate requests.
4691 Frequency : always, if extents are used on OSTs
4693 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4694 Details : OSTs using the extents format would not zero the high 16 bits of
4695 the index physical block number. This is not a problem for any
4696 OST filesystems smaller than 16TB, and no kernels support ext3
4697 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4698 new/modified files) and can be fixed for existing filesystems
4699 with e2fsprogs-1.39-cfs1.
4704 Description: import connection selection may be incorrect if timer wraps
4705 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4706 import connections to be ignored if the 32-bit jiffies counter
4707 wraps. Use a 64-bit jiffies counter.
4710 Frequency : very large clusters immediately after boot
4712 Description: LNET request buffers exhausted under heavy short-term load
4713 Details : If a large number of client requests are generated on a service
4714 that has previously never seen so many requests it is possible
4715 that the request buffer growth cannot keep up with the spike in
4716 demand. Instead of dropping incoming requests, they are held in
4717 the LND until the RPC service can accept more requests.
4720 Frequency : Sometimes during replay
4722 Description: Assertion failure in ll_local_open after replay.
4723 Details : If replay happened on an open request reply before we were able
4724 to set replay handler, reply will become not swabbed tripping the
4725 assertion in ll_local_open. Now we set the handler right after
4726 recognising of open request
4729 Frequency : very rare
4731 Description: kernel reports "badness in vsnprintf"
4732 Details : Reading from the "recovery_status" /proc file in small chunks
4733 may cause a negative length in lprocfs_obd_rd_recovery_status()
4734 call to vsnprintf() (which is otherwise harmless). Exit early
4735 if there is no more space in the output buffer.
4737 Severity : enhancement
4739 Description: clear OBD RPC statistics by writing to them
4740 Details : It is now possible to clear the OBD RPC statistics by writing
4741 to the "stats" file.
4746 Description: Client mtime is not the same on different clients after utimes
4747 Details : In some cases, the client was using the utimes() syscall on
4748 a file cached on another node. The clients now validate the
4749 ctime from the MDS + OSTs to determine which one is right.
4754 Description: Inability to activate failout mode
4755 Details : lconf script incorrectly assumed that in pythong string's numeric
4756 value is used in comparisons.
4759 Frequency : always with multiple stripes per file
4761 Description: Inefficient object allocation for mutli-stripe files
4762 Details : When selecting which OSTs to stripe files over, for files with
4763 a stripe count that divides evenly into the number of OSTs,
4764 the MDS is always picking the same starting OST for each file.
4765 Return the OST selection heuristic to the original design.
4770 Description: mount failures may take full timeout to return an error
4771 Details : Under some heavy load conditions it is possible that a
4772 failed mount can wait for the full obd_timeout interval,
4773 possibly several minutes, before reporting an error.
4774 Instead return an error as soon as the status is known.
4776 Frequency : quota enabled and large files being deleted
4778 Description: releasing more than 4GB of quota at once hangs OST
4779 Details : If a user deletes more than 4GB of files on a single OST it
4780 will cause the OST to spin in an infinite loop. Release
4781 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4786 Description: statfs data retrieved from /proc may be stale or zero
4787 Details : When reading per-device statfs data from /proc, in the
4788 {kbytes,files}_{total,free,avail} files, it may appear
4789 as zero or be out of date.
4792 Frequency : systems with MD RAID1 external journal devices
4794 Description: lconf's call to blkid is confused by RAID1 journal devices
4795 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4796 of returning all block devices that match the journal UUID.
4799 Frequency : always, for aggregate stripe size over 4GB
4801 Description: assertion fails when trying to use 4GB stripe size
4802 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4803 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4806 Frequency : always on ppc64
4808 Description: the first write on an ext3 filesystem with mballoc got stuck
4809 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4810 perform endianness conversion.
4812 ------------------------------------------------------------------------------
4814 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4816 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4817 INTEROPERATE with older versions automatically. Please read the
4818 user documentation before upgrading any part of a live system.
4819 * WARNING: Lustre networking configuration changes are required with
4820 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4823 * Support for kernels:
4824 2.6.9-22.0.2.EL (RHEL 4)
4825 2.6.5-7.244 (SLES 9)
4826 2.6.12.6 vanilla (kernel.org)
4829 Severity : enhancement
4830 Bugzilla : 7981/8208
4831 Description: Introduced Lustre Networking (LNET)
4832 Details : LNET is new networking infrastructure for Lustre, it includes
4833 a reorganized network configuration mode (see the user
4834 documentation for full details) as well as support for routing
4835 between different network fabrics. Lustre Networking Devices
4836 (LNDs) for the supported network fabrics have also been
4837 created for this new infrastructure.
4839 Severity : enhancement
4840 Description: Introduced Access control lists
4841 Details : clients can set ACLs on files and directories in order to have
4842 more fine-grained permissions than the standard Unix UGO+RWX.
4843 The MDS must be started with the "-o acl" mount option.
4845 Severity : enhancement
4846 Description: Introduced filesystem quotas
4847 Details : Administrators may now establish per-user quotas on the
4850 Severity : enhancement
4852 Description: Configuration change for the XT3
4853 The PTLLND is now used to run Lustre over Portals on the XT3
4854 The configure option(s) --with-cray-portals are no longer used.
4855 Rather --with-portals=<path-to-portals-includes> is used to
4856 enable building on the XT3. In addition to enable XT3 specific
4857 features the option --enable-cray-xt3 must be used.
4862 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4863 Details : A race between allocating a new llcd and re-getting the llcd_lock
4864 allowed another thread to grab newly-allocated llcd.
4866 Severity : enhancement
4868 Description: 2.6 OST async journal commit and locking fix to improve performance
4869 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4870 2.6 kernels are now async as they already were in 2.4 kernels so
4871 that they can commit concurrently with the network bulk transfer.
4872 For block-allocated files the filter allocation semaphore is held
4873 to avoid filesystem fragmentation during allocation. BKL lock
4874 removed for 2.6 xattr operations where it is no longer needed.
4879 Description: lconf incorrectly determined whether two IP networks could talk
4880 Details : In some more complicated routing and multiple-network
4881 configurations, lconf will avoid trying to make a network
4882 connection to a disjoint part of the IP space. It was doing the
4883 math incorrectly for one set of cases.
4888 Description: Fix for potential infinite loop processing records in an llog.
4889 Details : If an llog record is corrupted/zeroed, it is possible to loop
4890 forever in llog_process(). Validate the llog record length
4891 and skip the remainder of the block on error.
4894 Frequency : occasional (liblustre only)
4896 Description: liblustre could not open files whose last component is a symlink
4897 Details : sysio_path_walk() would incorrectly pass the open intent to
4898 intermediate path components.
4901 Frequency : rare (liblustre only with non-standard tuning)
4902 Bugzilla : 7201 (7350)
4903 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4904 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4905 after it has been released, only do this for referenced locks
4907 Severity : enhancement
4909 Description: specifying an (invalid) directory default stripe_size of -1
4910 would reset the directory default striping
4911 Details : stripe_size -1 was used internally to signal directory stripe
4912 removal, now use "all default" to signal dir stripe removal
4913 as a directory striping of "all default" is not useful
4916 Frequency : common for large clusters running liblustre clients
4918 Description: doing an ls when liblustre clients are running is slow
4919 Details : sending a glimpse AST to a liblustre client waits for every AST
4920 to time out, as liblustre clients will not respond. Since they
4921 cannot cache data we refresh the OST lock LVB from disk instead.
4923 Severity : enhancement
4925 Description: doing an ls at the same time as file IO can be slow
4926 Details : enqueue and other "small" requests can be blocked behind many
4927 large IO requests. Create a new OST IO portal for non-IO
4928 requests so they can be processed faster.
4931 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4933 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4934 created file with mode 000
4935 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4936 allow an ftruncate() as a non-root user to a file with mode 000.
4937 Lustre now respects this flag to disable mode checks when
4938 truncating a file owned by the user
4941 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4944 Description: Revoking locks from clients that went dead or catatonic might take
4946 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4947 cancellation of such locks instant on servers without waiting for
4948 any reply from clients. Clients drops these locks when cancel
4949 notification from server is received without replying.
4952 Frequency : liblustre-only, when liblustre client dies or becomes busy
4954 Description: Doing ls on Linux clients can take a long time with active
4956 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4957 granting such locks to it in the first place if possible. Locks
4958 are taken by proxy on the OST during the read or write and
4959 dropped immediately afterward. Add connect flags handling, do
4960 not grant locks to liblustre clients for glimpse ASTs.
4962 Severity : enhancement
4964 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4965 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4966 sequential reads and behaves badly with applications doing random
4967 reads. Improve it by reading ahead at least read region, and
4968 avoiding excessive large RPC for small reads.
4970 Severity : enhancement
4972 Description: Creating more than 1000 files for a single job may cause a load
4973 imbalance on the OSTs if there are also a large number of OSTs.
4974 Details : qos_prep_create() uses an OST index reseed value that is an
4975 even multiple of the number of available OSTs so that if the
4976 reseed happens in the middle of the object allocation it will
4977 still utilize the OSTs as uniformly as possible.
4982 Description: OST or MDS may oops in ping_evictor_main()
4983 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4984 but doesn't restart at beginning of obd_exports_timed list
4987 Severity : enhancement
4989 Description: improve by-nid export eviction on the MDS and OST
4990 Details : allow multiple exports with the same NID to be evicted at one
4991 time without re-searching the exports list.
4994 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4996 Description: MDS may oops in groups_free()
4997 Details : in rare race conditions a newly allocated group_info struct is
4998 freed again, and this can be NULL. The 2.4 compatibility code
4999 for groups_free() checked for a NULL pointer, but 2.6 did not.
5002 Frequency : common for liblustre clients doing little filesystem IO
5003 Bugzilla : 9352, 7313
5004 Description: server may evict liblustre clients accessing contended locks
5005 Details : if a client is granted a lock or receives a completion AST
5006 with a blocking AST already set it would not reply to the AST
5007 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5010 Frequency : lfs setstripe, only systems with more than 160 OSTs
5012 Description: unable to set striping with a starting offset beyond OST 160
5013 Details : llapi_create_file() incorrectly limited the starting stripe
5014 index to the maximum single-file stripe count.
5017 Frequency : LDAP users only
5019 Description: lconf did not handle in-kernel recovery with LDAP properly
5020 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5022 Severity : enhancement
5024 Description: bind OST threads to NUMA nodes to improve performance
5025 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5026 node and do their allocations there to localize memory access
5028 Severity : enhancement
5030 Description: llmount can determine client NID directly from Myrinet (GM)
5031 Details : the client NID code from gmnalnid was moved directly into
5032 llmount, removing the need to use this or specifying the
5033 client NID explicitly when mounting GM clients with zeroconf
5036 Frequency : if client is started with down MDS
5038 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5039 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5040 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5041 ptlrpc_import_delayed_req() because we didn't check if the
5042 request was interrupted, and we also didn't break out of the
5043 event loop if there was no timeout
5048 Description: data loss during non-page-aligned writes to a single file from
5049 both multiple nodes and multiple threads on one node at same time
5050 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5051 inconsistency led to false short-reads, that were cached and later
5052 used by ->prepare_write() to fill in partially written page,
5053 leading to data loss.
5056 Frequency : always, if lconf --abort_recovery used
5058 Description: lconf --abort_recovery fails with 'Operation not supported'
5059 Details : lconf was attempting to abort recovery on the MDT device and not
5062 Severity : enhancement
5064 Description: remove cleanup logs
5065 Details : replace lconf-generated cleanup logs with lustre internal
5066 cleanup routines. Eliminates the need for client-cleanup and
5069 Severity : enhancement
5071 Description: add support for EAs (user and system) on lustre filesystems
5072 Details : it is now possible to store extended attributes in the Lustre
5073 client filesystem, and with the user_xattr mount option it
5074 is possible to allow users to store EAs on their files also
5076 Severity : enhancement
5078 Description: Add possibility (config option) to show minimal available OST free
5080 Details : When compiled with --enable-mindf configure option, statfs(2)
5081 (and so, df) will return least minimal free space available from
5082 all OSTs as amount of free space on FS, instead of summary of
5083 free spaces of all OSTs.
5085 Severity : enhancement
5087 Description: do not expand extent locks acquired on OST-side
5088 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5089 by server: they are not cached anyway.
5092 Frequency : when mmap is used/binaries executed from Lustre
5094 Description: Unmmap pages before throwing them away from read cache.
5095 Details : llap_shrink cache now attempts to unmap pages before discarding
5096 them (if unmapping failed - do not discard). SLES9 kernel has
5097 extra checks that trigger if this unmapping is not done first.
5102 Description: lconf didn't resolve symlinks before checking to see whether a
5103 given mountpoint was already in use
5106 Frequency : when migrating failover services
5107 Bugzilla : 6395, 9514
5108 Description: When migrating a subset of services from a node (e.g. failback
5109 from a failover service node) the remaining services would
5110 time out and evict clients.
5111 Details : lconf --force (implied by --failover) sets the global obd_timeout
5112 to 5 seconds in order to quickly disconnect, but this caused
5113 other RPCs to time out too quickly. Do not change the global
5114 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5116 Severity : enhancement
5117 Frequency : if MDS is started with down OST
5118 Bugzilla : 9439,5706
5119 Description: Allow startup/shutdown of an MDS without depending on the
5120 availability of the OSTs.
5121 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5122 Add appropriate locking and lov-osc refcounts for safe
5123 cleaning. Add osc abort_inflight calls in case the
5124 synchronize never started.
5127 Frequency : occasional (Cray XT3 only)
5129 Description: root not authorized to access files in CRAY_PORTALS environment
5130 Details : The client process capabilities were not honoured on the MDS in
5131 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5132 been accessed by an authorized user then root was able to access
5133 the file on the local client also. The root user capabilities
5134 are now allowed on the MDS, as this environment has secure UID.
5137 Frequency : occasional
5139 Description: ldiskfs "too long searching" message happens too often
5140 Details : A debugging message (otherwise harmless) prints too often on
5141 the OST console. This has been reduced to only happen when
5142 there are fragmentation problems on the filesystem.
5147 Description: Division by zero in statfs when all OSCs are inactive
5148 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5149 lov_statfs divides by value returned by lov_get_stripecnt().
5153 Bugzilla : 9489, 3273
5154 Description: First write from each client to each OST was only 4kB in size,
5155 to initialize client writeback cache, which caused sub-optimal
5156 RPCs and poor layout on disk for the first written file.
5157 Details : Clients now request an initial cache grant at (re)connect time
5158 and so that they can start streaming writes to the cache right
5159 away and always do full-sized RPCs if there is enough data.
5160 If the OST is rebooted the client also re-establishes its grant
5161 so that client cached writes will be honoured under the grant.
5166 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5167 Details : Now I/O RPCs go to different portal number and (presumably) fast
5168 lock requests (and glimses) and other RPCs get their own service
5169 threads pool that should be able to service those RPCs
5172 Severity : enhancement
5174 Description: Ability to exchange lustre version between client and servers and
5175 issue warnings at client side if client is too old. Also for
5176 liblustre clients there is ability to refuse connection of too old
5178 Details : New 'version' field is added to connect data structure that is
5179 filled with version info. That info is later checked by server and
5183 Frequency : rare, liblustre only.
5184 Bugzilla : 9296, 9581
5185 Description: Two simultaneous writes from liblustre at offset within same page
5186 might proceed at the same time overwriting eachother with stale
5188 Details : I/O lock withing llu_file_prwv was released too early, before data
5189 actually was hitting the wire. Extended lock-holding time until
5190 server acknowledges receiving data.
5193 Frequency : extremely rare. Never observed in practice.
5195 Description: avoid generating lustre_handle cookie of 0.
5196 Details : class_handle_hash() generates handle cookies by incrementing
5197 global counter, and can hit 0 occasionaly (this is unlikely, but
5198 not impossible, because initial value of cookie counter is
5199 selected randonly). Value of 0 is used as a sentinel meaning
5200 "unassigned handle" --- avoid it. Also coalesce two critical
5201 sections in this function into one.
5203 Severity : enhancement
5205 Description: allow liblustre clients to delegate truncate locking to OST
5206 Details : To avoid overhead of locking, liblustre client instructs OST to
5207 take extent lock in ost_punch() on client's behalf. New connection
5208 flag is added to handle backward compatibility.
5210 Severity : enhancement
5211 Bugzilla : 4928, 7341, 9758
5212 Description: allow number of OST service threads to be specified
5213 Details : a module parameter allows the number of OST service threads
5214 to be specified via "options ost ost_num_threads={N}" in the
5215 OSS's /etc/modules.conf or /etc/modprobe.conf.
5219 Bugzilla : 6146, 9635, 9895
5220 Description: servers crash with bad pointer in target_handle_connect()
5221 Details : In rare cases when a client is reconnecting it was possible that
5222 the connection request was the last reference for that export.
5223 We would temporarily drop the export reference and get a new
5224 one, but this may have been the last reference and the export
5225 was just destroyed. Get new reference before dropping old one.
5227 Severity : enhancement
5228 Frequency : if client is started with failover MDS
5230 Description: Allow multiple MDS hostnames in the mount command
5231 Details : Try to read the configuration from all specified MDS
5232 hostnames during a client mount in case the "primary"
5235 Severity : enhancement
5237 Description: Stop sending data to evicted clients as soon as possible.
5238 Details : Check if the client we are about to send or are sending data to
5239 was evicted already. (Check is done every second of waiting,
5240 for which l_wait_event interface was extended to allow checking
5241 of exit condition at specified intervals).
5244 Frequency : rare, normally only when NFS exporting is done from client
5246 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5248 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5251 Frequency : always, if srand() is called before liblustre initialization
5253 Description: Liblustre uses system PRNG disturbing its usage by user application
5254 Details : Introduce internal to lustre fast and high-quality PRNG for
5255 lustre usage and make liblustre and some other places in generic
5256 lustre code to use it.
5258 Severity : enhancement
5259 Bugzilla : 9477, 9557, 9870
5260 Description: Verify that the MDS configuration logs are updated when xml is
5261 Details : Check if the .xml configuration logs are newer than the config
5262 logs stored on the MDS and report an error if this is the case.
5263 Request --write-conf, or allow starting with --old_conf.
5265 Severity : enhancement
5267 Description: Handle symlinks in the path when checking if Lustre is mounted.
5268 Details : Resolve intermediate symlinks when checking if a client has
5269 mounted a filesystem to avoid duplicate client mounts.
5274 Description: lconf can hit an error exception but still return success.
5275 Details : The lconf command catches the Command error exception at the top
5276 level script context and will exit with the associated exit
5277 status, but doesn't ensure that this exit status is non-zero.
5282 Description: failure of ptlrpc thread startup can cause oops
5283 Details : Starting a ptlrpc service thread can fail if there are a large
5284 number of threads or the server memory is very fragmented.
5285 Handle this without oopsing.
5288 Frequency : always, only if liblustre and non-default acceptor port was used
5290 Description: liblustre cannot connect to servers with non-default acceptor port
5291 Details : tcpnal_set_default_params() was not called and was therefore
5292 ignoring the environment varaible TCPNAL_PORT, as well as other
5293 TCPNAL_ environment variables
5298 Description: two objects could be created on the same OST for a single file
5299 Details : If an OST is down, in some cases it was possible to create two
5300 objects on a single OST for a single file. No problems other
5301 than potential performance impact and spurious error messages.
5305 Bugzilla : 5681, 9562
5306 Description: Client may oops in ll_unhash_aliases
5307 Details : Client dcache may become inconsistent in race condition.
5308 In some cases "getcwd" can fail if the current directory is
5314 Description: Inode refcounting problems in NFS export code
5315 Details : link_raw functions used to call d_instantiate without obtaining
5316 extra inode reference first.
5320 Bugzilla : 9942, 9903
5321 Description: Referencing freed requests leading to crash, memleaks with NFS.
5322 Details : We used to require that call to ll_revalidate_it was always
5323 followed by ll_lookup_it. Also with revalidate_special() it is
5324 possible to call ll_revalidate_it() twice for the same dentry
5325 even if first occurence returned success. This fix changes semantic
5326 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5327 reference on a request referred from the intent.
5328 ll_intent_release() then releases such a request.
5331 Frequency : rare, normally benchmark loads only
5333 Description: unlinked inodes were kept in memory on the client
5334 Details : If a client is repeatedly creating and unlinking files it
5335 can accumulate a lot of stale inodes in the inode slab cache.
5336 If there is no other client load running this can cause the
5337 client node to run out of memory. Instead flush old inodes
5338 from client cache that have the same inode number as a new inode.
5341 Frequency : SLES9 2.6.5 kernel and long filenames only
5342 Bugzilla : 9969, 10379
5343 Description: utime reports stale NFS file handle
5344 Details : SLES9 uses out-of-dentry names in some cases, which confused
5345 the lustre dentry revalidation. Change it to always use the
5349 Frequency : rare, unless heavy write-truncate concurrency is continuous
5350 Bugzilla : 4180, 6984, 7171, 9963, 9331
5351 Description: OST becomes very slow and/or deadlocked during object unlink
5352 Details : filter_destroy() was holding onto the parent directory lock
5353 while truncating+unlinking objects. For very large objects this
5354 may block other threads for a long time and slow overall OST
5355 responsiveness. It may also be possible to get a lock ordering
5356 deadlock in this case, or run out of journal credits because of
5357 the combined truncate+unlink. Solution is to do object truncate
5358 first in one transaction without parent lock, and then do the
5359 final unlink in a new transaction with the parent lock. This
5360 reduces the lock hold time dramatically.
5363 Frequency : rare, 2.4 kernels only
5365 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5366 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5367 before unmounting their filesystem in order to avoid deadlock.
5368 The kernel_locked() function in 2.4 kernels only checks whether
5369 the kernel lock is held, not whether it is this process that is
5370 holding it as 2.6 kernels do.
5375 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5376 Details : The client ptlrpc code may be trying to reconnect to a down
5377 server before a previous connection attempt has timed out.
5378 Increase the reconnect interval to be longer than the connection
5379 timeout interval to avoid sending duplicate connections to
5383 Frequency : echo_client brw_test command
5385 Description: fix echo_client to work with OST preallocated code
5386 Details : OST preallocation code (5137) didn't take echo_client IO path
5387 into account: echo_client calls filter methods outside of any
5388 OST thread and, hence, there is no per-thread preallocated
5389 pages and buffers to use. Solution: hijack pga pages for IO. As
5390 a byproduct, this avoids unnecessary data copying.
5394 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5395 Description: Client can oops in mdc_commit_close() after open replay
5396 Details : It was possible for the MDS to return an open request with no
5397 transaction number in mds_finish_transno() if the client was
5398 evicted, but without actually returning an error. Clients
5399 would later try to replay that open and may trip an assertion
5400 Simplify the client close codepath, and always return an error
5401 from the MDS in case the open is not successful.
5404 Frequency : rare, 2.6 OSTs only
5406 Description: OST may deadlock under high load on fragmented files
5407 Details : If there was a heavy load and highly-fragmented OST filesystems
5408 it was possible to have all the OST threads deadlock waiting on
5409 allocation of biovecs, because the biovecs were not released
5410 until the entire RPC IO was completed. Instead, release biovecs
5411 as soon as they are complete to ensure forward IO progress.
5413 Severity : enhancement
5415 Description: Support for specifying external journal device at mount
5416 Details : If an OST or MDS device is formatted with an external journal
5417 device, this device major/minor is stored in the ext3 superblock
5418 and may not be valid for failover. Allow detecting and
5419 specifying the external journal at mount time.
5424 Description: Mounting an MDS with pending unlinked files may cause oops
5425 Details : target_finish_recovery() calls mds_postrecov() which returned
5426 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5427 considers this an error and immediately begins cleaning up the
5428 lov, just after starting the mds_lov process
5430 Severity : enhancement
5432 Description: Implement 'lfs df' to report actual free space on per-OST basis
5433 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5434 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5435 usage of objects; '-h' to report in human readable format.
5437 ------------------------------------------------------------------------------
5439 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5446 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5447 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5448 at startup during orphan unlinking. Instead, skip buddy update
5449 before setup, as it will be regenerated after recovery is complete.
5454 Description: If an OST is inactive, its locks might reference stale inodes.
5455 Details : lov_change_cbdata() must iterate over all namespaces, even if
5456 they are inactive to clear inode references from the lock.
5458 Severity : enhancement
5459 Frequency : occasional, if non-standard max_dirty_mb used
5461 Description: Client will block write RPCs if not enough grant
5462 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5463 then the client will block writes while waiting for another RPC
5464 to complete instead of consuming its dirty limit. With change
5465 we get improved performance when max_dirty_mb is small.
5467 Severity : enhancement
5468 Bugzilla : 3389, 6253
5469 Description: Add support for supplementary groups on the MDS.
5470 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5471 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5472 lookups for user supplementary groups into a cache.
5476 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5477 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5478 for liblustre O_EXCL is enforced by MDS.
5482 Description: readdir never returns NULL in liblustre.
5483 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5484 dentry in struct dirent.
5488 Frequency : liblustre only, depends on application IO pattern
5489 Description: liblustre clients evicted if not contacting servers
5490 Details : Don't put liblustre clients into the ping_evictor list, so
5491 they will not be evicted by the pinger ever.
5493 Severity : enhancement
5495 Description: Add ability to evict clients by NID from MDS.
5496 Details : By echoing "nid:$NID" string into
5497 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5498 $NID would be instantly evicted from this MDS and from all active
5499 OSTs connected to it.
5503 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5504 Details : lookup_it_finish() used to query file size from OSTs that was not
5509 Description: service threads change working directory to that of init
5510 Details : Starting lustre service threads may pin the working directory
5511 of the parent thread, making that filesystem busy. Threads
5512 now change to the working directory of init to avoid this.
5516 Frequency : during shutdown only
5517 Description: shutdown with a failed MDS or OST can cause unmount to hang
5518 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5523 Frequency : 2.6 only
5524 Description: chmod/chown may include an extra supplementary group
5525 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5526 supplementary group and if none is specified this is used.
5529 Bugzilla : 5479 (6816)
5531 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5532 Details : If lookup is in progress on a file that is unlinked we might try
5533 to revalidate the inode and fail in revalidate after lookup is
5534 complete and ll_file_open() enqueues the open again but
5535 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5538 Frequency : always, if lconf --abort_recovery used
5540 Description: lconf --abort_recovery fails with 'Operation not supported'
5541 Details : lconf was attempting to abort recovery on the MDT device and not
5544 ------------------------------------------------------------------------------
5546 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5551 Frequency : rare (only unsupported configurations with a node running as an
5553 Bugzilla : 6514, 5137
5554 Description: Mounting a Lustre file system on a node running as an OST could
5556 Details : OSTs now preallocates memory needed to write out data at
5557 startup, instead of when needed, to avoid having to
5558 allocate memory in possibly low memory situations.
5559 Specifically, if the file system is mounted on on OST,
5560 memory pressure could force it to try to write out data,
5561 which it needed to allocate memory to do. Due to the low
5562 memory, it would be unable to do so and the node would
5563 become unresponsive.
5565 Severity : enhancement
5567 Description: Addition of lconf --service command line option
5568 Details : lconf now accepts a '--service <arg>' option, which is
5569 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5571 Severity : enhancement
5573 Description: Failover mode is now the default for OSTs.
5574 Details : By default, OSTs will now run in failover mode. To return to
5575 the old behaviour, add '--failout' to the lmc line for OSTs.
5577 Severity : enhancement
5579 Description: Health checks are now provided for MDS and OSTs
5580 Details : Additional detailed health check information on MSD and OSTs
5581 is now provided through the procfs health_check value.
5584 Frequency : occasional, depends on IO load
5586 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5587 after numerous create/delete cycles
5588 Details : The ext3 inode allocation policy would not allocate new inodes
5589 very well on the OSTs because there are no new directories
5590 being created. Instead we look for groups with free space if
5591 the parent directories are nearly full.
5596 Description: Network or server problems during mount may cause partially
5597 mounted clients instead of returning an error.
5598 Details : The config llog parsing code may overwrite the error return
5599 code during mount error handling, returning success instead
5605 Description: MDS can fail to allocate large reply buffers
5606 Details : After long uptimes the MDS can fail to allocate large reply
5607 buffers (e.g. zconf client mount config records) due to memory
5608 fragmentation or consumption by the buffer cache. Preallocate
5609 some large reply buffers so that these replies can be sent even
5610 under memory pressure.
5614 Frequency : rare (liblustre)
5615 Description: fsx running with liblustre complained that using truncate() to
5616 extend the file doesn't work. This patch corrects that issue.
5617 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5618 ATTR_SIZE and lsm use in llu_setattr_raw.
5622 Frequency : rare, only 2.6 kernels
5623 Description: Unusual file access patterns on the MDS may result in inode
5624 data being lost in very rare circumstances.
5625 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5626 code in the 2.6 kernel caused the inode and/or EA data not to
5627 be read from disk, causing single-file corruption.
5631 Frequency : rare, only 2.6 filesystems using extents
5632 Description: Heavy concurrent write and delete load may cause data corruption.
5633 Details : It was possible under high-load situations to have an extent
5634 metadata block in the block device cache from a just-unlinked
5635 file overwrite a newly-allocated data block. We now unmap any
5636 metadata buffers that alias just-allocated data blocks.
5640 Frequency : filesystems with default stripe_count larger than 77
5641 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5642 Details : lconf specifies an inode size of 4096 bytes when the default
5643 stripe_count is larger than 77. This conflicts with the default
5644 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5645 this case to avoid pinning too much memory for large EAs.
5647 ------------------------------------------------------------------------------
5649 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5654 Frequency : rare (extremely heavy IO load with hundreds of clients)
5656 Description: Client is evicted, gets IO error writing to file
5657 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5658 caused clients to be evicted for AST timeouts. The fixes in
5659 bug 5192 mean we no longer need to have such short AST timeouts
5660 so ldlm_timeout has been increased.
5663 Frequency : occasional during --force or --failover shutdown under load
5664 Bugzilla : 5949, 4834
5665 Description: Server oops/LBUG if stopped with --force or --failover under load
5666 Details : a collection of import/export refcount and cleanup ordering
5667 issues fixed for safer force cleanup
5670 Frequency : only filesystems larger than 120 OSTs
5671 Bugzilla : 5990, 6223
5672 Description: lfs getstripe would oops on a very large filesystem
5673 Details : lov_getconfig used kfree on vmalloc'd memory
5676 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5677 Bugzilla : 6242, 6243
5678 Description: reading from files that had been truncated to a non-zero size
5679 but never opened returned no data
5680 Details : ll_file_read() reads zeros from no-object files to EOF
5685 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5686 Details : There's little harm in aborting MDS/OSS recovery and letting it
5687 try again, so I removed the LASSERT and return an error instead.
5689 Severity : enhancement
5691 Description: New debugging infrastructure for tracking down data corruption
5692 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5693 (b) cover more of the client-side code path, and (c) try to narrow
5694 down where problems occurred
5698 Bugzilla : 3819, 4364, 4397, 6313
5699 Description: Racing close and eviction MDS could cause assertion in mds_close
5700 Details : It was possible to get multiple mfd references during close and
5701 client eviction, leading to one thread referencing a freed mfd.
5703 Severity: : enhancement
5704 Bugzilla : 3262, 6359
5705 Description: Attempts to reconnect to servers are now more aggressive.
5706 Details : This builds on the enhanced upcall-less recovery that was added
5707 in 1.4.2. When trying to reconnect to servers, clients will
5708 now try each server in the failover group every 10 seconds. By
5709 default, clients would previously try one server every 25 seconds.
5714 Description: After recovery, certain operations trigger a failed
5715 assertion on a client.
5716 Details : Failing over an mds, using lconf -d --failover, while a
5717 client was doing a readdir() call would cause the client to
5718 LBUG after recovery completed and the readdir() was resent.
5720 Severity : enhancement
5722 Description: Default groups are now added by lconf
5723 Details : You can now run lconf --group <servicename> without having to
5724 manually add groups with lmc.
5727 Frequency : occasional
5729 Description: Nodes with an elan id of 0 trigger a failed assertion
5732 Frequency : always when accessing e.g. tty/console device nodes
5734 Description: tty and some other devices nodes cannot be used on lustre
5735 Details : file's private_data field is used by device data and lustre
5736 values in there got lost. New field was added to struct file to
5737 store fs-specific private data.
5740 Frequency : when exporting Lustre via NFS
5742 Description: NFSD failed occasionally when looking up a path component
5743 Details : NFSD is looking up ".." which was broken in ext3 directories
5744 that had grown large enough to become hashed.
5747 Frequency : Clusters with multiple interfaces not on the same subnet
5749 Description: Nodes will repeatedly try to reconnect to an interface which it
5750 cannot reach and report an error to the log.
5751 Details : Extra peer list entries will be created by lconf with some peers
5752 unreachable. lconf now validates the peer before adding it.
5755 Frequency : Only if a default stripe is set on the filesystem root.
5757 Description: Setting a default stripe on the filesystem root prevented the
5758 filesystem from being remounted.
5759 Details : The client was sending extra request flags in the root getattr
5760 request and did not allocate a reply buffer for the dir EA.
5763 Frequency : occasional, higher if lots of files are accessed by one client
5764 Bugzilla : 6159, 6097
5765 Description: Client trips assertion regarding lsm mismatch/magic
5766 Details : While revalidating inodes the VFS looks up inodes with ifind()
5767 and in rare cases can find an inode that is being freed.
5768 The ll_test_inode() code will free the lsm during ifind()
5769 when it finds an existing inode and then the VFS later attaches
5770 this free lsm to a new inode.
5774 Bugzilla : 6422, 7030
5775 Description: MDS deadlock between mkdir and client eviction
5776 Details : Creating a new file via mkdir or mknod (starting a transaction
5777 and getting the ns lock) can deadlock with client eviction
5778 (gets ns lock and trying to finish a synchronous transaction).
5781 Frequency : occasional
5782 Description: While starting a server, the fsfilt_ext3 module could not be
5784 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5785 kernels. Previously, lconf would still use the ext3 name
5786 when trying to load modules. Now, it will correctly use
5787 ext3 on 2.4 and ldiskfs on 2.6.
5789 Severity : enhancement
5790 Description: The default stripe count has been changed to 1
5791 Details : The interpretation of the default stripe count (0, to lfs
5792 or lmc) has been changed to mean striping across a single
5793 OST, rather than all available. For general usage we have
5794 found a stripe count of 1 or 2 works best.
5796 Severity : enhancement
5797 Description: Add support for compiling against Cray portals.
5798 Details : Conditional compiling for some areas that are different
5802 Frequency : occasional
5803 Bugzilla : 6409, 6834
5804 Description: Creating files with an explicit stripe count may lead to
5805 a failed assertion on the MDS
5806 Details : If some OSTs are full or unavailable, creating files may
5807 trigger a failed assertion on the MDS. Now, Lustre will
5808 try to use other servers or return an error to the
5812 Frequency : occasional
5814 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5815 caused lock timeout during readahead (since 1.4.2).
5816 Details : Processes doing readahead might match a lock that hasn't been
5817 granted yet if there are overlapping and conflicting lock
5818 requests. The readahead process waits on ungranted lock
5819 (original lock is CBPENDING), while OST waits for that process
5820 to cancel CBPENDING read lock and eventually evicts client.
5822 Severity : enhancement
5824 Description: Initial enabling of flock support for clients
5825 Details : Implements fcntl advisory locking and file status functions.
5826 This feature is provided as an optional mount flag (default
5827 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5828 locking are implemented yet, and those that are are not guaranteed
5829 to be completely correct in production environments.
5830 mount -t lustre -o [flock|noflock] ...
5833 Frequency : occasional
5835 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5836 assertion in the ext3 JBD (journaling) layer.
5837 Details : The b_committed_data struct is protected by the big kernel lock
5838 in 2.4 kernels, serializing journal_commit_transaction() and
5839 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5840 access to this struct. In 2.6 kernels there is finer grained
5841 locking to improve SMP performance of the JBD layer.
5845 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5850 Description: merge and cleanup kernel patches.
5853 -----------------------------------------------------------------------------
5855 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5857 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5858 versions of Lustre. Please update all servers and clients to
5859 version 1.4.2 or later at the same time. You must also run
5860 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5861 to update the configuration logs.
5863 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5864 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5865 - swab llog records of type '0' so we get proper header size/idx (5861)
5866 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5867 - fix rename of one directory over another leaking an inode (5953)
5868 - avoid SetPageDirty on 2.6 (5981)
5869 - don't re-add just-being-destroyed locks to the waiting list (5653)
5870 - when creating new directories, inherit the parent's custom
5871 striping settings if present parent (3048)
5872 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5873 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5874 - don't hold client locks on temporary worklist from l_lru (5666)
5875 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5876 - automatically evict dead clients (5921)
5877 - Update file size properly in create+truncate+fstat case (6196)
5878 - Do not unhash mountpoint dentries, do not allow removal of
5880 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5881 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5882 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5883 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5884 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5885 - Correctly handle reads of files with no objects (6243)
5886 - lctl recover will also mark a device active if deactivate used (5933)
5888 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5889 - allow --write-conf on an MDS with different nettype than client (5619)
5890 - don't write config llogs to MDS for mounts not from that MDS (5617)
5891 - lconf should create multiple TCP connections from a client (5201)
5892 - init scripts are now turned off by default; run chkconfig --on
5893 lustre and chkconfig --on lustrefs to use them
5894 - upcalls are no longer needed for clients to recover to failover
5896 - add --abort-recovery option to lconf to abort recovery on device
5898 - add support for an arbitrary number of OSTs (3026)
5899 - Quota support protocol changes.
5900 - forward compatibility changes to wire structs (6007)
5901 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5902 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5903 - improved "lustre status" script
5904 - initialize blocksize for non-regular files (6062)
5905 - added --disable-server and --disable-client configure options (5782)
5906 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5907 - Vanilla 2.4.29 support
5908 - increase maximum number of obd devices to 520 (6242)
5909 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5910 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5911 SLES 9 2.6.5 kernels
5912 - limit stripes per file to 160 (the maximum EA size) (6093)
5914 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5917 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5918 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5919 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5920 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5921 - small ext3 extents cleanups and fixes (5733)
5922 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5923 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5924 - avoid needless client->OST connect, fix handle mismatch (5317)
5925 - fix DLM error path that led to out-of-sync client, long delays (5779)
5926 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5927 - fix several locking issues related to i_size (5492,5624,5654,5672)
5928 - don't move pending lock onto export if it is already evicted (5683)
5929 - fix kernel oops when creating .foo in unlinked directory (5548)
5930 - fix deadlock in obdfilter statistics vs. object create (5811)
5931 - use time_{before,after} to avoid timer jiffies wrap (5882)
5932 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5933 - Do not leak request if server was not able to process it (5154)
5934 - If mds_open unable to find parent dir, make that negative lookup(5154)
5935 - don't create new directories with extent-mapping (5909, 5936)
5937 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5938 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5939 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5940 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5941 - support for RHEL4 kernel on i686 (5773)
5942 - provide error messages when incompatible logs are encountered (5898)
5944 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5945 * version 1.4.0.10 (1.4.1 release candidate 1)
5947 - don't keep a lock reference when lock is not granted (4238)
5948 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5949 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5950 - drop import inflight refcount on signal_completed_replay error (5255)
5951 - unlock page after async write error during send (3677)
5952 - handle missing objects in filter_preprw_read properly (5265)
5953 - no transno return for symlink open, don't save no-trasno open (3440)
5954 - don't try to complete elan receive that already failed (4012)
5955 - free RPC server reply state on error (5406)
5956 - clean up thread from ptlrpc_start_thread() on error (5160)
5957 - readahead could read extra page into cache that wasn't ejected (5388)
5958 - prevent races in class_attach/setup/cleanup/detach (5260)
5959 - don't dereference de->d_inode after l_dput of de (5458)
5960 - use "int" for stripe value returned from lock_to_stripe (5544)
5961 - mballoc allocation and error-checking fixes in 2.6 (5504)
5962 - block device patches to fix I/O request sizes in 2.6 (5482)
5963 - look up hostnames for IB nals (5602)
5964 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5965 - don't start multiple acceptors for the same port (5277)
5966 - fix incorrect LASSERT in mds_getattr_name (5635)
5967 - export a proc file for general "ping" checking (5628)
5968 - fix "lfs check" to not block when the MDS is down (5628)
5970 - service request history (4965)
5971 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5972 - create an "evict_client" /proc entry on OSTs, like the MDS has
5973 - fix mount usage message, return errors per mount(8) (5168)
5974 - change grep [] to grep "[]" in tests so they work in more UMLs
5975 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5976 - remove ancient LOV_MAGIC_V0 EA support (5047)
5977 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5978 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5979 - allow readahead allocations to fail when low on memory (5383)
5980 - mmap locking landed again, after considerable improvement (2828)
5981 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5983 2004-11-23 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 - flock/lockf fixes (but it's still disabled, pending 5135)
5991 - don't use EXT3 constants in llite code (5094)
5992 - memory shortage at startup could cause assertion (5176)
5994 - reorganization of lov code
5995 - single portals codebase
5997 - add extents/mballoc support (5025)
5998 - direct I/O reads in the obdfilter (4048)
5999 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6001 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6004 - send OST transaction number in read/write reply to free req (4966)
6005 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6006 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6007 - the watchdog thread now runs as interruptible (5246)
6008 - handle missing objects in filter_preprw_read properly (5265)
6009 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6010 - drop import inflight refcount on signal_completed_replay error (5255)
6011 - unlock page after async write error during send (3677)
6012 - return original error code on reconstructed replies (3761)
6013 - no transno return for symlink open, don't save no-trasno open (3440)
6015 - add pid to ldlm debugging output (4922)
6016 - bump the watchdog timeouts -- we can't handle 30sec yet
6017 - extra debugging for orphan dentry/inode bug (5259)
6019 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6022 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6023 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6024 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6025 - fix NULL dereference in /proc/sys/portals/routes (4827)
6026 - allow failed mdc_close() operations to be interrupted (4561)
6027 - stop precreate on OST before MDS would time out on it (4778)
6028 - don't send partial-page writes before EOF from client (4410)
6029 - discard client grant for sub-page writes on large-page clients (4520)
6030 - don't free dentries not owned by NFS code, check generation (4806)
6031 - fix lsm leak if mds_create_objects() fails (4801)
6032 - limit debug_daemon file size, always print CERROR messages (4789)
6033 - use transno after validating reply (3892)
6034 - process timed out requests if import state changes (3754)
6035 - update mtime on OST during writes, return in glimpse (4829)
6036 - add mkfsoptions to LDAP (4679)
6037 - use ->max_readahead method instead of zapping global ra (5039)
6038 - don't interrupt __l_wait_event() during strace
6040 - add software watchdogs to catch hung threads quickly (4941)
6041 - make lustrefs init script start after nfs is mounted
6042 - fix CWARN/ERROR duplication (4930)
6043 - return async write errors to application if possible (2248)
6044 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6045 - print NAL number in %x format (4645)
6046 - update barely-supported suse-2.4.21-171 series (4842)
6047 - support for sles 9 %post scripts
6048 - support for building 2.6 kernel-source packages
6049 - support for sles km_* packages
6051 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6054 - ignore -ENOENT errors in osc_destroy (3639)
6055 - notify osc create thread that OSC is being cleaned up (4600)
6056 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6057 - reconstruct ost_handle() like mds_handle() (4657)
6058 - create a new thread to do import eviction to avoid deadlock (3969)
6059 - let lconf resolve symlinked-to devices (4629)
6060 - don't unlink "objects" from directory with default EA (4554)
6061 - hold socknal file ref over connect in case target is down (4394)
6062 - allow more than 32000 subdirectories in a single directory (3244)
6063 - fix blocks count for O_DIRECT writes (3751)
6064 - OST returns ENOSPC from object create when no space left (4539)
6065 - don't send truncate RPC if file size isn't changing (4410)
6066 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6067 - bind to privileged port in socknal and tcpnal (3689)
6069 - rate limit CERROR/CWARN console message to avoid overload (4519)
6070 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6071 - basic mmap support (3918)
6072 - kernel patch series update from b1_4 (4711)
6074 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6077 - avoid crash during MDS cleanup with OST shut down (2775)
6078 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6079 - don't use bad inodes on the MDS (3744)
6080 - dynamic object preallocation to improve recovery speed (4236)
6081 - don't hold spinlock over lock dumping or change debug flags (4401)
6082 - don't zero obd_dev when it is force cleaned (3651)
6083 - print grants to console if they go negative (4431)
6084 - "lctl deactivate" will stop automatic recovery attempts (3406)
6085 - look for existing locks in ldlm_handle_enqueue() (3764)
6086 - don't resolve lock handle twice in recovery avoiding race (4401)
6087 - revalidate should check working dir is a directory (4134)
6089 - don't always mark "slow" obdfilter messages as errors (4418)
6091 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6094 - don't close LustreDB during write_conf until it is done (3860)
6095 - fix typo in lconf for_each_profile (3821)
6096 - allow dumping logs from multiple threads at one time (3820)
6097 - don't allow multiple threads in OSC recovery (3812)
6098 - fix debug_size parameters (3864)
6099 - fix mds_postrecov to initialize import for llog ctxt (3121)
6100 - replace config semaphore with spinlock (3306)
6101 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6102 - don't allow enqueue to complete on a destroyed export (3822)
6103 - down write_lock before checking llog header bitmap (3825)
6104 - recover from lock replay timeout (3764)
6105 - up llog sem before sending rpc (3652)
6106 - reduce ns lock hold times when setting kms (3267)
6107 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6108 - fix NULL deref and obd_dev leak on setup error (3312)
6109 - replace some LBUG about llog ops with error handling (3841)
6110 - don't match INVALID dentries from d_lookup and spin (3784)
6111 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6112 - fix invalid assertion in ptlrpc_set_wait (3880)
6114 - add libwrap support for the TCP acceptor (3996)
6115 - add /proc/sys/portals/routes for non-root route listing (3994)
6116 - allow setting MDS UUID in .xml (2580)
6117 - print the stack of a process that LBUGs (4228)
6119 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6122 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6123 - make sure to unset replay flag from failed open requests (3440)
6124 - if default stripe count is 0, use OST count for inode size (3636)
6125 - update parent mtime/ctime on client for create/unlink (2611)
6126 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6127 - free recovery state on server during a forced cleanup (3571)
6128 - unregister_reply for resent reqs (3063)
6129 - loop back devices mounting and status check on 2.6 (3563)
6130 - fix resource-creation race that can provoke i_size == 0 (3513)
6131 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6132 - more debugging for page-accounting assertion (3746)
6133 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6134 - avoid deadlock after precreation failure (3758)
6135 - fix race and lock order deadlock in orphan handling (3450, 3750)
6136 - add validity checks when grabbing inodes from l_ast_data (3599)
6138 - add /proc/.../recovery_status to obdfilter (3428)
6139 - lightweight CDEBUG infrastructure, debug daemon (3668)
6140 - change default OSC RPC parameters to be better on small clusters
6141 - turn off OST read cache for files smaller than 32MB
6142 - install man pages and include them in rpms (3100)
6143 - add new init script for (un)mounting lustre filesystems (2593)
6144 - run chkconfig in %post for init scripts (3701)
6145 - drop scimac NAL (unmaintained)
6147 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6150 - clean kiobufs before and after use (3485)
6151 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6152 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6153 - change init script to fail more clearly if not run as root (1528)
6154 - allow clients to reconnect during replay (1742)
6155 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6156 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6157 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6158 - fix cond_resched (3554)
6159 - only evict unfinished clients after recovery (3515)
6160 - allow bulk resend, prevent data loss (3570)
6161 - dynamic ptlrpc request buffer allocation (2102)
6162 - don't allow unlinking open directory if it isn't empty (2904)
6163 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6164 - remove extraneous obd dereference causing LASSERT failure (3334)
6165 - don't use get_cycles() when creating temp. files on the mds (3156)
6166 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6167 - handle EEXIST for set-stripe, set proper directory name (3336)
6169 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6170 - fix ksocknal_fmb_callback() error messages (2918)
6172 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6175 - don't copy lvb into (possibly NULL) reply on error (2983)
6176 - don't deref dentry after dput, don't free lvb on error (2922)
6177 - use the kms to determine writeback rpc length (2947)
6178 - increment oti_logcookies when osc is inactive (2948)
6179 - update client's i_blocks count via lvb messages (2543)
6180 - handle intent open/close of special files properly (1557)
6181 - mount MDS with errors=remount-ro, like obdfilter (2009)
6182 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6183 - don't use cancelling-locks' kms values (2947)
6184 - use highest lock extent for kms, not last one (2925)
6185 - don't dereference ERR_PTR() dentry in error handling path (3107)
6186 - fix thread race in portals_debug_dumplog() (3122)
6187 - create lprocfs device entries at setup instead of at attach (1519)
6188 - common AST error handler, don't evict client on completion race (3145)
6189 - zero nameidata in detach_mnt in 2.6 (3118)
6190 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6191 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6192 - initialize RPC timeout timer earlier for 2.6 (3219)
6193 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6194 - print nal/nid for unknown nid (3258)
6195 - additional checks for oscc recovery before doing precreate (3284)
6196 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6197 - don't crash in mdc_close for bad permissions on open (3285)
6198 - zero i_rdev for non-device files (3147)
6199 - clear page->private before handing to FS, better assertion (3119)
6200 - tune the read pipeline (3236)
6201 - fix incorrect decref of invalidated dentry (2350)
6202 - provide read-ahead stats and refine rpc in flight stats (3328)
6203 - don't hold journal transaction open across create RPC (3313)
6204 - update atime on MDS at close time (3265)
6205 - close LDAP connection when recovering to avoid server load (3315)
6206 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6207 - don't leak open file on MDS after open resend (3325)
6208 - serialize filter_precreate and filter_destroy_precreated (3329)
6209 - loop device shouldn't call sync_dev() for nul device (3092)
6210 - clear page cache after eviction (2766)
6211 - resynchronize MDS->OST in background (2824)
6212 - refuse to mount the same filesystem twice on same mountpoint (3394)
6213 - allow llmount to create routes for mounting behind routers (3320)
6214 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6215 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6216 - fix rare problem with rename on htree directories (3417)
6218 - allow default OST striping configuration per directory (1414)
6219 - fix compilation for qswnal for 2.6 kernels (3125)
6220 - increase maximum number of MDS request buffers for large systems
6221 - change liblustreapi to be useful for external progs like lfsck (3098)
6222 - increase local configuration timeout for slow disks (3353)
6223 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6225 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6228 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6229 - don't overwrite extent policy data in reply if lock was blocked (2901)
6230 - drop filter export grants atomically with removal from device (2663)
6231 - del obd_self_export from work_list in class_disconnect_exports (2908)
6232 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6233 - swab reply message in mdc_close, other PPC fixes (2464)
6234 - fix destroying of named logs (2325)
6235 - overwrite old logs when running lconf --write_conf (2264)
6236 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6237 - fix race in target_handle_connect (2898)
6238 - mds_reint_create() should take same inode create lock (2926)
6239 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6240 - don't close files for self_export to avoid uninitialized obd (2936)
6241 - allow MDS with the same name as client node (2939)
6242 - hold dentry reference for closed log files for unlink (2325)
6243 - reserve space for all logs during transactions (2059)
6244 - don't evict page beyond end of stripe extent (2925)
6245 - don't oops on a deleted current working directory (2399)
6246 - handle hard links to targets without a parent properly (2517)
6247 - don't dereference NULL lock when racing during eviction (2867)
6248 - don't grow lock extents when lots of conflicting locks (2919)
6250 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6253 - account for cache space usage on clients to avoid data loss (974)
6254 - lfsck support in lustre kernel code (2349)
6255 - reduce journal credits needed for BRW writes (2370)
6256 - orphan handling to avoid losing space on client/server crashes
6257 - ptlrpcd can be blocked, stopping ALL progress (2477)
6258 - use lock value blocks to assist in proper KMS, faster stat (1021)
6259 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6260 - recovery for initial connections (2355)
6261 - fixes for mds_cleanup_orphans (1934)
6262 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6263 - block all file creations until orphan recovery completes (1901)
6264 - client remove rq_connection from request struct (2423)
6265 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6266 - recovery timer race (2670)
6267 - mdc_close recovey bug (2532)
6268 - ptlrpc cleanup bug (2710)
6269 - mds timeout on local locks (2588)
6270 - namespace lock held during RPCs (2431)
6271 - handle interrupted sync write properly (2503)
6272 - don't try to handle a message that hasn't been replied to (2699)
6273 - client assert failure during cleanup after abort recovery (2701)
6274 - leak mdc device after failed mount (2712)
6275 - ptlrpc_check_set allows timedout requests to complete (2714)
6276 - wait for inflight reqs when ptlrpcd finishes (2710)
6277 - make sure unregistered services are removed from the srv_list
6278 - reset bulk XID's when resending them (caught by 1138 test)
6279 - unregister_bulk after timeout
6280 - fix lconf error (2694)
6281 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6282 - readahead locks pages, leaves pending causing memory pressure (2673)
6283 - increase OST request buffers to 4096 on large machines (2729)
6284 - fix up permission of existing directories in simple_mkdir (2661)
6285 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6286 - don't assign transno to errored transactions (2742)
6287 - don't delete objects on OST if given a bogus objid from MDS (2751)
6288 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6289 - if rq_no_resend, then timeout request after recovery (2432)
6290 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6291 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6292 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6293 - MDS recovery shouldn't delete objects that it has given out (2730)
6294 - if enqueue arrives after completion, don't clobber LVB (2819)
6295 - don't unlock pages twice when trigger_group_io returns error (2814)
6296 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6297 - don't write pages to disk if there was an error (1450)
6298 - don't ping imports that have recovery disabled (2676)
6299 - take buffered bytes into account when balancing socknal conn (2817)
6300 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6301 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6302 - remove little-endian swabbing of llog records (1987)
6303 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6304 - retry reposting request buffers if they fail (1191)
6305 - grow extent at grant time to avoid granting a revoked lock (2809)
6306 - lock revoke doesn't evict page if covered by a second lock (2765)
6307 - disable VM readahead to avoid reading outside lock extents (2805)
6309 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6310 - updated kernel patches for hp-2.4.20 kernel (2681)
6312 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6315 - fix truncated write corruption (2366)
6316 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6318 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6319 - ensure len(uuid) < 37 in lmc (1171)
6320 - fix ia64 OOPS in llog_test (2255)
6321 - zero end of page at obdfilter for partial page writes (2648)
6322 - don't leave stale dentries around after renames (bug 2428)
6323 - fix timeouts when evicting a client with a single lock held (2642)
6324 - set deadline for the initial HELLO message to drain (2634)
6325 - print out dotted-quad IP addresses in the socknal (2302)
6327 - additional debugging for MDS client eviction problem (2443)
6328 - fix mkfsoptions support for osts (2603, 2604)
6330 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6333 - add series for the vanilla 2.6.0 kernel
6334 - add series for the vanilla 2.4.24 kernel
6335 - add series for a cray x86/64 UL kernel drop
6336 - fix xattr patches for the vanilla 2.4.19 series
6338 - generate true UUIDs in lmc (1171)
6339 - have portals stack dumping break in UML (2466)
6340 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6341 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6342 - wake callback waiting threads on client eviction (2460)
6343 - Add --ptldebug and --subsystem to lmc (1719)
6344 - update assertion to allow safe interrupt allocation
6345 - set rq_no_resend for cancel requests (2432)
6346 - recalculate ptlrpcd timeout after resend (2494)
6347 - call vfs_rmdir when removing pending directories (2368)
6348 - fix renaming a file to itself (2429)
6349 - lmc creates a default one-stripe lov (2454)
6350 - expand procfs space to handle large clusters (2326)
6351 - increase UML stack to avoid overflow
6352 - update lconf's list of debug and subsystem masks
6353 - fix lfs find --obd (2510)
6354 - /proc tunable for disabling filter read caching (2591)
6355 - stop rpm packages from altering slapd.conf (2301)
6356 - disable nagle in the socknal under 0conf (2578)
6357 - choose mds inode size based on stripe count (2572)
6358 - fix kernel-source rpm problems (2516)
6360 - add --disable-doc to avoid pdf generation (2421)
6361 - update documentation, tests, type-os, comments
6362 - avoid format warnings on ia64
6363 - remove the TOE NAL
6364 - tiny code cleanups by removing unused fields
6366 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6369 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6370 - remove the most blatant lies from BUILDING file (2371)
6371 - change default debug level to reasonable production setting
6372 - reduce client side cache size to reduce cache flush time
6373 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6374 - make TCP zerocopy and pinger support enabled by default (2476)
6375 - sync writes completed after process exits caused crashes (2319)
6376 - maintain correct mount count on the MDS (2356)
6377 - backout 1557, because 2316 wasn't really fixed
6378 - better file I/O statistics gathering in /proc
6379 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6380 - another kernel patch to fix zero-copy TCP function export
6381 - don't take duplicate lock when processing re-sent getattr (2420)
6382 - lctl uses obd_self_export instead of creating new conn (2353)
6383 - MDS/OST recovery case which requires object creation asserted (2425)
6384 - move lfs from /usr/sbin to /usr/bin in packages
6385 - fix race between mds_client_add and mds_client_free (2417)
6386 - use kmalloc instead of slabs in portals (2430)
6387 - don't create duplicate records when a failover MDS is present (2442)
6388 - remove unnecessary mount age check (2332)
6389 - don't remove directory inodes from locks prematurely (2451)
6390 - don't break if MDS service name is the same as hostname (2103)
6391 - fix races in client write RPC generation when cache full (2482)
6393 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6396 - remove now-unused request->rq_obd (278)
6397 - if an allocation fails, print out how much memory we've used (1933)
6398 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6399 - add the "configurable stack size" patch to most series files (1256)
6400 - ability to write large log records, for 100+ OST configs (2306)
6401 - fix NULL deref when filter_prep fails (2314)
6402 - fix operator precedence error in filter_sync
6403 - dynamic allocation of socknal TX descriptors (2315)
6404 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6405 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6406 - ensure that CWARN messages in Portals always get to the syslog
6407 - __init/__exit are not for prototype decls (ldlm_init/exit)
6408 - x86-64 compile warning fixes
6409 - fix gateway LMC keyword conflict (2318)
6410 - fix MDS lock inversions in getattr/reint paths (1844)
6411 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6412 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6413 - DLM race condition prevented some lock evictions (2328)
6414 - ENOMEM detection and retry on socknal sends (2230)
6415 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6416 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6417 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6418 - better error messages when a client is rejected during recovery (1505)
6419 - avoid cancelling locks which were never granted, after failure (2330)
6420 - fix i_sem/journal inversion in mds_client_add (2333)
6421 - fix truncate/getattr lock cycle deadlock (2334)
6422 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6423 - fix two rare exit paths which could leak an l_lock() ref (2321)
6424 - fencepost error in MDS/OST orphan recovery (2226)
6425 - make log record alignment 8 bytes (1988)
6426 - lstripe now fails when requested offset > ost_count (2237)
6427 - ensure that all kernel series have a complete list.h (1607)
6428 - fix crashes in special-file operations (2316)
6429 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6431 - allow configurable automake binary, for testing new versions
6432 - small update to the lfs documentation
6434 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6436 * fix negative export reference count in fsfilt_sync (2312)
6438 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6439 * release candidate 0.9.1
6441 - orphans are moved into the PENDING directory for possible recovery
6442 - replayed opens now open by fid for orphan/rename safety (1042)
6443 - last close of an orphan inode generates a transno (683)
6444 - chdir() and mount() now pin the directory entry (1020)
6445 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6446 - discard very old requests without processing them (1502)
6447 - remove some common, well-understood CERRORs (1505)
6448 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6449 - clear "grant" flags in OST replies until OST grant code lands (1644)
6450 - fix read performance by not clobbering i_blksize on client (1598)
6451 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6452 - make LRU size a /proc tunable, clears locks when reduced (707)
6453 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6454 - server locks take references on exports now (1558)
6455 - build fixes for 2.4.20-rh trees (1663)
6456 - return an error from lov_create if all OSCs are inactive (1751)
6457 - fix import levels when a reconnect happens without a timeout (1597)
6458 - exit early from mds_open if we get a lookup error (1749)
6459 - partial page read at EOF wouldn't wait for disk before sending (1642)
6460 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6461 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6462 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6463 - write/truncate lock inversion (1639)
6464 - Don't auto-load obdclass, portals modules during cleanup (1495)
6465 - fix timestamps from jumping to "now" (1763)
6466 - extra journal assertions (1648)
6467 - add an extra multiunlink test (1771)
6468 - fix read_record/write_record API (1776)
6469 - fix leak of offset_extent, possible incorrect i_size later (1772)
6470 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6471 - Debugging for the kqswnal_get_idle_tx problems (1820)
6472 - Allow recovery to be attempted multiple times (1536)
6473 - Write out MDS last_rcvd file after it is first created (1600)
6474 - Fix tx_descriptor leak in failed transmit situations (1827)
6475 - ext3 journaling fixes for assertion failure after IO error (1871)
6476 - class_export_put() on freed export after completion AST error (1896)
6477 - Fix revalidate looping in VFS (1322)
6478 - Don't access a freed export during MDS_REINT timeout (1521)
6479 - Add open-unlink recovery support on the MDS (1673,1764)
6480 - Return an error if no MDS data was read from last_rcvd (1946)
6481 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6482 - Don't setup a disk device that doesn't match exported UUID (317)
6483 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6484 - avoid committing NULL handle in force close
6485 - local.sh is now a one-stripe LOV configuration
6486 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6487 - fix ext3 htree duplicate directory entry corruption (1516)
6488 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6489 - update to new LOV EA format (2097)
6490 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6492 2003-06-15 Phil Schwan <phil@clusterfs.com>
6495 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6496 - per-import/export recovery handling (958, 931, 959)
6497 - multiple last-rcvd slots, for serving multiple FSes (949)
6498 - connections are again shared between multiple imp/exports (963, 964)
6499 - "umount -f" would hang if any requests needed to be sent (393, 978)
6500 - avoid pinning large req buffer by copying for queued messages (989)
6501 - add "uuid" to "lctl device" command to help upcalls (991)
6502 - "open" RPCs with transnos would confuse recovery counters (1037)
6503 - do proper endian conversion of all wire messages (288, 340, 891)
6504 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6505 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6506 - fix ptlrpc_connection leak in target_handle_connect (1174)
6507 - fix import refcounting bug in OST and MDS cleanup (1134)
6508 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6509 - fix very unlikely obd_types race condition (501)
6510 - remove osc_open hack for echo_client (1187)
6511 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6512 - a failure in read_inode2 leads to deadlock (1139)
6513 - cancel ack-locks as soon as transaction is committed (1072)
6514 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6515 - make sure to commitrw after any preprw to avoid deadlock (1162)
6516 - failing to execute a file in a lustre FS would lock inode (1203)
6517 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6518 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6519 - obd_uuid2tgt would walk off the end of the list (1255)
6520 - on IA64 the osc would give portals incorrect bulk size (1258)
6521 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6522 - fix lock inversion caused by new llite matching code (1282)
6523 - limit the number of dirty pages on a client to 10MB (1286)
6524 - timed out locks were not being corrected cancelled (1289)
6525 - fix O_DIRECT above 4GB on IA-32 (1292)
6526 * major user-visible changes
6527 - fail out/fail over policy now controlled by the upcall (993)
6529 - add OBD_PING to check server availability and failure (954)
6530 - lustre messages are now sent in sending host order (288, 340, 891)
6531 - add eadatalen to MDS getattr reply (340)
6532 - OST read replies may contain second buffer, with per-page status (593)
6534 2003-03-11 Phil Schwan <phil@clusterfs.com>
6537 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6538 - failed open()s could cause deadlock; fixed (867, 869)
6539 - stop cancelling OST locks when files are closed (481)
6540 - overlapping XID spaces caused network corruption (851, 853)
6541 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6542 - setattr_raw added, to do single-RPC, server-side setattrs
6543 - lmc/lconf syntax change for OST UUIDs
6544 - fix crashy race condition between ptlrpc_free_req and osc_close
6545 - don't use request in mdc_enqueue if we hit a timeout (889)
6546 - don't set the inode i_size for regular files from the MDS (896)
6547 - handle out of order completion AST (842)
6548 - don't LBUG if a lock request times out after receiving AST (913)
6549 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6550 - if a bad lock AST arrives, send an error instead of dropping entirely
6551 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6552 - fix leak in bulk IO when only partially completed (899, 900, 926)
6553 - fix O_DIRECT for ia64 (55)
6554 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6555 - C-z after timeout could hang a process forever; fixed (977)
6557 - client-side I/O cache (678, 924, 929, 941, 970)
6559 - READPAGE and SETATTRs which don't take server-side locks get
6562 2003-02-11 Phil Schwan <phil@clusterfs.com>
6565 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6566 - Fix fsfilt_extN_readpage() to read a full page of directory
6567 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6568 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6569 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6570 - Most intent operations were removed, in favour of a new RPC mode
6571 that does a single RPC to the server and bypasses most of the VFS
6572 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6573 - Aggressively cancel local locks on DLM servers
6574 - mds_reint_unlink sends EA to the client if it's the last nlink.
6575 client uses that EA to unlink OST objects.
6576 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6577 - recursive symlinks were fixed (439)
6578 - fixed NULL deref in DEBUG_REQ
6579 - filter_update_lastobjid no longer calls sync, which annoyed extN
6580 - fixed multi-client small-writes to a single file problem (445)
6581 - fixed mtime updates during file writes (607)
6582 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6583 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6584 - fixed Linux OST size reporting problem (444, 656)
6585 - OST now updates object mtime with writes or setattr (607, 619)
6586 - client verifies file size before zeroing page past EOF (445)
6587 - OST now writes last allocated objid to disk with allocation (108)
6588 - LOV on echo now works (409)
6590 - mds_reint_unlink sends a new buffer, with the EA included. this
6591 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6592 set if a regular file was being unlinked, and it was the last link
6593 - use PtlGet from the target for bulk writes (315)
6594 - OST now updates object mtime with writes or setattr (607, 619)
6595 - LDLM now has a grant-time callback to revalidate locked items, if
6597 - Many MDS operations were reorganized to combat race conditions
6599 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6600 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6602 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6605 - Fully reactivate OST imports after reconnection (512, others)
6606 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6607 - More graceful error handling for truncating on dead OST (515)
6608 - Don't error out unless we're actually accessing dead stripes (474)
6609 - Fix garbage sizes when stripes are missing (410)
6610 - LRU counters were broken, causing constant lock purge (433, 432)
6611 - garbage on read from stripes with failed OSTs (441)
6612 - mark OSCs as active before reconnecting during recovery (438)
6613 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6614 - lfind did not preserve OST order in output (443)
6615 - symlinks cause hung clients, incorrect data (439)
6616 - stop dereferencing request after dropping refcount (457)
6617 - don't LASSERT(spin_is_locked) on non-SMP (455)
6618 - fixes for many rename() bugs
6619 - fstat didn't correctly synchronize attributes (399)
6620 - server must handle lock cancellation during blocking AST prep (487)
6621 - bulk descriptors were free()d too soon (511)
6622 - fix paths in lconf, which would load incorrect modules (451, 507)
6623 - fix confusing lconf 'host not found' error message (386)
6624 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6625 - fix race condition in mdc_blocking_ast() for inode access (526)
6626 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6627 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6628 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6629 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6630 - fix lstripe in several ways (kernel side) (527)
6631 - fix request leak in ldlm_cli_enqueue (262)
6632 - incorrect OSC was marked inactive after OST failure
6633 - call mds_fs_cleanup before unmounting filesystem (524)
6634 - fix races between taking ns_lock and ldlm_lock_change_resource
6635 - fix races updating LOV export open file list
6636 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6637 - fix recovery NULL deref in ldlm_cli_cancel_unused
6638 - fix some DLM races by using new hash table for lock handles (419)
6639 - permit the client to specify desired inodes, at replay
6640 - duplicate requests when we queue them for replay reintegration
6641 - fix last_rcvd offset calculation
6642 - sync after each recovered transaction, so we always make progress
6643 - never, not always, ERESTART requests without transnos
6644 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6646 - skip replay if the MDS says that the client is already connected
6647 - don't check for a recovery-enabled export to match lctl's UUID
6648 - don't INC_USE_COUNT for phantom exports
6649 - don't crash when cleaning up phantom exports (567)
6650 - don't double-finish or set replay data for errored mdc_open requests
6651 - abort requests when they time out, so we don't get old replies
6652 - send/receive replies for AST messages again
6653 - if the client says that it doesn't have the lock, cancel it on the
6655 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6656 mark it as destroyed, it will all work out when decref is called
6657 - fix module use counts (22, 581)
6659 - ASTs now expect a reply (server cancels lock on error reply)
6661 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6664 - fix many simultaneous client startup (392)
6665 - fix dentry->d_it clobbering
6666 - credentials weren't being shipped for readdir/getattr operations
6667 - remove invalid assertions triggered during some concurrent MD
6669 - proper Lustre versions added (336, 389)
6670 - fix memory leak for create error case (398)
6671 - fix LOV locking bug that would get cli/srv out of sync
6672 - fix echo client over LOV (409)
6673 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6674 - fix double-O_EXCL intent crash (424)
6675 - avoid sending multiple lock CANCELs (352)
6677 - MDS can do multi-client recovery (modulo bugs in new code)
6679 - many updates, edits, cleanups
6681 2002-11-18 Phil Schwan <phil@clusterfs.com>
6684 - fix null d_it dereference (346)
6685 - fix full OST/dbench hang (333)
6686 - fix permission problem with file removal (286)
6687 - fix removal of OSCs from LOV when they fail
6688 - fix NULL deref during bulk timeout (214)
6689 - fix problems related to multiple filesystems on one MDS (241)
6690 - fixed serious subtle metadata locking bugs
6691 - free locks on clients when inodes are removed due to memory
6693 - fix inode pointer in lock data (285)
6694 - partial support for multiple MDS on a single host (241)
6695 - data locks weren't cancelled at clear_inode time (290, 311)
6696 - intent locks could lead to unbounded lock growth (205)
6697 - added a maximum lock count, an LRU list, and a flusher
6698 - fix multiple rename (365)
6699 - properly abstracted the echo client
6700 - OSC locked 1 byte too many; fixed
6701 - rewrote brw callback code:
6702 - fixed recovery bugs related to LOVs (306)
6703 - fixed too-many-pages-in-one-write crash (191)
6704 - fixed (again) crash in sync_io_timeout (214)
6705 - probably fixed callback-related race (385)
6707 - Add capability to MDS protocol
6708 - LDLM cancellations and callbacks on different portals
6710 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6713 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6714 - defer open object destruction to close time (601981, 138)
6715 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6716 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6717 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6718 - don't call obd_setattr in ll_file_release for destroyed objects
6719 * protocol change to lustre_msg: move |version| and add |flags|
6720 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6721 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6722 * added replay of create, unlink, link and rename operations during
6723 MDS failover; recovery should be much more robust now
6724 * remove failed OSCs from LOVs (only lov_create uses this so far)
6725 * the lustre-HOWTO was brought (more) up to date (582544)
6727 2002-10-23 Phil Schwan <phil@clusterfs.com>
6730 - in-use dentries weren't being reused properly (617851)
6731 - prevent multiple LDLM setup (599178)
6732 - fix LOV size calculations for truncate (617853)
6733 - fix client handling of MDS intent errors (POSIX)
6734 - fix permission bug in lovstripe.c test (624321)
6735 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6736 - truncate past end of file could corrupt data
6737 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6738 - a race in recovery could return ETIMEDOUT to apps (623947)
6739 - building outside the source directory was fixed
6740 * the lustre-HOWTO was brought (more) up to date (582544)
6741 * major progress was made on recovery functionality
6743 2002-10-10 Phil Schwan <phil@clusterfs.com>
6746 - recovery deadlock fix
6747 - rm -rf causes LBUG fix (617817)
6748 - file open by multiple tasks fix (618962)
6749 - directory permissions bugs (602707 and 620007)
6750 - journal_stop fixed with locking (611313)
6751 - O_APPEND failures resolved (618273, perhaps 614459)
6752 - lconf PATH fix (619770)
6753 - IA64 build fix (621450)
6754 - RPC buffer sizes scale with amount of memory
6756 2002-10-01 Phil Schwan <phil@clusterfs.com>
6759 - locks would be cancelled without throwing away data pages,
6760 resulting in inconsistent data (605627)
6761 - inode attributes were not always being refreshed (605627, 612449)
6762 - lconf now continues to cleanup after lctl reports an error
6763 - MDS now enforces user permissions (602707)
6764 - lprocfs cleanup fixed, but not yet enabled (614157)
6765 - fixed infinite server hang, should a client not respond to an AST
6766 - avoid going into recovery if user calls readlink() with a buffer
6767 that's too small (613941)
6768 - AST RPCs no longer require replies (614867) -- this may be changed
6769 - don't crash server if client sends an IOV that's too big (611336)
6770 - fixed lock conversion deadlock (611892)
6771 - fixed the following of symlinks (614622)
6772 * recovery: the server can remove locks from a client that dies, other
6773 clients can make progress
6774 * more extN patch fixes
6775 * compile-time configurable ptlrpc buffer allocations
6777 - collaborative read cache document
6778 - Lustre Lite Performance CDR document-in-progress
6780 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6783 - fix typo in patch-2.4.18
6785 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6788 - clear ptlrpc request each time in handle_incoming_request()
6789 - unlink of files now destroys the object on the OST
6791 2002-09-19 Peter Braam <braam@clusterfs.com>
6793 * add hard link support
6794 * change obdfile creation method
6795 * kernel patch changed
6797 2002-09-19 Peter Braam <braam@clusterfs.com>
6800 - stack overflow bug in extN fixed
6802 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6804 * documentation updates
6805 - add man pages for config tools
6806 - update tests/README to describe testing with new config tools
6807 - finish metadata API descriptions
6808 * bug fixes and cleanups
6809 - statfs workaround for 16TB limit
6810 - LOV stripe allocation improved, can stripe on subset of OSTs
6811 - LOV file size/IO offset was wrong for files > 4GB in size
6812 - object EA data was being dropped, caused files to be unreadable
6813 - memory overflow with non-LOV OST caused memory corruption
6814 - fixed regression tests to work with new config tools, obdfilter
6815 - fixed bug when directory size became larger than 1 block
6816 - fixed bug (for single client case) when PWD was deleted
6817 - invalidate local directory pages when doing intent-based ops
6818 - avoid LDLM oops when lock callback contained bad data
6820 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6822 * documentation updates
6823 * bug fixes and cleanups
6824 - configuration tools
6827 - 64-bit compile warnings
6828 - 64-bit internal statfs data
6830 * test_brw on persistent OST devices
6832 * lprocfs (disabled)
6834 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6836 * documentation updates
6837 * bug fixes and cleanups
6838 * configuration tools
6840 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6843 * many small fixes to 0_5_4
6844 * io/network handling
6845 * thinkos in MDS operations
6847 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6850 * crucial basic fixes to 0.5.3
6851 * IOR, Iozone work over Elan
6854 2002-08-07 Phil Schwan <phil@clusterfs.com>
6855 * version 0_5_3, our first alpha
6856 * we use the new Portals iovs
6857 * documentation updates
6858 * bug fixes and cleanups
6859 * small changes in the DLM wire protocol
6861 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6862 * version 0_5_1 with some initial stability,
6863 * locking on MD and file I/O.
6864 * documentation updates
6865 * several bug fixes since 0.5.0
6866 * small changes in wire protocol
6868 2002-07-18 Phil Schwan <phil@clusterfs.com>
6870 * delivered as Lustre Light Alpha
6871 * fixed a crash after handling invalid MDS requests
6872 * fixed directory pages for architectures with non-4k pages sizes
6874 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6875 * release version v0_4_4
6876 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6878 * Fixes a number of interruption problems with OST operations.
6879 * Update documentation for portals header changes
6880 * Move all wire protocol structs/defines to lustre_idl.h
6881 * Fixes symlink length bug.
6882 * Add tcpdump to repository.
6884 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6885 * release version v0_4_3
6886 * Fixes statfs for inodes on extN.
6887 * Fixes bug in runtests which would delete /etc/hosts.
6888 * Use 64-bit object IDs wherever possible (not into VFS though)
6889 Remove ost_get_info, which is unused by lustre, and out of date.
6891 2002-07-03 Peter Braam <braam@clusterfs.com>
6892 * release version v0_4_2 Fixes a lookup error (type not passed)
6893 * move forward to head of Portals
6894 * move forward to latest Lustre kernel
6896 2002-06-25 Peter Braam <braam@clusterfs.com>
6897 * release version v0_4_1. Hopefully stable on single node use.