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 4.18.0-425.3.1.el8 (RHEL8.7)
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 3.10.0-1160.80.1.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 4.18.0-348.23.1.el8 (RHEL8.5)
19 4.18.0-372.32.1.el8 (RHEL8.6)
20 4.4.120-92.70 (SLES12 SP2)
21 4.4.180-94.100 (SLES12 SP3)
22 4.4.0-131 (Ubuntu 16.04)
23 4.15.0-32 (Ubuntu 18.04)
24 5.4.0-48 (Ubuntu 20.04)
25 vanilla linux 5.4.0 (ZFS + ldiskfs)
26 vanilla linux 5.4.21 (ZFS + ldiskfs)
27 vanilla linux 5.4.136 (ZFS + ldiskfs)
28 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
29 * Client primary kernels built and tested during release cycle:
30 5.14.0-70.30.1.el9 (RHEL9.0)
31 4.18.0-425.3.1.el8 (RHEL8.7)
32 5.4.0-37 (Ubuntu 20.04)
33 5.14.21-150400.24.28 (SLES15 SP4)
34 * Other clients known to build on these kernels at some point (others may also work):
35 3.10.0-862.14.4.el7 (RHEL7.5)
36 3.10.0-957.27.2.el7 (RHEL7.6)
37 4.14.0-49.13.1.el7a (RHEL7.5)
38 4.14.0-115.2.2.el7a (RHEL7.6)
39 3.10.0-1062.18.1.el7 (RHEL7.7)
40 3.10.0-1127.19.1.el7 (RHEL7.8)
41 3.10.0-1160.80.1.el7 (RHEL7.9)
42 4.18.0-80.11.2.el8 (RHEL8.0)
43 4.18.0-147.8.1.el8 (RHEL8.1)
44 4.18.0-193.28.1.el8 (RHEL8.2)
45 4.18.0-240.22.1.el8 (RHEL8.3)
46 4.18.0-305.25.1.el8 (RHEL8.4)
47 4.18.0-348.23.1.el8 (RHEL8.5)
48 4.18.0-372.32.1.el8 (RHEL8.6)
49 4.4.120-92.70 (SLES12 SP2)
50 4.4.180-94.100 (SLES12 SP3)
51 4.12.14-95.48 (SLES12 SP4)
52 4.12.14-122.91 (SLES12 SP5)
53 4.12.14-197.75 (SLES15 SP1)
54 5.3.18-24.96 (SLES15 SP2)
55 5.3.18-150300.59.93 (SLES15 SP3)
56 4.4.0-131 (Ubuntu 16.04)
57 4.15.0-48 (Ubuntu 18.04)
58 5.8.0-53 (Ubuntu 20.04.2 HWE)
59 5.11.0-31 (Ubuntu 20.04.3 HWE)
60 5.11.0 (vanilla kernel.org)
61 * Recommended e2fsprogs version: 1.46.5.wc1 or newer
62 * Recommended ZFS version: 2.1.5
63 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
64 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
65 stack overflow. For more information, please refer to bugzilla 17630.
66 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
67 the re-exporting nodes to be mounted with "32bitapi" mount option
69 --------------------------------------------------------------------------------
73 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
74 for currently supported client and server kernel versions.
75 * Server primary kernels built and tested during release cycle:
76 4.18.0-372.9.1.el8 (RHEL8.6)
77 * Other server kernels known to build and work at some point (others may also work):
78 3.10.0-862.14.4.el7 (RHEL7.5)
79 3.10.0-957.27.2.el7 (RHEL7.6)
80 3.10.0-1062.18.1.el7 (RHEL7.7)
81 3.10.0-1127.19.1.el7 (RHEL7.8)
82 3.10.0-1160.71.1.el7 (RHEL7.9)
83 4.18.0-80.11.2.el8 (RHEL8.0)
84 4.18.0-147.8.1.el8 (RHEL8.1)
85 4.18.0-193.28.1.el8 (RHEL8.2)
86 4.18.0-240.22.1.el8 (RHEL8.3)
87 4.18.0-305.25.1.el8 (RHEL8.4)
88 4.18.0-348.2.1.el8 (RHEL8.5)
89 4.4.120-92.70 (SLES12 SP2)
90 4.4.180-94.100 (SLES12 SP3)
91 4.4.0-131 (Ubuntu 16.04)
92 4.15.0-32 (Ubuntu 18.04)
93 5.4.0-48 (Ubuntu 20.04)
94 vanilla linux 5.4.0 (ZFS + ldiskfs)
95 vanilla linux 5.4.21 (ZFS + ldiskfs)
96 vanilla linux 5.4.136 (ZFS + ldiskfs)
97 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
98 * Client primary kernels built and tested during release cycle:
99 4.18.0-372.9.1.el8 (RHEL8.6)
100 5.4.0-37 (Ubuntu 20.04)
101 5.3.18-59.27 (SLES15 SP3)
102 * Other clients known to build on these kernels at some point (others may also work):
103 3.10.0-862.14.4.el7 (RHEL7.5)
104 3.10.0-957.27.2.el7 (RHEL7.6)
105 4.14.0-49.13.1.el7a (RHEL7.5)
106 4.14.0-115.2.2.el7a (RHEL7.6)
107 3.10.0-1062.18.1.el7 (RHEL7.7)
108 3.10.0-1127.19.1.el7 (RHEL7.8)
109 3.10.0-1160.71.1.el7 (RHEL7.9)
110 4.18.0-80.11.2.el8 (RHEL8.0)
111 4.18.0-147.8.1.el8 (RHEL8.1)
112 4.18.0-193.28.1.el8 (RHEL8.2)
113 4.18.0-240.22.1.el8 (RHEL8.3)
114 4.18.0-305.25.1.el8 (RHEL8.4)
115 4.18.0-348.2.1.el8 (RHEL8.5)
116 4.4.120-92.70 (SLES12 SP2)
117 4.4.180-94.100 (SLES12 SP3)
118 4.12.14-95.48 (SLES12 SP4)
119 4.12.14-122.91 (SLES12 SP5)
120 4.12.14-197.75 (SLES15 SP1)
121 5.3.18-24.96 (SLES15 SP2)
122 4.4.0-131 (Ubuntu 16.04)
123 4.15.0-48 (Ubuntu 18.04)
124 5.8.0-53 (Ubuntu 20.04.2 HWE)
125 5.11.0-31 (Ubuntu 20.04.3 HWE)
126 5.11.0 (vanilla kernel.org)
127 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
128 * Recommended ZFS version: 2.1.2
129 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
130 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
131 stack overflow. For more information, please refer to bugzilla 17630.
132 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
133 the re-exporting nodes to be mounted with "32bitapi" mount option
135 --------------------------------------------------------------------------------
139 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
140 for currently supported client and server kernel versions.
141 * Server primary kernels built and tested during release cycle:
142 4.18.0-372.9.1.el8 (RHEL8.6)
143 * Other server kernels known to build and work at some point (others may also work):
144 3.10.0-862.14.4.el7 (RHEL7.5)
145 3.10.0-957.27.2.el7 (RHEL7.6)
146 3.10.0-1062.18.1.el7 (RHEL7.7)
147 3.10.0-1127.19.1.el7 (RHEL7.8)
148 3.10.0-1160.49.1.el7 (RHEL7.9)
149 4.18.0-80.11.2.el8 (RHEL8.0)
150 4.18.0-147.8.1.el8 (RHEL8.1)
151 4.18.0-193.28.1.el8 (RHEL8.2)
152 4.18.0-240.22.1.el8 (RHEL8.3)
153 4.18.0-305.25.1.el8 (RHEL8.4)
154 4.18.0-348.2.1.el8 (RHEL8.5)
155 4.4.120-92.70 (SLES12 SP2)
156 4.4.180-94.100 (SLES12 SP3)
157 4.4.0-131 (Ubuntu 16.04)
158 4.15.0-32 (Ubuntu 18.04)
159 5.4.0-48 (Ubuntu 20.04)
160 vanilla linux 5.4.0 (ZFS + ldiskfs)
161 vanilla linux 5.4.21 (ZFS + ldiskfs)
162 vanilla linux 5.4.136 (ZFS + ldiskfs)
163 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
164 * Client primary kernels built and tested during release cycle:
165 4.18.0-372.9.1.el8 (RHEL8.6)
166 5.4.0-37 (Ubuntu 20.04)
167 5.3.18-59.27 (SLES15 SP3)
168 * Other clients known to build on these kernels at some point (others may also work):
169 3.10.0-862.14.4.el7 (RHEL7.5)
170 3.10.0-957.27.2.el7 (RHEL7.6)
171 4.14.0-49.13.1.el7a (RHEL7.5)
172 4.14.0-115.2.2.el7a (RHEL7.6)
173 3.10.0-1062.18.1.el7 (RHEL7.7)
174 3.10.0-1127.19.1.el7 (RHEL7.8)
175 3.10.0-1160.49.1.el7 (RHEL7.9)
176 4.18.0-80.11.2.el8 (RHEL8.0)
177 4.18.0-147.8.1.el8 (RHEL8.1)
178 4.18.0-193.28.1.el8 (RHEL8.2)
179 4.18.0-240.22.1.el8 (RHEL8.3)
180 4.18.0-305.25.1.el8 (RHEL8.4)
181 4.18.0-348.2.1.el8 (RHEL8.5)
182 4.4.120-92.70 (SLES12 SP2)
183 4.4.180-94.100 (SLES12 SP3)
184 4.12.14-95.48 (SLES12 SP4)
185 4.12.14-122.91 (SLES12 SP5)
186 4.12.14-197.75 (SLES15 SP1)
187 5.3.18-24.96 (SLES15 SP2)
188 4.4.0-131 (Ubuntu 16.04)
189 4.15.0-48 (Ubuntu 18.04)
190 5.8.0-53 (Ubuntu 20.04.2 HWE)
191 5.11.0-31 (Ubuntu 20.04.3 HWE)
192 5.11.0 (vanilla kernel.org)
193 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
194 * Recommended ZFS version: 2.1.2
195 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
196 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
197 stack overflow. For more information, please refer to bugzilla 17630.
198 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
199 the re-exporting nodes to be mounted with "32bitapi" mount option
201 --------------------------------------------------------------------------------
205 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
206 for currently supported client and server kernel versions.
207 * Server primary kernels built and tested during release cycle:
208 3.10.0-1062.18.1.el7 (RHEL7.7)
209 3.10.0-1127.19.1.el7 (RHEL7.8)
210 3.10.0-1160.6.1.el7 (RHEL7.9)
211 * Other server kernels known to build and work at some point (others may also work):
212 3.10.0-862.14.4.el7 (RHEL7.5)
213 3.10.0-957.27.2.el7 (RHEL7.6)
214 4.18.0-80.11.2.el8 (RHEL8.0)
215 4.18.0-147.8.1.el8 (RHEL8.1)
216 4.18.0-193.28.1.el8 (RHEL8.2)
217 4.18.0-240.1.1.el8 (RHEL8.3)
218 4.4.120-92.70 (SLES12 SP2)
219 4.4.180-94.100 (SLES12 SP3)
220 4.4.0-131 (Ubuntu 16.04)
221 4.15.0-32 (Ubuntu 18.04)
222 5.4.0-48 (Ubuntu 20.04)
223 vanilla linux 5.4.0 (ZFS + ldiskfs)
224 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
225 * Client primary kernels built and tested during release cycle:
226 3.10.0-1062.18.1.el7 (RHEL7.7)
227 3.10.0-1127.19.1.el7 (RHEL7.8)
228 3.10.0-1160.6.1.el7 (RHEL7.9)
229 4.12.14-95.48 (SLES12 SP4)
230 4.12.14-122.20 (SLES12 SP5)
231 4.12.14-197.75 (SLES15 SP1)
232 5.3.18-24.24 (SLES15 SP2)
233 4.15.0-48 (Ubuntu 18.04)
234 * Other clients known to build on these kernels at some point (others may also work):
235 3.10.0-862.14.4.el7 (RHEL7.5)
236 3.10.0-957.27.2.el7 (RHEL7.6)
237 4.14.0-49.13.1.el7a (RHEL7.5)
238 4.14.0-115.2.2.el7a (RHEL7.6)
239 4.18.0-80.11.2.el8 (RHEL8.0)
240 4.18.0-147.8.1.el8 (RHEL8.1)
241 4.18.0-193.28.1.el8 (RHEL8.2)
242 4.18.0-240.1.1.el8 (RHEL8.3)
243 4.4.120-92.70 (SLES12 SP2)
244 4.4.180-94.100 (SLES12 SP3)
245 4.4.0-131 (Ubuntu 16.04)
246 5.4.0-37 (Ubuntu 20.04)
247 5.4.0 (vanilla kernel.org)
248 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
249 * Recommended ZFS version: 2.0.0
250 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
251 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
252 stack overflow. For more information, please refer to bugzilla 17630.
253 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
254 the re-exporting nodes to be mounted with "32bitapi" mount option
256 --------------------------------------------------------------------------------
260 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
261 for currently supported client and server kernel versions.
262 * Server primary kernels built and tested during release cycle:
263 3.10.0-1062.1.1.el7 (RHEL7.7)
264 * Other server kernels known to build and work at some point (others may also work):
265 3.10.0-862.14.4.el7 (RHEL7.5)
266 3.10.0-957.27.2.el7 (RHEL7.6)
267 4.18.0-80.7.1.el8 (RHEL8.0)
268 4.4.120-92.70 (SLES12 SP2)
269 4.4.180-94.100 (SLES12 SP3)
270 4.4.0-131 (Ubuntu 16.04)
271 4.15.0-32 (Ubuntu 18.04)
272 vanilla linux 4.6.7 (ZFS only)
273 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
274 * Client primary unpatched kernels built and tested during release cycle:
275 3.10.0-1062.1.1.el7 (RHEL7.7)
276 4.12.14-95.29 (SLES12 SP4)
277 4.15.0-48 (Ubuntu 18.04)
278 * Other clients known to build on unpatched kernels at some point (others may also work):
279 3.10.0-862.14.4.el7 (RHEL7.5)
280 3.10.0-957.27.2.el7 (RHEL7.6)
281 4.14.0-49.13.1.el7a (RHEL7.5)
282 4.14.0-115.2.2.el7a (RHEL7.6)
283 4.18.0-80.7.1.el8 (RHEL8.0)
284 4.4.120-92.70 (SLES12 SP2)
285 4.4.180-94.100 (SLES12 SP3)
286 4.4.0-131 (Ubuntu 16.04)
287 vanilla linux 4.15.0 (ZFS only)
288 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
289 * Recommended ZFS / SPL version: 0.7.13
290 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
291 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
292 stack overflow. For more information, please refer to bugzilla 17630.
293 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
294 the re-exporting nodes to be mounted with "32bitapi" mount option
295 * Note for Ubuntu 14.04.5:
296 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
297 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
299 --------------------------------------------------------------------------------
303 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
304 for currently supported client and server kernel versions.
305 * Server known to build on the following kernels (others may also work):
306 2.6.32-696.18.7.el6 (RHEL6.9)
307 3.10.0-862.14.4.el7 (RHEL7.5)
308 3.10.0-957.el7 (RHEL7.6)
309 4.4.120-92.70 (SLES12 SP2)
310 4.4.162-94.69 (SLES12 SP3)
311 4.4.0-131 (Ubuntu 16.04)
312 4.15.0-32 (Ubuntu 18.04)
313 vanilla linux 4.6.7 (ZFS only)
314 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
315 * Client known to build on unpatched kernels (others may also work):
316 2.6.32-696.18.7.el6 (RHEL6.9)
317 2.6.32-754.6.3.el6 (RHEL6.10)
318 3.10.0-862.14.4.el7 (RHEL7.5)
319 3.10.0-957.el7 (RHEL7.6)
320 4.14.0-49.13.1.el7a (RHEL7.5)
321 4.14.0-115.2.2.el7a (RHEL7.6)
322 4.4.120-92.70 (SLES12 SP2)
323 4.4.162-94.69 (SLES12 SP3)
324 4.4.0-131 (Ubuntu 16.04)
325 4.15.0-32 (Ubuntu 18.04)
326 vanilla linux 4.15.0 (ZFS only)
327 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
328 * Recommended ZFS / SPL version: 0.7.9
329 * Tested with ZFS / SPL version: 0.7.9
330 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
331 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
332 stack overflow. For more information, please refer to bugzilla 17630.
333 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
334 the re-exporting nodes to be mounted with "32bitapi" mount option
335 * Note for Ubuntu 14.04.5:
336 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
337 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
339 --------------------------------------------------------------------------------
341 2018-04-03 Intel Corporation
343 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
344 for currently supported client and server kernel versions.
345 * Server known to build on patched kernels:
346 2.6.32-431.29.2.el6 (RHEL6.5)
347 2.6.32-504.30.3.el6 (RHEL6.6)
348 2.6.32-573.26.1.el6 (RHEL6.7)
349 2.6.32-642.15.1.el6 (RHEL6.8)
350 2.6.32-696.18.7.el6 (RHEL6.9)
351 3.10.0-693.21.1.el7 (RHEL7.4)
352 3.0.101-0.47.71 (SLES11 SP3)
353 3.0.101-107 (SLES11 SP4)
354 3.12.74-60.64.40 (SLES12 SP1)
355 4.4.114-92.64 (SLES12 SP2)
356 4.4.114-94.11 (SLES12 SP3)
357 3.13.0-101 (Ubuntu 14.04, ZFS only)
358 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
359 4.4.0-116 (Ubuntu 16.04)
360 vanilla linux 4.6.7 (ZFS only)
361 * Client known to build on unpatched kernels:
362 2.6.32-431.29.2.el6 (RHEL6.5)
363 2.6.32-504.30.3.el6 (RHEL6.6)
364 2.6.32-573.26.1.el6 (RHEL6.7)
365 2.6.32-642.15.1.el6 (RHEL6.8)
366 2.6.32-696.18.7.el6 (RHEL6.9)
367 3.10.0-693.21.1.el7 (RHEL7.4)
368 3.0.101-0.47.71 (SLES11 SP3)
369 3.0.101-107 (SLES11 SP4)
370 3.12.74-60.64.40 (SLES12 SP1)
371 4.4.114-92.64 (SLES12 SP2)
372 4.4.114-94.11 (SLES12 SP3)
373 3.13.0-101 (Ubuntu 14.04)
374 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
375 4.4.0-116 (Ubuntu 16.04)
377 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
378 * Recommended ZFS / SPL version: 0.7.6
379 * Tested with ZFS / SPL version: 0.7.6
380 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
381 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
382 stack overflow. For more information, please refer to bugzilla 17630.
383 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
384 the re-exporting nodes to be mounted with "32bitapi" mount option
385 * Note for Ubuntu 14.04.5:
386 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
387 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
389 --------------------------------------------------------------------------------
391 2017-07-05 Intel Corporation
393 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
394 for currently supported client and server kernel versions.
395 * Server known to build on patched kernels:
396 2.6.32-431.29.2.el6 (RHEL6.5)
397 2.6.32-504.30.3.el6 (RHEL6.6)
398 2.6.32-573.26.1.el6 (RHEL6.7)
399 2.6.32-642.15.1.el6 (RHEL6.8)
400 2.6.32-696.3.1.el6 (RHEL6.9)
401 3.10.0-514.21.1.el7 (RHEL7.3)
402 3.0.101-0.47.71 (SLES11 SP3)
403 3.0.101-100 (SLES11 SP4)
404 3.12.74-60.64.40 (SLES12 SP1)
405 4.4.59-92.17 (SLES12 SP2)
406 vanilla linux 4.6.7 (ZFS only)
407 * Client known to build on unpatched kernels:
408 2.6.32-431.29.2.el6 (RHEL6.5)
409 2.6.32-504.30.3.el6 (RHEL6.6)
410 2.6.32-573.26.1.el6 (RHEL6.7)
411 2.6.32-642.15.1.el6 (RHEL6.8)
412 2.6.32-696.3.1.el6 (RHEL6.9)
413 3.10.0-514.21.1.el7 (RHEL7.3)
414 3.0.101-0.47.71 (SLES11 SP3)
415 3.0.101-100 (SLES11 SP4)
416 3.12.74-60.64.40 (SLES12 SP1)
417 4.4.59-92.17 (SLES12 SP2)
419 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
420 * Recommended ZFS / SPL version: 0.6.5.9
421 * Tested with ZFS / SPL version: 0.6.5.9
422 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
423 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
424 stack overflow. For more information, please refer to bugzilla 17630.
425 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
426 the re-exporting nodes to be mounted with "32bitapi" mount option
428 --------------------------------------------------------------------------------
430 12-07-2016 Intel Corporation
432 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
433 for currently supported client and server kernel versions.
434 * Server known to build on patched kernels:
435 2.6.32-431.29.2.el6 (RHEL6.5)
436 2.6.32-504.30.3.el6 (RHEL6.6)
437 2.6.32-573.26.1.el6 (RHEL6.7)
438 2.6.32-642.11.1.el6 (RHEL6.8)
439 3.10.0-514.el7 (RHEL7.3)
440 3.0.101-0.47.71 (SLES11 SP3)
441 3.0.101-80 (SLES11 SP4)
442 3.12.62-60.64.8 (SLES12 SP1)
443 vanilla linux 4.5.7 (ZFS only)
444 * Client known to build on unpatched kernels:
445 2.6.32-431.29.2.el6 (RHEL6.5)
446 2.6.32-504.30.3.el6 (RHEL6.6)
447 2.6.32-573.26.1.el6 (RHEL6.7)
448 2.6.32-642.11.1.el6 (RHEL6.8)
449 3.10.0-514.el7 (RHEL7.3)
450 3.0.101-0.47.71 (SLES11 SP3)
451 3.0.101-80 (SLES11 SP4)
452 3.12.62-60.64.8 (SLES12 SP1)
454 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
455 * Recommended ZFS / SPL version: 0.6.5.8
456 * Tested with ZFS / SPL version: 0.6.5.8
457 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
458 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
459 stack overflow. For more information, please refer to bugzilla 17630.
460 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
461 the re-exporting nodes to be mounted with "32bitapi" mount option
463 --------------------------------------------------------------------------------
465 02-29-2016 Intel Corporation
467 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
468 for currently supported client and server kernel versions.
469 * Server known to build on patched kernels:
470 2.6.32-431.29.2.el6 (RHEL6.5)
471 2.6.32-504.30.3.el6 (RHEL6.6)
472 2.6.32-573.12.1.el6 (RHEL6.7)
473 3.10.0-327.3.1.el7 (RHEL7.2)
474 3.0.101-0.47.71 (SLES11 SP3)
475 3.0.101-68 (SLES11 SP4)
476 vanilla linux 4.2.1 (ZFS only)
477 * Client known to build on unpatched kernels:
478 2.6.32-431.29.2.el6 (RHEL6.5)
479 2.6.32-504.30.3.el6 (RHEL6.6)
480 2.6.32-573.12.1.el6 (RHEL6.7)
481 3.10.0-327.3.1.el7 (RHEL7.2)
482 3.0.101-0.47.71 (SLES11 SP3)
483 3.0.101-68 (SLES11 SP4)
486 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
487 * Recommended ZFS / SPL version: 0.6.4.2
488 * Tested with ZFS / SPL version: 0.6.4.2
489 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
490 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
491 stack overflow. For more information, please refer to bugzilla 17630.
492 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
493 the re-exporting nodes to be mounted with "32bitapi" mount option
495 --------------------------------------------------------------------------------
497 03-10-2015 Intel Corporation
499 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
500 for currently supported client and server kernel versions.
501 * Server known to build on patched kernels:
502 2.6.32-431.29.2.el6 (RHEL6.5)
503 2.6.32-504.8.1.el6 (RHEL6.6)
504 3.0.101-0.46 (SLES11 SP3)
505 * Client known to build on unpatched kernels:
506 2.6.32-431.29.2.el6 (RHEL6.5)
507 2.6.32-504.8.1.el6 (RHEL6.6)
508 3.10.0-123.20.1.el7 (RHEL7)
509 3.0.101-0.46 (SLES11 SP3)
510 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
511 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
512 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
513 stack overflow. For more information, please refer to bugzilla 17630.
514 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
515 the re-exporting nodes to be mounted with "32bitapi" mount option
517 Severity : enhancement
519 Description: control OST-index in IDIF via ROCOMPAT flag.
520 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
521 last_rcvd file. For new formatted OST device, it will be auto set;
522 for the case of upgrading from old OST device, you can enable it
523 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
524 enabled, for new created OST-object, its IDIF-in-LMA will contain
525 the OST-index; for the existing OST-object, the OSD will convert
526 old format IDIF as new format IDIF with OST-index stored in the
527 LMA EA when accessing such OST-object or via OI scrub. Once such
528 flag is enabled, it cannot be reverted back, so the system cannot
529 be downgraded to the orignal incompatible version.
531 --------------------------------------------------------------------------------
533 07-30-2014 Intel Corporation
535 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
536 for currently supported client and server kernel versions.
537 * Server known to build on patched kernels:
538 2.6.32-431.20.3.el6 (RHEL6.5)
539 3.0.101-0.31 (SLES11 SP3)
540 * Client known to build on unpatched kernels:
541 2.6.32-431.20.3.el6 (RHEL6.5)
542 3.0.101-0.31 (SLES11 SP3)
545 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
546 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
547 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
548 stack overflow. For more information, please refer to bugzilla 17630.
549 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
550 the re-exporting nodes to be mounted with "32bitapi" mount option
552 --------------------------------------------------------------------------------
554 10-31-2013 Intel Corporation
556 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
557 * Server builds for kernels:
558 2.6.32-358.18.1.el6 (RHEL6.4)
559 * Client builds for unpatched kernels:
560 2.6.32-358.18.1.el6 (RHEL6.4)
561 3.0.92-0.8 (SLES11 SP3)
563 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
564 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
565 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
566 stack overflow. For more information, please refer to bugzilla 17630.
567 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
568 the re-exporting nodes to be mounted with "32bitapi" mount option
570 --------------------------------------------------------------------------------
572 05-30-2013 Intel Corporation
574 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
575 * Server builds for kernels:
576 2.6.32-358.6.2.el6 (RHEL6.4)
577 * Client builds for unpatched kernels:
578 2.6.32-358.6.2.el6 (RHEL6.4)
579 3.0.74-0.6.6 (SLES11 SP2)
581 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
582 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
583 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
584 stack overflow. For more information, please refer to bugzilla 17630.
585 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
586 the re-exporting nodes to be mounted with "32bitapi" mount option
588 --------------------------------------------------------------------------------
590 9-30-2012 Whamcloud, Inc.
592 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
593 * Server support for kernels:
594 2.6.32-279.5.1.el6 (RHEL6)
595 * Client support for unpatched kernels:
596 2.6.32-279.5.1.el6 (RHEL6)
598 2.6.18-238.19.1.el5 (RHEL5)
599 2.6.16.60-0.69.1 (SLES10)
600 * Recommended e2fsprogs version: 1.41.90.wc4
601 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
602 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
603 stack overflow. For more information, please refer to bugzilla 17630.
604 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
605 the re-exporting nodes to be mounted with "32bitapi" mount option
607 Severity : enhancement
609 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
611 Severity : enhancement
613 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
615 Severity : enhancement
617 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
619 Severity : enhancement
621 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
623 Severity : enhancement
625 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
627 --------------------------------------------------------------------------------
629 04-30-2012 Whamcloud, Inc.
631 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
632 * Server support for kernels:
633 2.6.32-220.4.2.el6 (RHEL6)
634 * Client support for unpatched kernels:
635 2.6.32-220.4.2.el6 (RHEL6)
637 2.6.18-238.19.1.el5 (RHEL5)
638 2.6.16.60-0.69.1 (SLES10)
639 * Recommended e2fsprogs version: 1.41.90.wc4
640 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
641 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
642 stack overflow. For more information, please refer to bugzilla 17630.
643 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
644 the re-exporting nodes to be mounted with "32bitapi" mount option
646 Severity : enhancement
648 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
650 Severity : enhancement
652 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
654 Severity : enhancement
656 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
658 Severity : enhancement
660 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
662 --------------------------------------------------------------------------------
664 09-30-2011 Whamcloud, Inc.
666 * Server support for kernels:
667 2.6.18-238.19.1.el5 (RHEL5)
668 2.6.32-131.6.1.el6 (RHEL6)
669 * Client support for unpatched kernels:
670 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
671 2.6.18-238.19.1.el5 (RHEL5)
672 2.6.32-131.6.1.el6 (RHEL6)
673 2.6.29.4-167.fc11 (FC11)
674 2.6.16.60-0.69.1 (SLES10)
676 * Recommended e2fsprogs version: 1.41.90.wc3
677 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
678 Lustre filesystem with 4K stack may cause a stack overflow. For more
679 information, please refer to bugzilla 17630.
680 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
681 mounted with "32bitapi" mount option
683 Severity : enhancement
684 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
686 Severity : enhancement
687 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
689 Severity : enhancement
690 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
693 Frequency : only with SLES10
695 Description: Use OFED "KMP" provided by Novell
696 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
697 Lustre is now built against this package. That means you need to
698 install the ofed-kmp package from Novell for the patchless client.
699 Note that the ofed-kmp that Novell ships may not exactly match the
700 kernel version but should still be compatible.
702 Severity : enhancement
704 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
705 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
707 --------------------------------------------------------------------------------
709 2010-07-15 Oracle, Inc.
711 * Server support for kernels:
712 2.6.18-164.11.1.el5 (RHEL 5)
713 2.6.18-164.11.1.0.1.el5 (OEL 5)
714 * Client support for unpatched kernels:
715 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
716 2.6.18-164.11.1.el5 (RHEL 5),
717 2.6.18-164.11.1.0.1.el5 (OEL 5)
718 2.6.16.60-0.42.8 (SLES 10),
720 2.6.29.4-167.fc11 (FC11)
721 * Recommended e2fsprogs version: 1.41.10-sun2
723 Severity : enhancement
725 Description: Distribute lfs_migrate script with lustre
726 Details : The lfs_migrate mentioned in the manual is now included in the
727 lustre tree in lustre/scripts and will distributed in lustre RPMs.
729 Severity : enhancement
731 Description: Failover nid list added to import proc file.
733 Severity : enhancement
735 Description: "som_preview" mount option required to use SOM
736 Details : The clients will need to use som_preview mount option
740 Description: Lustre does not do 1MB IOs to HW RAID
741 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
743 Severity : enhancement
745 Description: lfs quota output is not very convenient for awk/sed-parsing
746 Details : Some positions in lfs quota output table could be empty or
747 non-empty which made it hard to parse it with scripts, now
748 a dash is put instead of space where there is not supposed
753 Description: Update to OFED 1.5.1
754 Details : OFED 1.5.1 will be used in environments that don't use
755 vendor supplied OFED.
759 Description: Remove lnet. prefix from lctl params display
760 Details : Also renamed lnet.memused to lnet_memused to
765 Description: Use conf_param -d to remove permanant settings.
769 Description: Add list_param -R
770 Details : lctl list_param now supports recursive mode.
773 Bugzilla : 21829, 22312
774 Description: Fix llobdstat.
775 Details : 1. Now updating the write_bytes counter
776 2. Port script fixes from 1.8.
780 Description: decrease the usage of memory on clients.
781 Details : 1. On clients, recycle dentries and inodes unused.
782 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
787 Description: llobdstat fix and enhancement
788 Details : add a counter to set a limit to how many samples will be returned
789 fix a wildcard in the path to limit to obdfilter stats only
793 Description: lfs check only allowed for root
794 Details : Code cleanups and added a regresion test of non-root use of
799 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
800 Details : Fix hr->hr_index overload.
804 Description: Fix FIEMAP functionality.
808 Description: FC11 patchless client support added
812 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
813 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
817 Description: New acc-sm script and various test-framework enhancments.
821 Description: Previously landed patch was reverted.
825 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
829 Description: Add ldiskfs maxdirsize mount option.
833 Description: Fix mkfs.lustre support for 16TB LUNs.
837 Description: Print mmp_check_interval
838 Details : print mmp_check_interval and make it possible to abort
839 mount operation in case it takes too long.
843 Description: Add cascading_rw to lustre mpi tests.
847 Description: Store SOM data in EA.
848 Details : Also several SOM related cleanups.
852 Description: Add refcount to md_open_data
856 Description: Use refcount on lov_request_set to prevent memory corruption.
860 Description: Statahead: use parent dir's i_mutex for synchronization.
864 Description: loadgen improvements.
868 Description: Enable peer health detection in ksocklnd and ko2iblnd.
872 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
876 Description: Fix race on obd_devcie:md_stats
880 Description: Fix SLES 11 client builds.
881 Details : Resolved uninitalized pointer gcc warnings.
885 Description: Allow non-root access for "lfs check".
886 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
890 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
892 Severity : enhancement
894 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
895 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
899 Description: Sanity test_27p fails.
900 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
901 Sanity tests 27 were cleaned up and improved.
905 Description: Unify req format on client/servers
906 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
907 buffers and swabbers approach..
912 Description: drop join file stuff.
914 Severity : enhancement
916 Description: Add 2.6.27 and SLES11 (patchless client) support
920 Description: Fix in ptlrpc_expire_one_request() to print the signed time
923 Severity : enhancement
925 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
928 Severity : enhancement
930 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
932 Severity : enhancement
934 Description: correctly handle big reply message.
935 Details : send LNet event if reply is bigger than buffer and adjust this buffer
940 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
942 Severity : enhancement
944 Description: Add support for OEL5.
948 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
950 Severity : enhancement
952 Description: "weak-modules" support
953 Details : Implement "weak-modules" support which enables kernel modules
954 to be used with any kernel that implements the same kABI. In
955 order to achieve this modules are now installed in
956 /lib/modules/$(uname -r)/updates/kernel on all distributions.
960 Description: client could not reconnect to OST because of an active request.
961 Details : The client now retries to reconnect to the same server, if a
962 connect request failed with EBUSY or -EAGAIN. Test case 35b
963 conf-sanity.sh has been created.
967 Description: Don't allow make backward step on assignin osc next id.
968 Details : race between allocation next id and ll_sync thread can be cause
969 of set wrong osc next id and can be kill valid ost objects.
973 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
977 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
981 Description: Error handling in osc_statfs_interpret() has been improved.
982 Details : Check in osc_statfs_interpret() for EBADR.
986 Description: Do not send statfs() requests to OSTs disabled by administrator.
987 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
989 Severity : enhancement
991 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
993 Severity : enhancement
995 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
997 Severity : enhancement
999 Description: Enhancement for lfs(1) command to use numeric uid/gid.
1003 Description: Improved cleanup procedure for conf-sanity test_46a.
1007 Description: Do not start bulk transfer, if client tried to reconnect.
1008 Details : Do not start bulk transfer, if client tried to reconnect,
1009 but failed due to active rpcs. Check for exp_abort_active_req
1010 has been implemented.
1014 Description: Prevent inconsistences between linux and lustre mount structures.
1015 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
1016 Make the sleep interruptible.
1018 Severity : enhancement
1020 Description: provide server to client comms path
1021 Details : server to client communications path via new LDLM_SET_INFO rpc
1025 Description: truncate starts GFP_FS allocation under transaction causing deadlock
1026 Details : ldiskfs_truncate calls grab_cache_page which may start page
1027 allocation under an open transaction. This may lead to
1028 calling prune_icache with consequent lustre reentrance.
1032 Description: send correctly lsm on open replay
1033 Details : MDS is trust to LSM size on replay open, but client can set wrong size
1038 Description: enable client interface failover
1039 Details : When a child reconnects from another NID, properly update export
1040 nid hash position and ldlm reverse import.
1042 Severity : enhancement
1044 Description: Communicate OST degraded/readonly state via statfs to MDS
1045 Details : Flags in the statfs returned from OSTs indicate whether the
1046 OST is in a degraded RAID state, or if the filesystem has
1047 turned read-only after a filesystem error is detected.
1052 Description: don't panic if EPROTO was hit when reading symlink
1053 Details : correctly handling request reference in error cases.
1055 Severity : enhancement
1057 Description: Add LustreNetLink, a kernel-userspace communcation path.
1059 Severity : enhancement
1061 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1066 Description: access to llog context before init.
1067 Details : move handling CATALOGS file at osc layer and forbid access to llog
1068 context before init.
1071 Frequency : rare, only when client is writing to OST on local node
1073 Description: Avoid deadlock for local client writes
1074 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1075 memory freeing context. This allows OST threads to set the
1076 PF_MEMALLOC flag on task structures in order to allocate memory
1077 from reserved pools and complete IO.
1078 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1079 so that the OST threads generate memory pressure and allow
1080 inactive pages to be reclaimed.
1082 Severity : enhancement
1084 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1089 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1090 Details : move ldlm namespace creation in setup phase to avoid grab
1091 _lprocfs_lock with cli_sem held.
1095 Description: Temporarily disable grant shrink.
1096 Details : Disable the feature for debugging.
1099 Frequency : only when formatting test filesystems
1101 Description: Unable to run several mkfs.lustre on loop devices at the same time
1102 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1103 formatting. The solution is to proper handle the error.
1105 Severity : enhancement
1107 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1109 Severity : enhancement
1111 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1116 Description: resolve race between obd_disconnect and class_disconnect_exports
1117 Details : if obd_disconnect will be called to already disconnected export he
1118 forget release one reference and osc module can't unloaded.
1120 Severity : enhancement
1122 Description: Allow tuning service thread via /proc
1123 Details : For each service a new
1124 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1125 created that can be used to set min/max thread counts, and get the
1126 current number of running threads.
1131 Descriptoin: don't return error if have partially created objects for file.
1132 Details : lov_update_create_set uses set->set_success as index for created
1133 objects, so if some requests will be failed, they will have hole
1134 at end of array and qos_shrink_lsm for allocate correct lsm.
1136 Severity : enhancement
1138 Description: Update OFED support to 1.4.1
1143 Description: Reduce small size read RPC
1144 Details : Set read-ahead limit for every file and only do read-ahead when
1145 available read-ahead pages are bigger than 1M to avoid small size
1148 Severity : enhancement
1150 Description: move AT tunable parameters for more consistent usage
1151 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1153 Severity : enhancement
1155 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1156 Details : allow skip disconnected ost for send statfs request and hide error
1162 Description: fix racy locking of mballoc block bitmaps causing BUG
1163 Details : The locking of the mballoc buddy bitmap and the in-memory
1164 block bitmap was using two different spin locks in some
1165 cases. This made it possible to incorrectly access the
1166 mballoc bitmap while another process was modifying it,
1167 causing a sanity assertion to fail. While no on-disk corruption
1168 was reported, there was some risk of this happening.
1171 Frequency : rare, on llog test 6
1173 Descriptoin: don't allow connect to already connected import
1174 Details : allowing connect to already connected import is hide connecting problem.
1179 Descriptoin: allow kill process which wait statahead result
1180 Details : for some reasons 'ls' can stick in waiting result from statahead,
1181 in this case need way for kill this process.
1183 Severity : enhancement
1185 Description: Add state history info file, enhance import info file
1186 Details : Track import connection state changes in a new osc/mdc proc file;
1187 add overview-type data to the osc/mdc import proc file.
1189 Severity : enhancement
1191 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1193 Severity : enhancement
1195 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1200 Descriptoin: OST grant too much space to client even there are not enough space.
1201 Details : Client will shrink its grant cache to OST if there are no write
1202 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1203 this grant cache if there are already not enough avaible space
1204 (left_space < total_clients * 32M).
1207 Frequency : start MDS on uncleanly shutdowned MDS device
1209 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1210 Details : stay in waiting mds<>ost recovery finished produce random bugs
1211 due race between two ll_sync thread for one lov target. send
1212 ACTIVATE event only if connect realy finished and import have
1216 Frequency : rare, connect and disconnect target at same time
1218 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1219 Details : don't call obd_disconnect under lov_lock. this long time
1220 operation and can block ptlrpcd which answer to connect request.
1225 Descriptoin: don't lose wakeup for imp_recovery_waitq
1226 Details : recover_import_no_retry or invalidate_import and import_close can
1227 both sleep on imp_recovery_waitq, but we was send only one wakeup
1231 Frequency : always with long access acl
1233 Descriptoin: mds can't pack reply with long acl.
1234 Details : mds don't control size of acl but they limited by reint/getattr
1237 Severity : enhancement
1239 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1241 Severity : enhancement
1243 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1246 Frequency : start MDS on uncleanly shutdowned MDS device
1248 Descriptoin: aborting recovery hang on MDS
1249 Details : don't throttle destroy RPCs for the MDT.
1252 Frequency : on remount
1254 Description: external journal device not working after the remount
1255 Details : clear dev_rdonly flag for external journal devices in
1261 Description: shutdown vs evict race
1262 Details : client_disconnect_export vs connect request race.
1263 if client will evicted at this time - we start invalidate
1264 thread without referece to import and import can be freed
1268 Frequency : rare, need acl's on inode.
1270 Description: client can't handle ost additional correctly
1271 Details : if ost was added after client connected to mds client can have
1272 hit lnet_try_match_md ... to big messages to wide striped files.
1273 in this case need teach client to handle config events about add
1274 lov target and update client max ea size at that event.
1276 Severity : enhancement
1278 Description: Changelogs
1279 Details : Changelogs are a lightweight mechanism to track filesystem
1280 metadata and namespace changes. The changelog is recorded
1281 permanently on the MDTs, and is periodically "consumed" / purged
1282 when records are no longer needed.
1284 Severity : enhancement
1286 Description: compact fld format with extents
1287 Details : Store range of seq rather than every seq in FLD. Seq
1288 controller update FLD rather than clients. In Case of CMD, mdt0
1289 has FLD, all other metadata server act as non persistent proxy
1290 for FLD queries and cache fld entries in fld cache.
1295 Description: don't skip ost target if they assigned to file
1296 Details : Drop slow OSCs if we can, but not for requested start idx.
1297 This means "if OSC is slow and it is not the requested
1298 start OST, then it can be skipped, otherwise skip it only
1299 if it is inactive/recovering/out-of-space.
1303 Description: more cleanup in mds_lov
1304 Details : not send LOV EA under replay, we can't know about they size at this
1305 time. Don't allow client connect to mds before any ost connected,
1306 for avoid problems with LOV EA size and returning EIO to client.
1308 Severity : enhancement
1310 Description: Interoperability at server side (Disk interoperability)
1312 Severity : enhancement
1314 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1316 Severity : enhancement
1318 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1320 Severity : enhancement
1322 Description: New client IO stack (CLIO).
1324 Severity : enhancement
1326 Description: Commit on sharing. Eliminate inter-client dependencies between
1327 uncommitted transactions by doing transaction commits.
1328 Thereby clients may recovery independently.
1331 Frequency : Create a symlink file with a very long name
1333 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1334 Details : If there is no extra space in the request for early cancels,
1335 ldlm_req_handles_avail() returns 0 instead of a negative value.
1337 Severity : enhancement
1339 Description: Add /proc entry for import status
1340 Details : The mdc, osc, and mgc import directories now have
1341 an import directory that contains useful import data for debugging
1342 connection problems.
1344 Severity : enhancement
1346 Description: Re-disable certain /proc logging
1347 Details : Enable and disable client's offset_stats, extents_stats and
1348 extents_stats_per_process stats logging on the fly.
1351 Frequency : Only on FC kernels 2.6.22+
1353 Description: oops in statahead
1354 Details : Do not drop reference count for the dentry from VFS when lookup,
1355 VFS will do that by itself.
1357 Severity : enhancement
1359 Description: Generic /proc file permissions
1360 Details : Set /Proc file permissions in a more generic way to enable non-
1361 root users operate on some /proc files.
1365 Description: Hitting mdc_commit_close() ASSERTION
1366 Details : Properly handle request reference release in
1367 ll_release_openhandle().
1371 Description: quota recovery deadlock during mds failover
1372 Details : This patch includes att18982, att18236, att18237 in bz14840.
1374 1. fix osts hang when mds does failover with quotaon
1375 2. prevent watchdog storm when osts threads wait for the
1380 Frequency : only patchless client
1381 Description: add workaround for race between add/remove dentry from hash
1383 Severity : enhancement
1385 Description: Allow OST glimpses to return PW locks
1389 Description: LBUG when llog conf file is full
1390 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1394 Description: Prevent import from entering FULL state when server in recovery
1398 Description: service mount cannot take device name with ":"
1399 Details : Only when device name contains ":/" will mount treat it as
1405 Description: replace ptlrpcd with the statahead thread to interpret the async
1406 statahead RPC callback
1410 Frequency : on recovery
1411 Description: I/O failures after umount during fail back
1412 Details : if client reconnected to restarted server we need join to recovery
1413 instead of find server handler is changed and process self
1414 eviction with cancel all locks.
1416 Severity : enhancement
1418 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1420 Severity : enhancement
1422 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1424 Severity : enhancement
1426 Description: Upcall on Lustre log has been dumped
1427 Details : Allow for a user mode script to be called once a Lustre log has
1428 been dumped. It passes the filename of the dumped log to the
1429 script, the location of the script can be specified via
1430 /proc/sys/lnet/debug_log_upcall.
1435 Description: avoid idr_remove called for id which is not allocated.
1436 Details : Move assigment s_dev for clustered nfs to end of initialization,
1437 to avoid problem with error handling.
1442 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1443 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1444 into CONN_USED_HASH and this prodice warning when put connection
1445 again in unused hash.
1448 Severity : enhancement
1450 Description: File striping can now be set to use an arbitrary pool of OSTs.
1452 Severity : enhancement
1454 Description: Export bytes_read/bytes_write count on OSC/OST.
1458 Description: Early reply size mismatch, MGC loses connection
1459 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1460 the connect flags are properly negotiated.
1465 Description: quotas are not honored with O_DIRECT
1466 Details : all writes with the flag O_DIRECT will use grants which leads to
1467 this problem. Now using OBD_BRW_SYNC to guard this.
1471 Description: add quota statistics
1472 Details : 1. sort out quota proc entries and proc code.
1473 2. add quota statistics
1475 Severity : enhancement
1477 Description: enable quota support for HEAD.
1481 Description: Properly propagate oinfo flags from lov to osc for statfs
1482 Details : restore missing copy oi_flags to lov requests.
1484 Severity : enhancement
1486 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1487 : l_getgroups(8), lst(8), routerstat(8)
1488 Details : included man pages for llobdstat(8), llstat(8),
1489 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1491 Severity : enhancement
1493 Description: configure's --enable-quota should check the
1494 : kernel .config for CONFIG_QUOTA
1495 Details : configure is terminated if --enable-quota is passed but
1496 : no quota support is in kernel
1500 Description: Remove portals compatibility
1501 Details : Remove portals compatibility, not interoperable with releases
1506 Description: Resolve device initialization race
1507 Details : Prevent proc handler from accessing devices added to the
1508 obd_devs array but yet be intialized.
1510 Severity : enhancement
1512 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1514 Severity : enhancement
1516 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1521 Description: Using wrong pointer in osc_brw_prep_request
1522 Details : Access to array[-1] can produce panic if kernel compiled with
1523 CONFIG_PAGE_ALLOC enabled
1527 Description: Client runs out of low memory
1528 Details : Consider only lowmem when counting initial number of llap pages
1532 Description: *optional* service tags registration
1533 Details : if the "service tags" package is installed on a Lustre node
1534 When the filesystem is mounted, a local-node service tag will
1535 be created. See http://inventory.sun.com/ for more information
1536 about the Service Tags asset management system.
1540 Description: Kernel BUG tries to release flock
1541 Details : Lustre does not destroy flock lock before last reference goes
1542 away. So always drop flock locks when client is evicted and
1543 perform unlock regardless of successfulness of speaking to MDS.
1547 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1551 Description: sanity test 65a fails if stripecount of -1 is set
1552 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1557 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1558 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1561 Severity : enhancement
1563 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1567 Description: do not process already freed flock
1568 Details : flock can possibly be freed by another thread before it reaches
1569 to ldlm_flock_completion_ast.
1573 Description: LBUG during stress test
1574 Details : Need properly lock accesses the flock deadlock detection list.
1578 Description: oops in page fault handler
1579 Details : kernel page fault handler can return two special 'pages' in error
1580 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1584 Description: timeout with invalidate import.
1585 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1586 should be handled by ptlrpcd. This produce long age waiting and
1587 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1589 Severity : enhancement
1591 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1595 Description: enable MGS and MDT services start separately
1596 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1597 a co-located MGS without starting the MGS, which is a complement
1598 to 'nosvc' mount option.
1602 Frequency : after recovery
1603 Description: precreate to many object's after del orphan.
1604 Details : del orphan st in oscc last_id == next_id and this triger growing
1605 count of precreated objects. Set flag LOW to skip increase count
1606 of precreated objects.
1610 Frequency : rare, on clear nid stats
1611 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1612 Details : when clean nid stats sometimes try destroy live entry,
1613 and this produce panic in free.
1617 Description: Stack overflow during MDS log replay
1618 ease stack pressure by using a thread dealing llog_process.
1622 Description: wait until IO finished before start new when do lock cancel.
1623 Details : VM protocol want old IO finished before start new, in this case
1624 need wait until PG_writeback is cleared until check dirty flag and
1625 call writepages in lock cancel callback.
1627 Severity : enhancement
1629 Description: using special macro for print time and cleanup in includes.
1633 Description: mds_mfd_close() ASSERTION(rc == 0)
1634 Details : In mds_mfd_close(), we need protect inode's writecount change
1635 within its orphan write semaphore to prevent possible races.
1639 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1643 Frequency : rare, on shutdown ost
1644 Description: don't hit live lock with umount ost.
1645 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1646 use shrink_dcache_sb instead.
1650 Description: don't panic with use echo client
1651 Details : echo client pass NULL as client nid pointer and this produce null
1652 pointer dereference.
1656 Description: fix build on ppc32
1657 Details : compile code with -m64 flag produce wrong object file for ppc32.
1661 Description: add message levels for liblustreapi
1665 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1666 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1667 will result in an -EAGAIN and in the exit call path we call
1668 alloc_rr() which will with increasing aggressiveness attempt to
1669 aquire precreated objects on the minimum number of required OSCs.
1673 Description: Use old size assignment to avoid deadlock
1674 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1675 the scheduling while holding a spinlock. We do not need locking
1676 for size in ll_update_inode() because size is only updated from
1677 the MDS for directories or files without objects, so there is no
1678 other place to do the update, and concurrent access to such inodes
1679 are protected by the inode lock.
1683 Description: resolve "_IOWR redefined" build error on SLES10
1687 Description: dump the memory debugging after all modules are unloaded to
1688 suppress false negative in conf_sanity test 39
1690 Severity : enhancement
1692 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1697 Description: fix SLES kernel versioning
1698 Details : the kernel version for our SLES 10 kernel did not include a "-"
1699 before the "smp" at the end. while this was not a problem in
1700 general, it did mean that software trying to use the kernel
1701 version to try to detect a vendor specific kernel would fail.
1702 this was most evident by the OFED build scripts.
1706 Description: Don't update lov_desc members until making sure they are valid
1707 Details : When updating lov_desc members via proc fs, need fix their
1708 validities before doing the real update.
1712 Description: don't put request into delay list while invalidate in flight.
1713 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1714 import in flight. this produce timeout for invalidate and sometimes
1715 can cause stale data.
1719 Frequency : on ppc only
1720 Description: not convert ost objects for directory because it's not exist.
1721 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1725 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1727 Severity : enhancement
1729 Description: Update to support 2.6.22.14 vanilla kernel.
1733 Frequency : rare, on recovery
1734 Description: read procfs can produce deadlock in some situation
1735 Details : Holding lprocfs lock which send rpc can produce block for destroy
1736 obd objects and this also block reconnect with -EALREADY. This
1737 isn't fix all lprocfs bugs - but make it rare.
1739 Severity : enhancement
1741 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1744 Frequency : frequent on X2 node
1746 Description: mdc_set_open_replay_data LBUG
1747 Details : Set replay data for requests that are eligible for replay.
1751 Description: lustre_mgs: operation 101 on unconnected MGS
1752 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1753 MGC, and late on MGC cannot successfully connect to MGS and a lot
1754 of the error messages complaining that MGS is not connected.
1758 Frequency : on network error
1759 Description: panic with double free request if network error
1760 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1761 but it's true only for synchronus enqueue, for async enqueue
1762 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1765 Severity : enhancement
1767 Description: client-side metadata stat-ahead during readdir(directory readahead)
1768 Details : perform client-side metadata stat-ahead when the client detects
1769 readdir and sequential stat of dir entries therein
1772 Frequency : on start mds
1774 Description: Implement get_info(last_id) in obdfilter.
1777 Frequency : occasional
1779 Description: Correctly check stale fid, not start epoch if ost not support SOM
1780 Details : open with flag O_CREATE need set old fid in op_fid3 because
1781 op_fid2 was overwritten with new generated fid, but mds can answer
1782 with one of these two fids and both is not stale. Setattr
1783 incorrectly started an epoch and assume will be called
1784 done_writing, but without SOM done_writing ever being called.
1787 Frequency : rare, depends on device drivers and load
1789 Description: MDS or OSS nodes crash due to stack overflow
1790 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1791 In some cases, in conjunction with device drivers that use a lot
1792 of stack the MDS (or possibly OSS) service threads could overflow
1793 the stack. One change which was identified to consume additional
1794 stack has been reworked to avoid the extra stack usage.
1797 Frequency : occasional
1799 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1800 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1801 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1802 oscc_has_objects and see OSC has no objests and call
1803 oscc_internal_create to resend create request.
1805 Severity : enhancement
1807 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1809 Severity : enhancement
1811 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1814 Frequency : very rare
1816 Description: Fix replay if there is an un-replied request and open
1817 Details : In some cases, older replay request will revert the
1818 mcd->mcd_last_xid on MDS which is used to record the client's
1819 latest sent request.
1821 Severity : enhancement
1823 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1825 Severity : enhancement
1827 Description: Add rhel5 support to HEAD.
1829 Serverity : enhancement
1831 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1836 Description: Don't allow skipping OSTs if index has been specified.
1837 Details : Don't allow skipping OSTs if index has been specified, make
1838 locking in internal create lots better.
1842 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1843 Details : access to bitfield in structure is always rounded to long
1844 and this produce problem with not atomic change any bit.
1848 Description: Lustre make rpms failed.
1849 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1850 builds Lustre rpms from tarball.
1853 Frequency : rare on shutdown ost
1855 Description: If llog cancel was not send before clean_exports phase, this can
1856 produce deadlock in llog code.
1857 Details : If llog thread has last reference to obd and call class_import_put
1858 this produce deadlock because llog_cleanup_commit_master wait when
1859 last llog_commit_thread exited, but this never success because was
1860 called from llog_commit_thread.
1864 Description: allow userland application know is lost one of stripes.
1865 Details : fill lvb_blocks with error code on ost and return it to
1866 application if error flag found.
1870 Description: NULL lov_tgts causing MDS oops
1871 Details : more safe checks for NULL lov_tgts for avoid oops.
1873 Severity : enhancement
1875 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1879 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1880 Details : If an OST is down the MDS will hang indefinitely in
1881 obd_statfs_rqset() waiting for the statfs data. While for
1882 MDS QOS usage of statfs, it should not stuck in waiting.
1884 Severity : enhancement
1886 Description: remote_acl support
1887 Details : Support ACL-based permission check for remote user.
1888 Support setfacl/getfacl for remote user with the utils
1889 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1890 the system "{s,g}etfacl" utils.
1892 Severity : enhancement
1894 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1896 Severity : enhancement
1898 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1902 Description: make mgs_setparam() handle fsname containing dash
1903 Details : fsname containing a dash does not work with lctl conf_param
1905 Severity : enhancement
1907 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1909 Severity : enhancement
1911 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1915 Frequency : rare, at shutdown
1916 Description: access already free / zero obd_namespace.
1917 Details : if client_disconnect_export was called without force flag set,
1918 and exist connect request in flight, this can produce access to
1919 NULL pointer (or already free pointer) when connect_interpret
1920 store ocd flags in obd_namespace.
1924 Frequency : only at startup
1925 Description: not alloc memory with spinlock held.
1926 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1927 if any spinlock held.
1929 Severity : enhancement
1931 Description: make lustre randomly fail allocating memory
1932 Details : Make lustre randomly failed allocating memory for testing purpose.
1934 Severity : enhancement
1936 Description: lost problems with lov objid file
1937 Details : Fixes some scability and access to not inited memory problems
1938 in work with lov objdid file.
1943 Description: lfs find does not continue on file error
1944 Details : Continue other files processing when a file/dir is absent.
1948 Description: Inconsistent usage of lustre_pack_reply()
1949 Details : Standardize the usage of lustre_pack_reply() such that it
1950 always generate a CERROR on failure.
1953 Frequency : occasional
1955 Description: MDS hang or stay in waiting lock
1956 Details : If client receive lock with CBPENDING flag ldlm need send lock
1957 cancel as separate rpc, to avoid situation when cancel request
1958 can't processed due all i/o threads stay in wait lock.
1962 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1966 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1970 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1974 Description: Update to RHEL4 latest kernel.
1976 Severity : enhancement
1978 Description: Build SLES10 patchless client fails
1979 Details : The configure was broken by run ./configure with
1980 --with-linux-obj=.... argument for patchless client. When the
1981 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1982 can't search header adequately. Use absolute path such as
1983 -I($LINUX)/include instead.
1987 Description: interrupt oig_wait produce painc on resend.
1988 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1989 private set, and this produce situation when rq_ptlrpcd_data not
1990 copyed to new allocated request and triggered LBUG on assert
1991 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1992 ptlrpcd set if request is changed and can be safety changed to use
1997 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1998 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1999 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
2000 these macros and replace them with their difinitions instead.
2004 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
2005 Details : Update which_patch & target file for SLES10 latest kernel.
2008 Description: Remove client patches from SLES 10 kernel.
2009 Details : This causes SLES 10 clients to behave as patchless clients
2010 even on a Lustre-patched (server) kernel.
2012 Severity : enhancement
2014 Description: self-adjustable client's lru lists
2015 Details : use adaptive algorithm for managing client cached locks lru
2016 lists according to current server load, other client's work
2017 pattern, memory activities, etc. Both, server and client
2018 side namespaces provide number of proc tunables for controlling
2021 Severity : enhancement
2023 Description: light-weight GSS support
2024 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
2025 and reduce performance overhead.
2027 Severity : enhancement
2029 Description: Linux keyring support
2030 Details : Support using service of Linux keyring for Lustre GSS internal
2031 context refresh/cache mechanism.
2035 Description: Fix errors in lfs documentation
2036 Details : Fixes man pages
2040 Description: don't use GFP_* in generic Lustre code.
2041 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2045 Description: obdclass is limited by single OBD_ALLOC(idarray)
2046 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2050 Description: warnings with build patchless client with vanila 2.6.19 and up
2051 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2052 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2056 Frequency : Only for SLES
2057 Description: sanity_quota fail test_1
2058 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2059 file, which makes TSTID[2] inunique.
2063 Frequency : Only for SLES9
2064 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2065 Details : Update target/ChangeLog/which_patch .
2069 Frequency : Only for test_10 in sanity-quota.sh
2070 Description: a bug in quota test code
2071 Details : checking a test flag in a wrong place
2075 Frequency : Only for test_13 in sanity-quota.sh
2076 Description: a bug in quota test script
2077 Details : shouldn't check overfull allocated quota
2081 Description: jbd statistics
2082 Details : Port older jbd statistics patch for sles10
2084 Severity : enhancement
2086 Description: Add printing inode info into message about error in writepage.
2091 Description: reply_lock_interpret crash due to race with it and lock cancel.
2092 Details : Do not replay locks that are being cancelled. Do not reference
2093 locks by their address during replay, just by their handle.
2097 Frequency : When flocks are used.
2098 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2099 Details : Flock locks might destroy just granted lock if it could be merged
2100 with another existing flock, this is done in completion handler,
2101 so teach ldlm_cli_enquque_fini that this is a valid case for
2107 Description: Oops in read and write path when failing to allocate lock.
2108 Details : Check if lock allocation failed and return error back.
2112 Description: lstripe command fails for valid OST index
2113 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2114 instead of lov->desc.ld_active_tgt_count.
2118 Description: sanity.sh failed test 103
2119 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2124 Description: df doesn't work properly if diskfs blocksize != 4K
2125 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2129 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2130 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2132 Severity : enhancement
2134 Description: Adaptive timeouts
2135 Details : RPC timeouts adapt to changing server load and network
2136 conditions to reduce resend attempts and improve recovery time.
2140 Description: llapi_file_create() does not allow some changes
2141 Details : add llapi_file_open() that allows specifying the mode and
2142 open flags, and also returns an open file handle.
2146 Description: merge and cleanup kernel patches.
2147 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2151 Description: Add journal checksum support.(Kernel part)
2152 Details : The journal checksum feature adds two new flags i.e
2153 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2154 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2155 indicates that the commit block contains the checksum for
2156 the blocks described by the descriptor blocks. Now commit
2157 record can be sent to disk without waiting for descriptor
2158 blocks to be written to disk. This behavior is controlled
2159 using JBD2_FEATURE_ASYNC_COMMIT flag.
2163 Description: OSS needs mutliple precreate threads
2164 Details : Add ability to start more than one create thread per OSS.
2168 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2169 Details : execute lfs setstripe on client
2173 Description: mds_obd_create error creating tmp object
2174 Details : When the user sets quota on root, llog will be affected and can't
2175 create files and write files.
2178 Frequency : Always on ia64 patchless client, and possibly others.
2180 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2181 Details : This allows the patchless client to be loaded on architectures
2182 without this export.
2186 Description: RedHat Update kernel for RHEL5
2187 Details : Add the kernel config file for RHEL5.
2191 Description: RedHat Update kernel for RHEL5
2192 Details : Modify the kernel config file more closer RHEL5.
2196 Description: Build failure against Centos5 (RHEL5)
2197 Details : Define PAGE_SIZE when it isn't present.
2201 Description: test_7 on sanity-quota.sh can't be used on separate machines
2202 Details : add facet to handle it
2206 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2207 Details : with NFS, the anon dentry's parent was set to itself in
2208 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2209 corresponding dentry other than use rec->ur_name.
2211 Severity : enhancement
2213 Description: lfs setstripe enhancement
2214 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2218 Description: enable data checksumming by default
2219 Details : enable checksum by default, allow --disable-checksum
2220 configure option and "-o nochecksum" mount option. Checksums
2221 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2225 Description: lustre support for RHEL5
2226 Details : Add support for RHEL5.
2230 Description: Client eviction due to failover config
2231 Details : after a connection loss, the lustre client should attempt to
2232 reconnect to the last active server first before trying the
2233 other potential connections.
2237 Description: when mds and osts use different quota unit(32bit and 64bit),
2238 quota will be released repeatly.
2239 Details : void sending multiple quota reqs to mds, which will keep the
2240 status between the reqs.
2244 Description: osts not allocated evenly to files
2245 Details : change the condition to increase offset_idx
2248 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2250 Description: Data corruption for OSTs that are formatted larger than 2TB
2252 Details : When generating the bio request for lustre file writes the
2253 sector number would overflow a temporary variable before being
2254 used for the IO. The data reads correctly from Lustre (which
2255 will overflow in a similar manner) but other file data or
2256 filesystem metadata may be corrupted in some cases.
2260 Description: Tune the kernel for good SCSI performance.
2261 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2262 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2267 Description: rewrite ext2-derived code in obdclass/uuid.c
2268 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2269 from scratch preserving functionality.
2273 Description: "lfs find -obd UUID" prints directories
2274 Details : "lfs find -obd UUID" will return all directory names instead
2275 of just file names. It is incorrect because the directories
2276 do not reside on the OSTs.
2280 Description: use i_size_read and i_size_write in 2.6 port
2281 Details : replace inode->i_size access with i_size_read/write()
2284 Frequency : when removing large files
2286 Description: scheduling issue during removal of large Lustre files
2287 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2288 It causes scheduling issues when removing large files (17TB in the
2292 Frequency : only with liblustre clients on XT3
2294 Description: evictions taking too long
2295 Details : allow llrd to evict clients directly on OSTs
2298 Frequency : only on ppc
2300 Description: /proc/fs/lustre/devices broken on ppc
2301 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2302 We should make sure the type of 'index' is loff_t and then cast
2303 explicitly as needed below. Do not assign an explicitly cast
2307 Frequency : only for rhel5
2309 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2310 Details : Modify the target file & which_kernel.
2312 Severity : enhancement
2314 Description: omit set fsid for export NFS
2315 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2316 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2320 Frequency : Always, for kernels after 2.6.16
2321 Description: Fix warning idr_remove called for id=.. which is not allocated.
2322 Details : Last kernels save old s_dev before kill super and not allow
2323 to restore from callback - restore it before call kill_anon_super.
2327 Description: buffer overruns could theoretically occur
2328 Details : llapi_semantic_traverse() modifies the "path" argument by
2329 appending values to the end of the origin string, and a
2330 overrun may occur. Adding buffer overrun check in liblustreapi.
2334 Description: Fix error on 'ls .' at the top of the Lustre mount.
2335 Details : Don't revalidate dentry if it is a root dentry.
2339 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2340 Details : Modify vm-tunables-rhel4.patch.
2344 Description: Kernel config for 2.6.18-vanilla.
2345 Details : Modify targets/2.6-vanilla.target.in.
2346 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2347 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2348 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2349 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2353 Description: improve handling recoverable errors
2354 Details : if request processig with error which can be recoverable on server
2355 request should be resend, otherwise page released from cache and
2360 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2361 Details : Modify target file & which_patch.
2362 A flaw was found in the IA32 system call emulation provided
2363 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2364 value could be stored in the %RAX register, which could trigger an
2365 out-of-bounds system call table access. An untrusted local user
2366 could exploit this flaw to run code in the kernel
2367 (ie a root privilege escalation). (CVE-2007-4573).
2371 Description: O_DIRECT bypasses client statistics.
2372 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2373 read_ahead_stats were updated. copied the stats section from
2374 osc_send_oap_rpc() into async_internal().
2378 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2382 Description: change order of libsysio includes
2383 Details : '#include sysio.h' should always come before '#include xtio.h'
2387 Description: handle "serious error: objid * already exists" more gracefully
2388 Details : If LAST_ID value on disk is smaller than the objects existing in
2389 the O/0/d* directories, it indicates disk corruption and causes an
2390 LBUG(). If the object is 0-length, then we should use the existing
2391 object. This will help to avoid a full fsck in most cases.
2393 Severity : enhancement
2395 Description: adapt the lustre_config script to support the upgrade case
2396 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2397 server targets to 1.6 in parallel.
2401 Description: To avoid grant space > avaible space when the disk is almost
2402 full. Without this patch you might see the error "grant XXXX >
2403 available" or some LBUG about grant, when the disk is almost
2405 Details : In filter_check_grant, for non_grant cache write, we should
2406 check the left space by if (*left > ungranted + bytes), instead
2407 of (*left > ungranted), because only we are sure the left space
2408 is enough for another "bytes", then the ungrant space should be
2409 increase. In client, we should update cl_avail_grant only there
2410 is OBD_MD_FLGRANT in the reply.
2414 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2415 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2416 described in CVE-2007-4573. This problem could allow untrusted
2417 local users to gain root access.
2420 Frequency : when using O_DIRECT and quotas
2422 Description: Incorrect file ownership on O_DIRECT output files
2423 Details : block usage reported by 'lfs quota' does not take into account
2424 files that have been written with O_DIRECT.
2429 Description: touch file failed when fs is not full
2430 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2431 otherwise we can get ENOSP while fs is not full.
2435 Description: parallel lock callbacks
2436 Details : Instead of sending blocking and completion callbacks as separated
2437 requests, adding them to a set and sending in parallel.
2440 Frequency : only for Cray XT3
2441 Bugzilla : 12829/13455
2442 Description: Changing primary group doesn't change the group lustre assigns to
2444 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2445 overridden with the primary group provided by the group upcall,
2446 whereas the supplied fsgid can be trusted if it is in the list of
2447 supplementary groups returned by the group upcall.
2449 Severity : enhancement
2451 Description: Allow masking D_WARNING, D_ERROR messages from console
2452 Details : Console messages can now be disabled via lnet.printk.
2456 Description: User code with malformed file open parameter crashes client node
2457 Details : Before packing join_file req, all the related reference should be
2458 checked carefully in case some malformed flags cause fake
2459 join_file req on client.
2463 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2464 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2465 mode is equal to the granted mode. However, it should also check
2466 that LDLM_FL_CP_REQD is not set.
2470 Description: Heavy nfs access might result in deadlocks
2471 Details : After ELC code landed, it is now improper to enqueue any mds
2472 locks under och_sem, because enqueue might want to decide to
2473 cancel open locks for same inode we are holding och_sem for.
2477 Description: Client eviction while running blogbench
2478 Details : A lot of unlink operations with concurrent I/O can lead to a
2479 deadlock causing evictions. To address the problem, the number of
2480 oustanding OST_DESTROY requests is now throttled to
2481 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2482 callbacks are processed in priority.
2486 Description: enable ACLs on MDS by default
2487 Details : ACLs must be enabled on MDS by default.
2490 Frequency : PPC/PPC64 only
2492 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2493 Details : fix duplicated definitions between asm-ppc64/types.h and
2494 lustre_types.h on PPC.
2497 Frequency : PPC/PPC64 only
2499 Description: asm-ppc/segment.h does not exist
2500 Details : fix compile issue on PPC.
2504 Description: better handle error messages in extents code
2507 Frequency : RHEL4 only
2509 Description: mkfs is very slow on IA64/RHEL4
2510 Details : A performance regression has been discovered in the MPT Fusion
2511 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2512 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2513 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2515 Severity : enhancement
2517 Description: SNMP support enhancement
2518 Details : Adding total number of sampled request for an MDS node in snmp
2521 Severity : enhancement
2523 Description: Optimize ldlm waiting list processing for PR extent locks
2524 Details : When processing waiting list for read extent lock and meeting read
2525 lock that is same or wider to it that is not contended, skip
2526 processing rest of the list and immediatelly return current
2527 status of conflictness, since we are guaranteed there are no
2528 conflicting locks in the rest of the list.
2532 Description: Time out and refuse to reconnect
2533 Details : When the failover node is the primary node, it is possible
2534 to have two identical connections in imp_conn_list. We must
2535 compare not conn's pointers but NIDs, otherwise we can defeat
2536 connection throttling.
2540 Description: port llog fixes from b1_6 into HEAD
2541 Details : Port llog reference couting and some llog cleanups from b1_6
2542 (bug 10800) into HEAD, for protect from panic and access to already
2543 free llog structures.
2547 Description: Detect stride IO mode in read-ahead
2548 Details : When a client does stride read, read-ahead should detect that and
2549 read-ahead pages according to the detected stride pattern.
2553 Description: data checksumming impacts single node performance
2554 Details : add support for several checksum algorithm. Currently, only CRC32
2555 and Adler-32 are supported. The checksum type can be changed on
2556 the fly via /proc/fs/lustre/osc/*/checksum_type.
2560 Description: use adler32 for page checksums
2561 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2566 Description: build for x2 fails
2567 Details : fix compile issue on Cray systems.
2571 Description: Properly match for duplicate locks
2572 Details : Due to different lock order from skiplists code, we need to
2573 traverse entire list for now
2576 Frequency : only on PPC/SLES10
2578 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2579 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2580 be defined. Add a hack in mkfs_lustre.c to work around this compile
2585 Description: LASSERT on MDS when client holding flock lock dies
2586 Details : ldlm pool logic depends on number of granted locks equal to
2587 number of released locks which is not true for flock locks, so
2588 just exclude such locks from consideration.
2592 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2593 Details : Use fsfilt_sync() for both the whole filesystem sync and
2594 individual file sync to eliminate dangerous inode locking
2595 with I_LOCK that can lead to a deadlock.
2599 Description: performance in 1.6.3
2600 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2601 sure that 1MB requests are not fragmented by the block layer.
2603 Severity : enhancement
2605 Description: organize the server-side client stats on per-nid basis
2606 Details : Change the structure of stats under obdfilter and mds to
2614 The "uuid"s file would list the uuids of _active_ exports.
2615 And the clear entry is to clear all stats and stale nids.
2617 Severity : enhancement
2619 Description: eliminate client locks in face of contention
2620 Details : file contention detection and lockless i/o implementation
2621 for contended files.
2625 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2628 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2630 Description: MDS may lose file striping (and hence file data) in some cases
2631 Details : If there are additional extended attributes stored on the MDS,
2632 in particular ACLs, SELinux, or user attributes (if user_xattr
2633 is specified for the client mount options) then there is a risk
2634 of attribute loss. Additionally, the Lustre file striping
2635 needs to be larger than default (e.g. striped over all OSTs),
2636 and an additional attribute must be stored initially in the
2637 inode and then increase in size enough to be moved to the
2638 external attribute block (e.g. ACL growing in size) for file
2643 Description: skiplist implementation simplification
2644 Details : skiplists are used to group compatible locks on granted list
2645 that was implemented as tracking first and last lock of each
2646 lock group the patch changes that to using doubly linked lists
2650 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2651 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2652 because LASSERT on both the data supplied by a client, and the
2653 data on disk is dangerous and incorrect.
2655 Severity : enhancement
2657 Description: Slow truncate/writes to huge files at high offsets.
2658 Details : Directly associate cached pages to lock that protect those pages,
2659 this allows us to quickly find what pages to write and remove
2660 once lock callback is received.
2664 Description: more ldlm soft lockups
2665 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2666 starve other threads from the resource lock for a long time in
2667 case of long waiting queue, so change the debug level from
2668 D_OTHER to the less frequently used D_INFO.
2670 Severity : enhancement
2672 Description: add -gid, -group, -uid, -user options to lfs find
2676 Description: Hung threads in invalidate_inode_pages2_range
2677 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2678 once one is completed. As a result, some RPCs are stuck in the
2679 queue and are never sent.
2683 Description: filter threads hungs on waiting journal commit
2684 Details : Cleanup filter group llog code, then only filter group llog will
2685 be only created in the MDS/OST syncing process.
2689 Description: Procfs and llog threads access destoryed import sometimes.
2690 Details : Sync the import destoryed process with procfs and llog threads by
2691 the import refcount and semaphore.
2693 Severity : enhancement
2695 Description: openlock cache of b1_6 port to HEAD
2700 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2701 Details : If insertion of an extent fails, then discard the inode
2702 preallocation and free data blocks else it can lead to duplicate
2707 Description: don't always update ctime in ext3_xattr_set_handle()
2708 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2709 In some cases the ctime should not be updated, for example for
2710 2.0->1.8 compatibility it is necessary to delete an xattr and it
2711 should not update the ctime.
2715 Bugzilla : 15713/16362
2716 Description: Assertion in iopen_connect_dentry in 1.6.3
2717 Details : looking up an inode via iopen with the wrong generation number can
2718 populate the dcache with a disconneced dentry while the inode
2719 number is in the process of being reallocated. This causes an
2720 assertion failure in iopen since the inode's dentry list contains
2721 both a connected and disconnected dentry.
2725 Description: assertion failure in ldlm_handle2lock()
2726 Details : fix a race between class_handle_unhash() and class_handle2object()
2727 introduced in lustre 1.6.5 by bug 13622.
2732 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2733 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2738 Description: oops in sd_iostats_seq_show()
2739 Details : unloading/reloading the scsi low level driver triggers a kernel
2740 bug when trying to access the sd iostat file.
2745 Description: Kernel panics during QLogic driver reload
2746 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2747 patch, causing memory corruption.
2752 Description: journal_dev option does not work in b1_6
2753 Details : pass mount option during pre-mount.
2755 Severity : enhancement
2757 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2758 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2759 extent information of a file. It can be used to map logical blocks
2760 in a file to physical blocks in the block device.
2764 Description: LDLM soft lockups - improvement
2765 Details : It is be possible to send the lock handle along with each read
2766 or write request because the client is already doing a lock match
2767 itself so there isn't any reason the OST should have to re-do that
2771 Frequency : only on Cray X2
2773 Description: X2 build failures
2774 Details : fix build failures on Cray X2.
2778 Description: xid & resent requests
2779 Details : Initialize RPC XID from clock at startup (randomly if clock is
2782 Severity : enhancement
2784 Description: Add lustre_start utility to start or stop multiple Lustre servers
2789 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2790 Details : In case of memory pressure, list_del() can be called twice on
2791 req->rq_history_list, causing a kernel oops.
2795 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2796 Details : fix stack overflow in the distributed lock manager by defering
2797 export eviction after a failed AST to the elt thread instead of
2798 handling it in the dlm interpret routine.
2802 Description: Convert some comments to new format.
2803 Details : Update documenting comments to match doxygen conventions.
2807 Description: Grammar fixes.
2808 Details : A couple of trivial spelling fixes.
2812 Description: OSD_COUNTERS-mandatory
2813 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2814 to simplify the code.
2818 Description: simplify cmm_device freeing logic.
2819 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2823 Description: Add lockdep support to dt_object_operations locking interface.
2824 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2825 parameter indicating lock ordering. Update mdd code to use new
2830 Description: Introduce failloc constants for lockless IO tests.
2831 Details : Add two new failloc constants to test lockless IO. Only one of
2832 them in implemented---another is checked in yet to be landed
2837 Description: Add lockdep support for inode mutex.
2838 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2839 in the situations where Lustre has to lock more than one inode
2844 Description: Add optional invariants checking support.
2845 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2846 configure switch. This macro is to be used for consistency and
2847 sanity checks that are too expensive to be left in `production'
2852 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2853 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2858 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2859 Details : Use spin_lock_nested() in (the only) situation where more than
2860 one ldlm_lock is locked simultaneously. Also, fix possible
2861 dead-lock in ldlm_lock_change_resource() by enforcing particular
2866 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2867 Details : Instead of specifying each ldlm_lock call-back through separate
2868 parameter, wrap them into struct ldlm_callback_suite.
2872 Description: Kill join_lru obd method and its callers.
2873 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2874 in memory---a requirement implemented through ->o_join_lru() obd
2875 method in HEAD. Remove this method and its users.
2879 Description: Add asynchronous ldlm ENQUEUE completion handler.
2880 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2881 case that stock ldlm_completion_ast() cannot handle as it waits
2882 until lock is granted. Introduce new ldlm_completion_ast_async()
2883 for this. Also comment ldlm_completion_ast().
2887 Description: ldlm_error <-> errno conversion.
2888 Details : Add functions to map (rather arbitrary) between LDLM error codes
2889 and standard errno values. CLIO needs this to prevent LDLM specific
2890 constants from escaping ldlm and osc.
2894 Description: Kill unused ldlm_handle2lock_ns() function.
2895 Details : Kill unused ldlm_handle2lock_ns() function.
2899 Description: Add lu_ref support to ldlm_lock
2900 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2901 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2902 and ldlm_resource. LDLM interface has to be changed, because code that
2903 releases a reference on a lock, has to "know" what reference this is.
2904 In the most frequent case
2906 lock = ldlm_handle2lock(handle);
2908 LDLM_LOCK_PUT(lock);
2910 no changes are required. When any other reference (received _not_ from
2911 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2912 instead of LDLM_LOCK_PUT().
2914 Arguably, changes are pervasive, and interface requires some discipline
2915 for proper use. On the other hand, it was very instrumental in finding
2916 a few leaked lock references.
2920 Description: Add ldlm_lock_addref_try().
2921 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2922 attempts to addref a lock that might be being canceled
2927 Description: Add ldlm_weigh_callback().
2928 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2929 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2930 hard-coded `number of pages' logic.
2934 Description: Add lockdep annotations to llog code.
2935 Details : Use appropriately tagged _nested() locking calls in the places
2936 where llog takes more than one ->lgh_lock lock.
2940 Description: Add loi_kms_set().
2941 Details : Wrap kms updates into a helper function.
2945 Description: Constify instances of struct lsm_operations.
2946 Details : Constify instances of struct lsm_operations.
2950 Description: lu_conf support.
2951 Details : On a server, a file system object is uniquely identified
2952 by a fid, which is sufficient to locate and load all object
2953 state (inode). On a client, on the other hand, more data are
2954 necessary instantiate an object. Change lu_object_find() and
2955 friends to take additional `lu_conf' argument describing object.
2956 Typically this includes layout information.
2960 Description: lu_context fixes.
2961 Details : Introduce new lu_context functions that are needed on the client
2962 side, where some system threads (ptlrpcd) are shared by multiple
2963 modules, and so cannot be stopped during module shutdown.
2967 Description: Add start and stop methods to lu_device_type_operations.
2968 Details : Introduce two new methods in lu_device_type_operations, that are
2969 invoked when first instance of a given type is created and last one
2970 is destroyed respectively. This is need by CLIO.
2974 Description: Add lu_ref support to struct lu_device.
2975 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2976 track leaked references.
2980 Description: Introduce lu_kmem_descr.
2981 Details : lu_kmem_descr and its companion interface allow to create
2982 and destroy a number of kmem caches at once.
2986 Description: Fix lu_object finalization race.
2987 Details : Fix a race between lu_object_find() finding an object and its
2988 concurrent finalization. This race is (most likely) not possible
2989 on the server, but might happen on the client.
2993 Description: Introduce lu_ref interface.
2994 Details : lu_ref is a debugging module allowing to track references to
2995 a given object. It is quite cpu expensive, and has to be
2996 explicitly enabled with --enable-lu_ref. See usage description
3001 Description: Factor lu_site procfs stats into a separate function.
3002 Details : Separate lu_site stats printing code into a separate function
3003 that can be reused on a client.
3007 Description: Constify instances of struct {lu,dt,md}_device_operations.
3008 Details : Constify instances of struct {lu,dt,md}_device_operations.
3012 Description: Introduce struct md_site and move meta-data specific parts of
3013 struct lu_site here.
3014 Details : Move md-specific fields out of struct lu_site into special struct
3015 md_site, so that lu_site can be used on a client.
3019 Description: Kill mdd_lov_destroy().
3020 Details : Remove unused mdd code.
3024 Description: Add st_block checking to multistat.c.
3025 Details : Add st_block checking to multistat.c.
3029 Description: Add lu_ref support to struct obd_device.
3030 Details : Add lu_ref tracking to obd_device.
3034 Description: Kill obd_set_fail_loc().
3035 Details : Remove unused code.
3039 Description: Add special type for ptlrpc_request interpret functions.
3040 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
3041 there. Actual usage will be in CLIO.
3045 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
3046 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
3047 doesn't work with lockdep.
3051 Description: Add rwv.c test program.
3052 Details : New testing program exercising readv(2) and writev(2) (Qian).
3056 Description: Add sendfile.c test program.
3057 Details : New testing program exercising sendfile(2) (Jay).
3061 Description: Ratelimit a message that can be very frequent.
3062 Details : Ratelimit a memory allocation failure message that can
3067 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3068 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3072 Description: Liblustre build fixes.
3073 Details : Liblustre build fixes.
3077 Description: libcfs: add cfs_{need,cond}_resched() interface.
3078 Details : libcfs: add cfs_{need,cond}_resched() definition and
3079 implementations for Linux, NT, and liblustre.
3081 Severity : enhancement
3083 Description: More exported tunables for mballoc
3084 Details : Add support for tunable preallocation window and new tunables for
3085 large/small requests
3089 Description: Detect corruption of block bitmap and checking for preallocations
3090 Details : Checks validity of on-disk block bitmap. Also it does better
3091 checking of number of applied preallocations. When corruption is
3092 found, it turns filesystem readonly to prevent further corruptions.
3096 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3097 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3098 window to 0. If it does hit cache miss, and read-pattern is still
3099 stride-io mode, does not reset the stride window, but also does not
3100 increase the stride window length in this case.
3104 Frequency : only for big-endian servers
3105 Description: Check if system is big-endian while mounting fs with extents feature
3106 Details : Mounting a filesystem with extents feature will fail on big-endian
3107 systems since ext3-based ldiskfs is not supported on big-endian
3108 systems. This can be over-riden with "bigendian_extents" mount option.
3110 Severity : enhancement
3112 Description: The root squash functionality
3113 Details : A security feature, which is to prevent users from being able
3114 to mount lustre on their desktop, run as root, and delete
3115 all of the files in the filesystem. The goal is accomplished by
3116 remapping user id (UID) and group id (GID) of the root user to
3117 a UID and GID specified by the system administartor via Lustre
3118 configuration management server (MGS). The functionality also
3119 allows to specify sets of clients for which the remapping does
3124 Description: Excessive recovery window
3125 Details : With AT enabled, the recovery window can be excessively long (6000+
3126 seconds). To address this problem, we no longer use
3127 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3128 timeout no longer depends on the service time, it is set to
3129 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3130 time via pb_service_time.
3134 Description: Watchdog triggered on MDS failover
3135 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3136 the OSTs know that the MDS "UUID" can be reused for the same export
3137 from a different NID, so we do not need to wait for the export to be
3141 Frequency : rare, only if using MMP with Linux RAID
3143 Description: MMP doesn't work with Linux RAID
3144 Details : While using HA for Lustre servers with Linux RAID, it is possible
3145 that MMP will not detect multiple mounts. To make this work we
3146 need to unplug the device queue in RAID when the MMP block is being
3147 written. Also while reading the MMP block, we should read it from
3148 disk and not the cached one.
3150 Severity : enhancement
3152 Description: open file using fid
3153 Details : A file can be opened using just its fid, like
3154 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3157 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3160 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3161 Details : add proper locking for creation of jbd_4k slab cache
3165 Description: MMP check in ext3_remount() fails without displaying any error
3166 Details : When multiple mount protection fails during remount, proper error
3169 Severity : enhancement
3171 Description: Allow stripe size to be up to 4G-64k
3172 Details : Fix math logic to allow large stripe sizes.
3176 Description: add check for >8TB ldiskfs filesystems
3177 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3178 Don't allow >8TB ldiskfs filesystems to be mounted without
3179 force_over_8tb mount option
3184 Descriptoin: OSS complains "too many blocks" and crashes
3185 Details : In filter_commitrw_write():
3186 1. lquota_chkquota();
3187 2. obdo_from_inode();
3188 3. lquota_pending_commit();
3189 "2." has changed uid and gid in this case, then caused
3190 lquota_chkquota() and lquota_pending_commit() to use
3191 different uid and gid. That is the root of the bug.
3195 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3196 Details : Detect and print out the 'stripe count = -1' case properly.
3200 Description: Out or order replies might be lost on replay
3201 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3202 request with tid smaller than one currently being added, add it
3203 to the start, not end of the list.
3205 --------------------------------------------------------------------------------
3207 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3209 * Support for kernels:
3210 2.6.5-7.283 (SLES 9),
3211 2.6.9-55.EL (RHEL 4),
3212 2.6.16.46-0.14 (SLES 10),
3213 2.6.18.8 vanilla (kernel.org)
3214 * Client support for unpatched kernels:
3215 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3216 2.6.16 - 2.6.22 vanilla (kernel.org)
3217 * Due to recently discovered recovery problems, we do not recommend
3218 using patchless RHEL 4 clients with this or any earlier release.
3219 * Recommended e2fsprogs version: 1.39.cfs8
3220 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3221 * Starting with this release, the ldiskfs backing filesystem required
3222 by Lustre is now in its own package, lustre-ldiskfs. This package
3223 should be installed. It is versioned separately from Lustre and
3224 may be released separately in future.
3228 Description: block reactivating mgc import until all deactivates complete
3229 Details : Fix race when failing back MDT/MGS to itself (testing)
3232 Frequency : at statup only
3234 Description: mds_lov_synchronize race leads to various problems
3235 Details : simultaneous MDT->OST connections at startup can cause the
3236 sync to abort, leaving the OSC in a bad state.
3238 Severity : enhancement
3240 Description: add optional extra BUILD_VERSION info
3241 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3242 to override the lustre version.
3245 Frequency : 2.6.18 servers only
3247 Description: ll_kern_mount() doesn't release the module reference
3248 Details : The ldiskfs module reference count never drops down to 0
3249 because ll_kern_mount() doesn't release the module reference.
3254 Description: server LBUG when using old ost_num_threads parameter
3255 Details : Accept the old ost_num_threads parameter but warn that it
3256 is deprecated, and fix an off-by-one error that caused an LBUG.
3261 Description: Transient SCSI error results in persistent IO issue
3262 Details : iobuf->dr_error is not reinitialized to 0 between two
3266 Frequency : sometimes when underlying device returns I/O errors
3268 Description: OSTs not going read-only during write failures
3269 Details : OSTs are not remounted read-only when the journal commit threads
3270 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3271 instead of the ext3 wrappers.
3275 Description: poor connect scaling with increasing client count
3276 Details : Don't run filter_grant_sanity_check for more than 100 exports
3277 to improve scaling for large numbers of clients.
3280 Frequency : SLES10 only
3282 Description: sanity-quota.sh quotacheck failed: rc = -22
3283 Details : Quotas cannot be enabled on SLES10.
3286 Frequency : liblustre clients only
3288 Description: getdirentries does not give error when run on compute nodes
3289 Details : getdirentries does not fail when the size specified as an argument
3290 is too small to contain at least one entry
3292 Severity : enhancement
3294 Description: Add LNET router traceability for debug purposes
3295 Details : If a checksum failure occurs with a router as part of the
3296 IO path, the NID of the last router that forwarded the bulk data
3297 is printed so it can be identified.
3302 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3303 Details : Due to a race condition, liblustre clients were occasionally
3304 evicted incorrectly.
3306 Severity : enhancement
3308 Description: lfs setstripe use optional parameters instead of postional
3311 Severity : enhancement
3313 Description: Nanosecond timestamp support for ldiskfs
3314 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3315 resolution timestamps. There is not yet support for this at
3316 the Lustre filesystem level.
3319 Frequency : during server recovery
3321 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3322 Details : request with rq_no_resend flag not awake l_wait_event if they get
3326 Frequency : nfs export on patchless client
3328 Description: connectathon hang when test nfs export over patchless client
3329 Details : Disconnected dentry cannot be found with lookup, so we do not need
3330 to unhash it or make it invalid
3333 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3334 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3335 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3340 Description: open req refcounting wrong on reconnect
3341 Details : If reconnect happened between getting open reply from server and
3342 call to mdc_set_replay_data in ll_file_open, we will schedule
3343 replay for unreferenced request that we are about to free.
3344 Subsequent close will crash in variety of ways.
3345 Check that request is still eligible for replay in
3346 mdc_set_replay_data().
3351 Description: disable writes to filesystem when reading health_check file
3352 Details : the default for reading the health_check proc file has changed
3353 to NOT do a journal transaction and write to disk, because this
3354 can cause reads of the /proc file to hang and block HA state
3355 checking on a healthy but otherwise heavily loaded system. It
3356 is possible to return to the previous behaviour during configure
3357 with --enable-health-write.
3359 Severity : enhancement
3361 Description: 64-bit inode version
3362 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3363 to the inode. This will be required for version-based recovery.
3368 Description: MDS fails to start if a duplicate client export is detected
3369 Details : in some rare cases it was possible for a client to connect to
3370 an MDS multiple times. Upon recovery the MDS would detect this
3371 and fail during startup. Handle this more gracefully.
3373 Severity : enhancement
3375 Description: Add -o localflock option to simulate old noflock
3377 Details : This will achieve local-only flock/fcntl locks
3383 Description: log_commit_thread vs filter_destroy race leads to crash
3384 Details : Take import reference before releasing llog record semaphore
3389 Description: Wrong request locking in request set processing
3390 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3391 imp_delayed_list, in this place should be used imp_lock.
3394 Frequency : when reconnection
3396 Description: Grant Leak when osc reconnect to OST
3397 Details : When osc reconnect ost, OST(filter) should check whether it
3398 should grant more space to client by comparing fed_grant and
3399 cl_avail_grant, and return the granted space to client instead
3400 of "new granted" space, because client will call osc_init_grant
3401 to update the client grant space info.
3404 Frequency : when client reconnect to OST
3406 Description: Grant Leak when osc do resend and replay bulk write
3407 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3408 bulk write request, because the grant info while be sync between
3409 OSC and OST when reconnect, and we should ignore the grant info
3410 these of resend/replay write req.
3415 Description: Grant space more than available left space sometimes.
3416 Details : When then OST is about to be full, if two bulk writing from
3417 different clients came to OST. Accord the avaliable space of the
3418 OST, the first req should be permitted, and the second one
3419 should be denied by ENOSPC. But if the seconde arrived before
3420 the first one is committed. The OST might wrongly permit second
3421 writing, which will cause grant space > available space.
3424 Frequency : when client is evicted
3426 Description: Grant might be wrongly erased when osc is evicted by OST
3427 Details : when the import is evicted by server, it will fork another
3428 thread ptlrpc_invalidate_import_thread to invalidate the
3429 import, where the grant will be set to 0. While the original
3430 thread will update the grant it got when connecting. So if
3431 the former happened latter, the grant will be wrongly errased
3432 because of this race.
3437 Description: Checking Stale with correct fid
3438 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3439 to check whether it is stale, because sometimes, we want the
3440 enqueue happened anyway, and op_data.fid2 will not be initialized.
3442 Severity : enhancement
3444 Description: update patchless client
3445 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3448 Frequency : only with 2.4 kernel
3450 Description: random memory corruption
3451 Details : size of struct ll_inode_info is to big for union inode.u and this
3452 can be cause of random memory corruption.
3457 Description: Memory leak in recovery
3458 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3459 It should also check obd_fail before fsfilt_start, otherwise if
3460 fsfilt_start return -EROFS,(failover mds during mds recovery).
3461 then the req will return with repmsg->transno = 0 and rc = EROFS.
3462 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3463 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3464 be freed no matter whether fsfilt_commit success or not.
3467 Frequency : only with huge count clients
3469 Description: Prevents from taking the superblock lock in llap_from_page for
3471 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3472 for llap_from_page call in ll_removepage prevents from taking the
3473 superblock lock for a soon died page.
3478 Description: Not check open intent error before release open handle
3479 Details : in some rare cases, the open intent error is not checked before
3480 release open handle, which may cause
3481 ASSERTION(open_req->rq_transno != 0), because it tries to release
3482 the failed open handle.
3487 Description: Set cat log bitmap only after create log success.
3488 Details : in some rare cases, the cat log bitmap is set too early. and it
3489 should be set only after create log success.
3493 Description: Accessing a block bevice can re-enable I/O when Lustre is
3494 tearing down a device.
3495 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3500 Description: service threads may hog cpus when there are a lot of requests
3501 Details : Insert cond_resched to give other threads a chance to use some CPU
3506 Description: the cat log was not initialized in recovery
3507 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3508 unlink log on mds will not be initialized until mds post
3509 recovery. And also in mds post recovery, the unlink log will
3510 initialization will be done asynchronausly, so there will be race
3511 between add unlink log and unlink log initialization.
3515 Description: brw_stats were being printed incorrectly
3516 Details : brw_stats were being printed as log2 but all of them were not
3517 recorded as log2. Also remove some code duplication arising from
3518 filter_tally_{read,write}.
3522 Frequency : rare, only in recovery.
3523 Description: ASSERTION(req->rq_type != LI_POISON) failed
3524 Details : imp_lock should be held while iterating over imp_sending_list for
3525 prevent destroy request after get timeout in ptlrpc_queue_wait.
3529 Description: replay-single.sh test 52 fails
3530 Details : A lock's skiplist need to be cleanup when it being unlinked
3531 from its resource list.
3535 Description: Short directio read returns full requested size rather than
3537 Details : Direct I/O operations should return actual amount of bytes
3538 transferred rather than requested size.
3540 Severity : enhancement
3542 Description: metadata RPC reduction (e.g. for rm performance)
3543 Details : decrease the amount of synchronous RPC between clients and servers
3544 by canceling conflicing lock before the operation on the client
3545 and packing thier handles into the main operation RPC to server.
3547 Severity : enhancement
3549 Description: Async OSC create to avoid the blocking unnecessarily.
3550 Details : If an OST has no remaining object, system will block on the
3551 creation when it needs to create a new object on this OST. Now,
3552 ways use pre-created objects when available, instead of blocking on
3553 an empty osc while others are not empty. If we must block, we
3554 block for the shortest possible period of time.
3558 Description: improve handling recoverable errors
3559 Details : if request processig with error which can be recoverable on server
3560 request should be resend, otherwise page released from cache and
3563 Severity : enhancement
3565 Description: refine locking for avoid write wrong info into lov_objid file
3566 Details : fix possible races with add new target and write/update data in
3569 --------------------------------------------------------------------------------
3571 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3576 Frequency : on some architectures
3578 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3579 Details : Uninitialized flags sometimes cause configuration commands to
3583 Frequency : patchless clients only
3585 Description: missing __iget() symbol export
3586 Details : The __iget() symbol export is missing. To avoid the need for
3587 this on patchless clients the deathrow inode reaper is turned
3588 off, and we depend on the VM to clean up old inodes. This
3589 dependency was during via the fix for bug 12181.
3591 --------------------------------------------------------------------------------
3593 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3595 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3596 INTEROPERATE with older versions automatically. In many cases a
3597 special upgrade step is needed. Please read the
3598 user documentation before upgrading any part of a live system.
3599 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3600 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3601 * WARNING: Lustre configuration and startup changes are required with
3602 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3606 Severity : enhancement
3608 Description: MountConf
3609 Details : Lustre configuration is now managed via mkfs and mount
3610 commands instead of lmc and lconf. New obd types (MGS, MGC)
3611 are added for dynamic configuration management. See
3612 https://mail.clusterfs.com/wikis/lustre/MountConf for
3615 Severity : enhancement
3617 Description: dynamic OST addition
3618 Details : OSTs can now be added to a live filesystem
3620 Severity : enhancement
3622 Description: startup order invariance
3623 Details : MDTs and OSTs can be started in any order. Clients only
3624 require the MDT to complete startup.
3626 Severity : enhancement
3628 Description: parallel, asynchronous orphan cleanup
3629 Details : orphan cleanup is now performed in separate threads for each
3630 OST, allowing parallel non-blocking operation.
3632 Severity : enhancement
3634 Description: optimized stripe assignment
3635 Details : stripe assignments are now made based on ost space available,
3636 ost previous usage, and OSS previous usage, in order to try
3637 to optimize storage space and networking resources.
3639 Severity : enhancement
3641 Description: Permanently set tunables
3642 Details : All writable /proc/fs/lustre tunables can now be permanently
3643 set on a per-server basis, at mkfs time or on a live system.
3645 Severity : enhancement
3647 Description: Lustre message v2
3648 Details : Add lustre message format v2.
3650 Severity : enhancement
3652 Description: client OST exclusion list
3653 Details : Clients can be started with a list of OSTs that should be
3654 declared "inactive" for known non-responsive OSTs.
3658 Description: SPEC SFS validation failure on NFS v2 over lustre.
3659 Details : Changes the blocksize for regular files to be 2x RPC size,
3660 and not depend on stripe size.
3662 Severity : enhancement
3664 Description: Multiple MD RPCs in flight.
3665 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3666 To avoid overly-overloading MDS, introduce a limit on number of
3667 MDS RPCs in flight for a single client and add /proc controls
3668 to adjust this limit.
3670 Severity : enhancement
3672 Description: client read/write statistics
3673 Details : Add client read/write call usage stats for performance
3674 analysis of user processes.
3675 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3676 file access. extents_stats shows chunk size distribution.
3677 extents_stats_per_process show chunk size distribution per
3680 Severity : enhancement
3682 Description: mds statistics
3683 Details : Add detailed mds operations statistics in
3684 /proc/fs/lustre/mds/*/stats.
3688 Description: Failure of copying files with lustre special EAs.
3689 Details : Client side always return success for setxattr call for lustre
3690 special xattr (currently only "trusted.lov").
3695 Description: Refcount LNET uuids
3696 Details : The global LNET uuid list grew linearly with every startup;
3697 refcount repeated list entries instead of always adding to
3700 Severity : enhancement
3702 Description: Dynamic service threads
3703 Details : Within a small range, start extra service threads
3704 automatically when the request queue builds up.
3707 Frequency : mixed-endian client/server environments
3709 Description: mixed-endian crashes
3710 Details : The new msg_v2 system had some failures in mixed-endian
3713 Severity : enhancement
3715 Description: Easy OST removal
3716 Details : OSTs can be permanently deactivated with e.g. 'lctl
3717 conf_param lustre-OST0001.osc.active=0'
3719 Severity : enhancement
3721 Description: MGS proc entries
3722 Details : Added basic proc entries for the MGS showing what filesystems
3725 Severity : enhancement
3727 Description: provide MGS failover
3728 Details : Added config lock reacquisition after MGS server failover.
3730 Severity : enhancement
3732 Description: add Linux 2.4 support
3733 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3737 Description: a large application tries to do I/O to the same resource and dies
3738 in the middle of it.
3739 Details : Check the req->rq_arrival time after the call to
3740 ost_brw_lock_get(), but before we do anything about
3741 processing it & sending the BULK transfer request. This
3742 should help move old stale pending locks off the queue as
3743 quickly as obd_timeout.
3746 Frequency : when an incorrect nid is specified during startup
3748 Description: ptlrpc connect to non-existant node causes kernel crash
3749 Details : LNET can't be re-entered from an event callback, which
3750 happened when we expire a message after the export has been
3751 cleaned up. Instead, hand the zombie cleanup off to another
3754 Severity : enhancement
3756 Description: plain/inodebits lock performance improvement
3757 Details : Grouping plain/inodebits in granted list by their request modes
3758 and bits policy, thus improving the performance of search through
3762 Frequency : only if OST filesystem is corrupted
3764 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3765 Details : for a short time RPCs with bulk IO are in the replay list,
3766 but replay of bulk IOs is unimplemented. If the OST filesystem
3767 is corrupted due to disk cache incoherency and then replay is
3768 started it is possible to trip an assertion. Avoid putting
3769 committed RPCs into the replay list at all to avoid this issue.
3772 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3774 Description: System hang on startup
3775 Details : This bug allowed the liblustre (e.g. catamount) client to
3776 return to the app before handling all startup RPCs. This
3777 could leave the node unresponsive to lustre network traffic
3778 and manifested as a server ptllnd timeout.
3780 Severity : enhancement
3782 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3783 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3784 boolean to control whether to print peer debug info when a
3785 client's RPC times out.
3788 Frequency : only for kernels with patches from Lustre below 1.4.3
3790 Description: Remove old rdonly API
3791 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3794 Frequency : only for devices with external journals
3796 Description: Set external device read-only also
3797 Details : During a commanded failover stop, we set the disk device
3798 read-only while the server shuts down. We now also set any
3799 external journal device read-only at the same time.
3802 Frequency : when upgrading from 1.4 while trying to change parameters
3804 Description: The wrong (new) MDC name was used when setting parameters for
3805 upgraded MDT's. Also allows changing of OSC (and MDC)
3806 parameters if --writeconf is specified at tunefs upgrade time.
3809 Frequency : when setting specific ost indicies
3811 Description: QOS code breaks on skipped indicies
3812 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3813 created with --index need not be sequential.
3815 Severity : enhancement
3817 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3818 Details : The uninit_groups feature works in conjunction with the kernel
3819 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3820 up the pass1 processing of e2fsck. This is a read-only feature
3821 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3822 mount filesystems that have had this feature enabled.
3824 Severity : enhancement
3826 Description: Improve multi-block allocation algorithm to avoid fragmentation
3827 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3828 allocation locality and avoid filesystem fragmentation.
3830 ------------------------------------------------------------------------------
3832 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3834 * Support for kernels:
3835 2.4.21-47.0.1.EL (RHEL 3)
3836 2.6.5-7.283 (SLES 9)
3837 2.6.9-42.0.10.EL (RHEL 4)
3838 2.6.12.6 vanilla (kernel.org)
3839 2.6.16.27-0.9 (SLES 10)
3840 * Recommended e2fsprogs version: 1.39.cfs5
3842 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3846 Frequency : occasional, depends on client load and configuration
3847 Bugzilla : 12181, 12203
3848 Description: data loss for recently-modified files
3850 Details : In some cases it is possible that recently written or created
3851 files may not be written to disk in a timely manner (this should
3852 normally be within 30s unless client IO load is very high).
3853 The problem appears as zero-length files or files that are a
3854 multiple of 1MB in size after a client crash or client eviction
3855 that are missing data at the end of the file.
3857 This problem is more likely to be hit on clients where files are
3858 repeatedly created and unlinked in the same directory, clients
3859 have a large amount of RAM, have many CPUs, the filesystem has
3860 many OSTs, the clients are rebooted frequently, and/or the files
3861 are not accessed by other nodes after being written.
3863 The presence of the problem can be detected by looking at
3864 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3865 smaller than the second (nr_unused) then dirty files will not
3866 be flushed automatically to disk. "sync; sleep 10" should be
3867 run several times on the node before unmounting it to update
3868 Lustre (this is also safe to run on nodes without this problem).
3870 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3871 that can cause this same problem, so customers using that kernel
3872 also need to update the kernel in addition to Lustre. In order
3873 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3875 It is normal that files written just before a client crash (less
3876 than 30s) may not yet have been flushed to disk, even for local
3880 Frequency : frequent on thin XT3 nodes
3882 Description: UUID collision on thin XT3 Linux nodes
3883 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3884 randomly, since we relied on an insufficiently-seeded PRNG.
3889 Description: OSS hangs after "All ost request buffers busy"
3890 Details : A deadlock between quota and journal operations caused OSS
3891 hangs after printing "All ost request buffers busy."
3894 Frequency : always on liblustre builds
3896 Description: Cleanup compiler warnings on liblustre
3899 Frequency : always on liblustre builds on XT3
3901 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3905 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3906 > 32000 subdirectories
3907 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3908 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3909 subdirectory count crosses 32000. This will aid e2fsck to
3910 correctly handle more than 32000 subdirectories.
3913 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3915 Description: System hang on startup
3916 Details : This bug allowed the liblustre (e.g. catamount) client to
3917 return to the app before handling all startup RPCs. This
3918 could leave the node unresponsive to lustre network traffic
3919 and manifested as a server ptllnd timeout.
3921 Severity : enhancement
3923 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3924 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3925 boolean to control whether to print peer debug info when a
3926 client's RPC times out.
3931 Description: make O_SYNC working on 2.6 kernels
3932 Details : 2.6 kernels use different method for mark pages for write,
3933 so need add a code to lustre for O_SYNC work.
3938 Description: Failure to close file and release space on NFS
3939 Details : Put inode details into lock acquired in ll_intent_file_open.
3940 Use mdc_intent_lock in ll_intent_open to properly
3941 detect all kind of errors unhandled by mdc_enqueue.
3946 Description: proc file read during shutdown sometimes raced obd removal,
3948 Details : Add lock to prevent obd access after proc file removal.
3951 Frequency : Only for files larger than 4GB on 32-bit clients.
3953 Description: improperly doing page alignment of locks
3954 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3955 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3960 Description: RPCs being resent when they shouldn't be
3961 Details : Some RPCs that should not be resent are being resent. This
3962 can cause inconsistencies in the RPC state machine. Do not
3963 resend such requests.
3966 Frequency : rare, only with NFS export
3968 Description: Crash on NFS re-export node
3969 Details : under very unusual load conditions an assertion is hit in
3970 ll_intent_file_open()
3973 Frequency : only if OST filesystem is corrupted
3975 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3976 Details : for a short time RPCs with bulk IO are in the replay list,
3977 but replay of bulk IOs is unimplemented. If the OST filesystem
3978 is corrupted due to disk cache incoherency and then replay is
3979 started it is possible to trip an assertion. Avoid putting
3980 committed RPCs into the replay list at all to avoid this issue.
3985 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3986 Details : Large single O_DIRECT read and write calls can fail to allocate
3987 a sufficiently large buffer to process the request. In case of
3988 allocation failure the allocation is retried with a smaller
3989 buffer and broken into smaller requests.
3991 Severity : enhancement
3993 Description: Add -o localflock option to simulate old noflock behaviour.
3994 Details : This will achieve local-only flock/fcntl locks coherentness.
3999 Description: versioning check is incomplete
4000 Details : Checking the version difference of client vs. server, report
4001 error if the gap is too big.
4006 Description: add support PG_writeback bit
4007 Details : add support for PG_writeback bit for Lustre, for more carefull
4008 work with page cache in 2.6 kernel. This also fix some deadlocks
4009 and remove hack for work O_SYNC with 2.6 kernel.
4011 Severity : enhancement
4013 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4014 Details : The uninit_groups feature works in conjunction with the kernel
4015 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4016 up the pass1 processing of e2fsck. This is a read-only feature
4017 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4018 mount filesystems that have had this feature enabled.
4020 Severity : enhancement
4022 Description: Improve multi-block allocation algorithm to avoid fragmentation
4023 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4024 allocation locality and avoid filesystem fragmentation.
4026 ------------------------------------------------------------------------------
4028 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
4030 * Support for kernels:
4031 2.6.9-42.0.3.EL (RHEL 4)
4032 2.6.5-7.276 (SLES 9)
4033 2.4.21-47.0.1.EL (RHEL 3)
4034 2.6.12.6 vanilla (kernel.org)
4035 2.6.16.21-0.8 (SLES10)
4036 * Recommended e2fsprogs version: 1.39.cfs2-0
4038 * The backwards-compatible /proc/sys/portals symlink has been removed
4039 in this release. Before upgrading, please ensure that you change
4040 any configuration scripts or /etc/sysctl.conf files that access
4041 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4042 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4043 in advance of the upgrade on any system running Lustre 1.4.6 or
4044 newer, since /proc/sys/lnet was added in that version.
4045 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4049 Frequency : only when quota is used
4051 Description: avoid scanning export list for quota master
4052 Details : Change the algorithms to avoid scanning export list in order
4053 to improve the efficiency.
4056 Frequency : MDS failover only, very rarely
4058 Description: "went back in time" messages on mds failover
4059 Details : The greatest transno may be lost when the current operation
4060 finishes with an error (transno==0) and the client's last_rcvd
4061 record is over-written. Save the greatest transno in the
4062 mds_last_transno for this case.
4065 Frequency : always for specific kernels and striping counts
4067 Description: client may get "Matching packet too big" without ACL support
4068 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4069 when trying to access files in certain configurations. The
4070 clients should in fact be denied when mounting because they do
4071 not understand ACLs.
4074 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4076 Description: many clients connecting with IO in progress causes connect timeouts
4077 Details : Avoid synchronous journal commits to avoid delays caused by many
4078 clients connecting/disconnecting when bulk IO is in progress.
4079 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4080 OST_IO_PORTAL to avoid delays behind potentially many pending
4084 Frequency : occasionally with multiple writers to a single file
4086 Description: shared writes to file may result in wrong size reported by stat()
4087 Details : Allow growing of kms when extent lock is cancelled
4090 Frequency : always with random mmap IO to multi-striped file
4092 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4093 Details : Only if the hole is at the end of OST object so that kms is too
4094 small. Fix is to increase kms accordingly in ll_nopage.
4097 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4099 Description: writes to a missing object would leak memory on the OST
4100 Details : If there is an inconsistency between the MDS and OST filesystems,
4101 such that the MDS references an object that doesn't exist, writes
4102 to that object will leak memory due to incorrect cleanup in the
4103 error handling path, eventually running out of memory on the OST.
4108 Description: Creating too long symlink causes lustre errors
4109 Details : Check symlink and name lengths before sending requests to MDS.
4112 Frequency : only if flock is enabled (not on by default)
4114 Description: posix locks not released on fd closure on 2.6.9+
4115 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4116 kernels, this caused such locks not to be released on fd close and
4117 then assertions on fs unmount about still used locks.
4120 Frequency : MDS failover only, very rarely
4122 Description: clients may get ASSERTION(granted_lock != NULL)
4123 Details : When request was taking a long time, and a client was resending
4124 a getattr by name lock request. The were multiple lock requests
4125 with the same client lock handle and
4126 mds_getattr_name->fixup_handle_for_resent_request found one of the
4127 lock handles but later failed with ASSERTION(granted_lock != NULL).
4132 Description: handle->h_buffer_credits > 0, assertion failure
4133 Details : h_buffer_credits is zero after truncate, causing assertion
4134 failure. This patch extends the transaction or creates a new
4138 Frequency : NFS re-export or patchless client
4139 Bugzilla : 11179, 10796
4140 Description: Crash on NFS re-export node (__d_move)
4141 Details : We do not want to hash the dentry if we don't have a lock.
4142 But if this dentry is later used in d_move, we'd hit uninitialised
4143 list head d_hash, so we just do this to init d_hash field but
4144 leave dentry unhashed.
4147 Frequency : NFS re-export or patchless client
4149 Description: NFS exports has problem with symbolic link
4150 Details : lustre client didn't properly install dentry when re-exported
4151 to NFS or running patchless client.
4154 Frequency : NFS re-export or patchless client
4156 Description: Various nfs/patchless fixes.
4157 Details : fixes reuse disconected alias for lookup process - this fixes
4158 warning "find_exported_dentry: npd != pd",
4159 fix permission error with open files at nfs.
4160 fix apply umask when do revalidate.
4163 Frequency : occasional
4165 Description: Crash on NFS re-export node
4166 Details : calling clear_page() on the wrong pointer triggered oops in
4167 generic_mapping_read().
4170 Frequency : rarely, using O_DIRECT IO
4172 Description: unaligned directio crashes client with LASSERT
4173 Details : check for unaligned buffers before trying any requests.
4176 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4178 Description: stale data returned from RAID cache
4179 Details : If only a small amount of IO is done to the RAID device before
4180 reading it again it is possible to get stale data from the RAID
4181 cache instead of reading it from disk.
4184 Frequency : always for sles10 kernel
4186 Description: sles10 support
4187 Details : ll_follow_link: compile fixes and using of nd_set_link
4188 under newer kernels.
4191 Frequency : depends on arch, kernel and compiler version, always on sles10
4194 Description: recursive or deep enough symlinks cause stack overflow
4195 Details : getting rid of large stack-allocated variable in
4199 Frequency : depends on hardware
4201 Description: lustre write performance loss in the SLES10 kernel
4202 Details : the performance loss is caused by using of write barriers in the
4203 ext3 code. The SLES10 kernel turns barrier support on by
4204 default. The fix is to undo that change for ldiskfs.
4206 ------------------------------------------------------------------------------
4208 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4210 * Support for kernels:
4211 2.6.9-42.0.3EL (RHEL 4)
4212 2.6.5-7.276 (SLES 9)
4213 2.4.21-47.0.1.EL (RHEL 3)
4214 2.6.12.6 vanilla (kernel.org)
4218 Frequency : quota enabled and large files being deleted
4220 Description: releasing more than 4GB of quota at once hangs OST
4221 Details : If a user deletes more than 4GB of files on a single OST it
4222 will cause the OST to spin in an infinite loop. Release
4223 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4228 Description: statfs data retrieved from /proc may be stale or zero
4229 Details : When reading per-device statfs data from /proc, in the
4230 {kbytes,files}_{total,free,avail} files, it may appear
4231 as zero or be out of date.
4234 Frequency : systems with MD RAID1 external journal devices
4236 Description: lconf's call to blkid is confused by RAID1 journal devices
4237 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4238 of returning all block devices that match the journal UUID.
4241 Frequency : always, for aggregate stripe size over 4GB
4243 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4244 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4245 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4248 Frequency : always if "lfs find" used on a local file/directory
4250 Description: "lfs find" segfaults if used on a local file/directory
4251 Details : The case where a directory component was not specified wasn't
4252 handled correctly. Handle this properly.
4255 Frequency : always on ppc64
4257 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4258 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4259 perform endianness conversion.
4262 Frequency : rarely (truncate to non-zero file size after write under load)
4263 Bugzilla : 10730, 10687
4264 Description: Files padded with zeros to next 4K multiple
4265 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4266 it is possible that files that are truncated to a non-zero size
4267 immediately after being written are filled with zero bytes beyond
4268 the truncated size. No file data is lost.
4270 Severity : enhancement
4272 Description: Allow recovery/failover for liblustre clients.
4273 Details : liblustre clients were unaware of failover configurations until
4276 Severity : enhancement
4278 Description: user file locks should fail when not mounting with flock option
4279 Details : Set up an error-returning stub in ll_file_operations.lock field
4280 to prevent incorrect behaviour when client is mounted without
4281 flock option. Also, set up properly f_op->flock field for
4285 Frequency : always on ia64
4287 Description: "lfs df" loops on printing out MDS statfs information
4288 Details : The obd_ioctl_data was not initialized and in some systems
4289 this caused a failure during the ioctl that did not return
4290 an error. Initialize the struct and return an error on failure.
4293 Frequency : SLES 9 only
4295 Description: Error of copying files with lustre special EAs as root
4296 Details : Client side always return success for setxattr call for lustre
4297 special xattr (currently only "trusted.lov").
4300 Frequency : rarely on clusters with both ia64+i386 clients
4302 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4303 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4304 clients are doing concurrent writes to the same file it is
4305 possible that the ia64 clients may LASSERT because the OST
4306 extent locks are not PAGE_SIZE aligned. Ensure that grown
4307 locks are always aligned on the request boundary.
4310 Frequency : specific use, occasional
4312 Description: Overwriting in use executable truncates on-disk binary image
4313 Details : If one node attempts to overwrite an executable in use by
4314 another node, we now correctly return ETXTBSY instead of
4315 truncating the file.
4317 Severity : enhancement
4319 Description: Async OSC create to avoid the blocking unnecessarily.
4320 Details : If an OST has no remaining object, system will block on the creating
4321 when need to create a new object on this OST. Now, ways use
4322 pre-created objects when available, instead of blocking on an
4323 empty osc while others are not empty. If we must block, we block
4324 for the shortest possible period of time.
4329 Description: chmod on Lustre root is propagated to other clients
4330 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4331 invalid by the follow_mount time.
4334 Frequency : liblustre clients only
4336 Description: Race in 'instant cancel' lock handling could lead to such locks
4337 never to be granted in case of SMP MDS
4338 Details : Do not destroy not yet granted but cbpending locks in
4342 Frequency : replay/resend of open
4344 Description: non null lock assetion failure in mds_intent_policy
4345 Details : Trying to replay/resend lockless open requests resulted in
4346 mds_open() returning 0 with no lock. Now it sets a flag if
4347 a lock is going to be returned.
4349 Severity : enhancement
4351 Description: Checksum enhancements
4352 Details : New checksum enhancements allow for resending RPCs that failed
4355 Severity : enhancement
4357 Description: Tunables on number of dirty pages in cacche
4358 Details : Allow to set limit on number of dirty pages cached.
4363 Description: client crash on unmount - lock still has references
4364 Details : In some error handling cases it was possible to leak a lock
4365 reference on a client while accessing a file. This was not
4366 harmful to the client during operation, but would cause the
4367 client to crash when the filesystem is unmounted.
4370 Frequency : specific case, rare
4372 Description: ETXTBSY on mds though file not in use by client
4373 Details : ETXTBSY is no longer incorrectly returned when attempting to
4374 chmod or chown a directory that the user previously tried to
4375 execute or a currently-executing binary.
4378 Frequency : extremely rare except on liblustre-based clients
4380 Description: Lustre space not freed when files are deleted
4381 Details : Clean up open-unlinked files after client eviction. Previously
4382 the unlink was skipped and the files remained as orphans.
4387 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4388 Details : In some cases it was possible to send a blocking callback to
4389 a client doing a glimpse, even though that client didn't get
4390 a lock granted. When the glimpse lock is cancelled on the OST
4391 the freed lock is left on the waiting list and corrupted the list.
4394 Frequency : all core dumps
4396 Description: Broke core dumps to lustre
4397 Details : Negative dentry may be unhashed if parent does not have UPDATE
4398 lock, but some callers, e.g. do_coredump, expect dentry to be
4399 hashed after successful create, hash it in ll_create_it.
4401 ------------------------------------------------------------------------------
4403 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4405 * Support for kernels:
4406 2.6.9-42.0.2.EL (RHEL 4)
4407 2.6.5-7.276 (SLES 9)
4408 2.4.21-40.EL (RHEL 3)
4409 2.6.12.6 vanilla (kernel.org)
4413 Frequency : always on RHEL 3
4415 Description: Number of open files grows over time
4416 Details : The number of open files grows over time, whether or not
4417 Lustre is started. This was due to a filp leak introduced
4418 by one of our kernel patches.
4420 ------------------------------------------------------------------------------
4422 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4424 * Support for kernels:
4425 2.6.9-42.EL (RHEL 4)
4426 2.6.5-7.276 (SLES 9)
4427 2.4.21-40.EL (RHEL 3)
4428 2.6.12.6 vanilla (kernel.org)
4433 Bugzilla : 5719, 9635, 9792, 9684
4434 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4435 Details : If a server is under heavy load and cannot reply to new
4436 connection requests before the client resends the (re)connect,
4437 the connection handling code can behave badly if two service
4438 threads are concurrently handing separate (re)connections from
4439 the same client. Add better locking to the connection handling
4440 code, and ensure that only a single connection will be processed
4441 for a given client UUID, even if the lock is dropped.
4443 Severity : enhancement
4445 Description: add TCP zero-copy support to kernel
4446 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4447 sends if the hardware supports scatter-gather and checksumming.
4448 This allows socklnd to do client-write and server-read more
4449 efficiently and reduce CPU utilization from skbuf copying.
4452 Frequency : only if NFS exporting from client
4454 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4455 Details : If setting attributes on a file created under NFS that had
4456 never been opened it would be possible to oops the client
4457 if the file had no objects.
4460 Frequency : always for liblustre
4462 Description: liblustre client does MDS+OSTs setattr RPC for each write
4463 Details : When doing a write from a liblustre client, the client
4464 incorrectly issued an RPC to the MDS and each OST the file was
4465 striped over in order to update the timestamps. When writing
4466 with small chunks and many clients this could overwhelm the MDS
4467 with RPCs. In all cases it would slow down the write because
4468 these RPCs are unnecessary.
4470 Severity : enhancement
4472 Description: allow number of MDS service threads to be changed at module load
4473 Details : It is now possible to change the number of MDS service threads
4474 running. Adding "options mds mds_num_threads={N}" to the MDS's
4475 /etc/modprobe.conf will set the number of threads for the next
4476 time Lustre is restarted (assuming the "mds" module is also
4477 reloaded at that time). The default number of threads will
4478 stay the same, 32 for most systems.
4483 Description: OST crash if filesystem is unformatted or corrupt
4484 Details : If an OST is started on a device that has never been formatted
4485 or if the filesystem is corrupt and cannot even mount then the
4486 error handling cleanup routines would dereference a NULL pointer.
4491 Description: NULL pointer deref in llap_from_page.
4492 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4493 incorrect) mapping if the page was truncated/reclaimed while it was
4494 searched for. Check for this condition and skip such pages when
4495 doing readahead. Introduce extra check to llap_from_page() to
4496 verify page->mapping->host is non-NULL (so page is not anonymous).
4499 Frequency : Sometimes when using sys_sendfile
4501 Description: "page not covered by a lock" warnings from ll_readpage
4502 Details : sendfile called ll_readpage without right page locks present.
4503 Now we introduced ll_file_sendfile that does necessary locking
4504 around call to generic_file_sendfile() much like we do in
4508 Frequency : with certain MDS communication failures at client mount time
4510 Description: NULL pointer deref after failed client mount
4511 Details : a client connection request may delayed by the network layer
4512 and not be sent until after the PTLRPC layer has timed out the
4513 request. If the client fails the mount immediately it will try
4514 to clean up before the network times out the request. Add a
4515 reference from the request import to the obd device and delay
4516 the cleanup until the network drops the request.
4519 Frequency : occasionally during client (re)connect
4521 Description: assertion failure during client (re)connect
4522 Details : processing a client connection request may be delayed by the
4523 client or server longer than the client connect timeout. This
4524 causes the client to resend the connection request. If the
4525 original connection request is replied in this interval, the
4526 client may trip an assertion failure in ptlrpc_connect_interpret()
4527 which thought it would be the only running connect process.
4530 Frequency : only with obd_echo servers and clients that are rebooted
4532 Description: kernel BUG accessing uninitialized data structure
4533 Details : When running an obd_echo server it did not start the ping_evictor
4534 thread, and when a client was evicted an uninitialized data
4535 structure was accessed. Start the ping_evictor in the RPC
4536 service startup instead of the OBD startup.
4538 Severity : enhancement
4539 Bugzilla : 10193 (patchless)
4540 Description: Remove dependency on various unexported kernel interfaces.
4541 Details : No longer need reparent_to_init, exit_mm, exit_files,
4542 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4545 Frequency : rare (only users of deprecated and unsupported LDAP config)
4547 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4548 Details : LDAP apparently contains 'lustreName' attributes instead of
4549 'name'. A simple remapping of the name is sufficient.
4552 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4554 Description: waiting_locks_callback trips kernel BUG if client is evicted
4555 Details : Running with the dump_on_timeout debug flag turned on makes
4556 it possible that the waiting_locks_callback() can try to dump
4557 the Lustre kernel debug logs from an interrupt handler. Defer
4558 this log dumping to the expired_lock_main() thread.
4560 Severity : enhancement
4562 Description: Support NFS exporting on 2.6 kernels.
4563 Details : Implement non-rawops metadata methods for NFS server to use without
4564 changing NFS server code.
4567 Frequency : very rare (synthetic metadata workload only)
4569 Description: two racing renames might cause an MDS thread to deadlock
4570 Details : Running the "racer" program may cause one MDS thread to rename
4571 a file from being the source of a rename to being the target of
4572 a rename at exactly the same time that another thread is doing
4573 so, and the second thread has already enqueued these locks after
4574 doing a lookup of the target and is trying to relock them in
4575 order. Ensure that we don't try to re-lock the same resource.
4578 Frequency : only very large systems with liblustre clients
4580 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4581 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4582 all OSTs in parallel. This allows the eviction of stale liblustre
4583 clients to proceed much faster than if they were done in series,
4584 and also offers similar improvements for other set_info RPCs.
4589 Description: excessive CPU usage during initial read phase on client
4590 Details : During the initial read phase on a client, it would agressively
4591 retry readahead on the file, consuming too much CPU and impacting
4592 performance (since 1.4.5.8). Improve the readahead algorithm
4593 to avoid this, and also improve some other common cases (read
4594 of small files in particular, where "small" is files smaller than
4595 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4600 Description: MDS crash when receiving packet with unknown intent.
4601 Details : Do not LBUG in unknown intent case, just return -EFAULT
4603 Severity : enhancement
4604 Bugzilla : 9293, 9385
4605 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4606 Details : Do not serialize getattr (non-intent version) and statfs.
4609 Frequency : occasional, when OST network is overloaded/intermittent
4611 Description: client evicted by OST after bulk IO timeout
4612 Details : If a client sends a bulk IO request (read or write) the OST
4613 may evict the client if it is unresposive to its data GET/PUT
4614 request. This is incorrect if the network is overloaded (takes
4615 too long to transfer the RPC data) or dropped the OST GET/PUT
4616 request. There is no need to evict the client at all, since
4617 the pinger and/or lock callbacks will handle this, and the
4618 client can restart the bulk request.
4621 Frequency : Always when mmapping file with no objects
4623 Description: client crashes when mmapping file with no objects
4624 Details : Check that we actually have objects in a file before doing any
4625 operations on objects in ll_vm_open, ll_vm_close and
4631 Description: Request leak when working with deleted CWD
4632 Details : Introduce advanced request refcount tracking for requests
4633 referenced from lustre intent.
4635 Severity : Enhancement
4637 Description: Cache open file handles on client.
4638 Details : MDS now will return special lock along with openhandle, if
4639 requested and client is allowed to hold openhandle, even if unused,
4640 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4641 closing files for every read/write openration.
4643 Severity : Enhancement
4645 Description: Cache open negative dentries on client when possible.
4646 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4647 negative dentries on lock revocation.
4652 Description: Remounting a client read-only wasn't possible with a zconf mount
4653 Details : It wasn't possible to remount a client read-only with llmount.
4655 Severity : enhancement
4656 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4657 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4658 MPICH is included to add Lustre-specific ADIO interfaces.
4659 This is based closely on the UFS ADIO layer and only differs
4660 in file creation, in order to allow the OST striping to be set.
4661 This is user-contributed code and not supported by CFS.
4666 Description: extended inode attributes (immutable, append-only) work improperly
4667 when 2.4 and 2.6 kernels are used on client/server or vice versa
4668 Details : Introduce kernel-independent values for these flags.
4670 Severity : enhancement
4673 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4674 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4675 granularity. Now, Fractional MB granularity is be supported,
4676 this is very useful for low memory system.
4678 Severity : enhancement
4680 Description: Getattr by fid
4681 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4682 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4683 have locks and avoids getattr rpc requests for it on every stat.
4686 Frequency : Always, for filesystems larger than 2TB
4688 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4689 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4690 filesystems larger than 2TB to be mounted with mballoc enabled.
4693 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4695 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4696 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4697 filesystems larger than 2TB is mounted then the kernel will
4698 silently corrupt the start of the filesystem. CONFIG_LBD is
4699 enabled for all CFS-supported kernels, but the possibility of
4700 this happening with a modified kernel config exists.
4702 Severity : enhancement
4704 Description: add client O_DIRECT support for 2.6 kernels
4705 Details : It is now possible to do O_DIRECT reads and writes to files
4706 in the Lustre client mountpoint on 2.6 kernel clients.
4708 Severity : enhancement
4710 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4711 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4712 parallel, not waiting for response from every OST before sending
4713 a rpc to the next OST.
4718 Description: setattr vs write race when updating file timestamps
4719 Details : Client processes that update a file timestamp into the past
4720 right after writing to the file (e.g. tar) it is possible that
4721 the updated file modification time can be reset to the current
4722 time due to a race between processing the setattr and write RPC.
4724 Severity : enhancement
4726 Description: Bring 'lfs find' closer in line with regular Linux find.
4727 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4728 -print0 options and obtains all the needed info through the lustre
4731 Severity : enhancement
4733 Description: support up to 1024 configured devices on one node
4734 Details : change obd_dev array from statically allocated to dynamically
4735 allocated structs as they are first used to reduce memory usage
4740 Description: Flush dirty partially truncated pages during truncate
4741 Details : Immediatelly flush partially truncated pages in filter_setattr,
4742 this way we completely avoid having any pages in page cache on OST
4743 and can retire ugly workarounds during writes to flush such pages.
4748 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4749 Details : protect inode from truncation within vfs_unlink() context
4750 just take a reference before calling vfs_unlink() and release it
4751 when parent's i_sem is free.
4756 Description: last_id value checked outside lock on OST caused LASSERT failure
4757 Details : If there were multiple MDS->OST object precreate requests in
4758 flight, it was possible that the OST's last object id was checked
4759 outside a lock and incorrectly tripped an assertion. Move checks
4760 inside locks, and discard old precreate requests.
4763 Frequency : always, if extents are used on OSTs
4765 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4766 Details : OSTs using the extents format would not zero the high 16 bits of
4767 the index physical block number. This is not a problem for any
4768 OST filesystems smaller than 16TB, and no kernels support ext3
4769 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4770 new/modified files) and can be fixed for existing filesystems
4771 with e2fsprogs-1.39-cfs1.
4776 Description: import connection selection may be incorrect if timer wraps
4777 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4778 import connections to be ignored if the 32-bit jiffies counter
4779 wraps. Use a 64-bit jiffies counter.
4782 Frequency : very large clusters immediately after boot
4784 Description: LNET request buffers exhausted under heavy short-term load
4785 Details : If a large number of client requests are generated on a service
4786 that has previously never seen so many requests it is possible
4787 that the request buffer growth cannot keep up with the spike in
4788 demand. Instead of dropping incoming requests, they are held in
4789 the LND until the RPC service can accept more requests.
4792 Frequency : Sometimes during replay
4794 Description: Assertion failure in ll_local_open after replay.
4795 Details : If replay happened on an open request reply before we were able
4796 to set replay handler, reply will become not swabbed tripping the
4797 assertion in ll_local_open. Now we set the handler right after
4798 recognising of open request
4801 Frequency : very rare
4803 Description: kernel reports "badness in vsnprintf"
4804 Details : Reading from the "recovery_status" /proc file in small chunks
4805 may cause a negative length in lprocfs_obd_rd_recovery_status()
4806 call to vsnprintf() (which is otherwise harmless). Exit early
4807 if there is no more space in the output buffer.
4809 Severity : enhancement
4811 Description: clear OBD RPC statistics by writing to them
4812 Details : It is now possible to clear the OBD RPC statistics by writing
4813 to the "stats" file.
4818 Description: Client mtime is not the same on different clients after utimes
4819 Details : In some cases, the client was using the utimes() syscall on
4820 a file cached on another node. The clients now validate the
4821 ctime from the MDS + OSTs to determine which one is right.
4826 Description: Inability to activate failout mode
4827 Details : lconf script incorrectly assumed that in pythong string's numeric
4828 value is used in comparisons.
4831 Frequency : always with multiple stripes per file
4833 Description: Inefficient object allocation for mutli-stripe files
4834 Details : When selecting which OSTs to stripe files over, for files with
4835 a stripe count that divides evenly into the number of OSTs,
4836 the MDS is always picking the same starting OST for each file.
4837 Return the OST selection heuristic to the original design.
4842 Description: mount failures may take full timeout to return an error
4843 Details : Under some heavy load conditions it is possible that a
4844 failed mount can wait for the full obd_timeout interval,
4845 possibly several minutes, before reporting an error.
4846 Instead return an error as soon as the status is known.
4848 Frequency : quota enabled and large files being deleted
4850 Description: releasing more than 4GB of quota at once hangs OST
4851 Details : If a user deletes more than 4GB of files on a single OST it
4852 will cause the OST to spin in an infinite loop. Release
4853 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4858 Description: statfs data retrieved from /proc may be stale or zero
4859 Details : When reading per-device statfs data from /proc, in the
4860 {kbytes,files}_{total,free,avail} files, it may appear
4861 as zero or be out of date.
4864 Frequency : systems with MD RAID1 external journal devices
4866 Description: lconf's call to blkid is confused by RAID1 journal devices
4867 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4868 of returning all block devices that match the journal UUID.
4871 Frequency : always, for aggregate stripe size over 4GB
4873 Description: assertion fails when trying to use 4GB stripe size
4874 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4875 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4878 Frequency : always on ppc64
4880 Description: the first write on an ext3 filesystem with mballoc got stuck
4881 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4882 perform endianness conversion.
4884 ------------------------------------------------------------------------------
4886 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4888 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4889 INTEROPERATE with older versions automatically. Please read the
4890 user documentation before upgrading any part of a live system.
4891 * WARNING: Lustre networking configuration changes are required with
4892 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4895 * Support for kernels:
4896 2.6.9-22.0.2.EL (RHEL 4)
4897 2.6.5-7.244 (SLES 9)
4898 2.6.12.6 vanilla (kernel.org)
4901 Severity : enhancement
4902 Bugzilla : 7981/8208
4903 Description: Introduced Lustre Networking (LNET)
4904 Details : LNET is new networking infrastructure for Lustre, it includes
4905 a reorganized network configuration mode (see the user
4906 documentation for full details) as well as support for routing
4907 between different network fabrics. Lustre Networking Devices
4908 (LNDs) for the supported network fabrics have also been
4909 created for this new infrastructure.
4911 Severity : enhancement
4912 Description: Introduced Access control lists
4913 Details : clients can set ACLs on files and directories in order to have
4914 more fine-grained permissions than the standard Unix UGO+RWX.
4915 The MDS must be started with the "-o acl" mount option.
4917 Severity : enhancement
4918 Description: Introduced filesystem quotas
4919 Details : Administrators may now establish per-user quotas on the
4922 Severity : enhancement
4924 Description: Configuration change for the XT3
4925 The PTLLND is now used to run Lustre over Portals on the XT3
4926 The configure option(s) --with-cray-portals are no longer used.
4927 Rather --with-portals=<path-to-portals-includes> is used to
4928 enable building on the XT3. In addition to enable XT3 specific
4929 features the option --enable-cray-xt3 must be used.
4934 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4935 Details : A race between allocating a new llcd and re-getting the llcd_lock
4936 allowed another thread to grab newly-allocated llcd.
4938 Severity : enhancement
4940 Description: 2.6 OST async journal commit and locking fix to improve performance
4941 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4942 2.6 kernels are now async as they already were in 2.4 kernels so
4943 that they can commit concurrently with the network bulk transfer.
4944 For block-allocated files the filter allocation semaphore is held
4945 to avoid filesystem fragmentation during allocation. BKL lock
4946 removed for 2.6 xattr operations where it is no longer needed.
4951 Description: lconf incorrectly determined whether two IP networks could talk
4952 Details : In some more complicated routing and multiple-network
4953 configurations, lconf will avoid trying to make a network
4954 connection to a disjoint part of the IP space. It was doing the
4955 math incorrectly for one set of cases.
4960 Description: Fix for potential infinite loop processing records in an llog.
4961 Details : If an llog record is corrupted/zeroed, it is possible to loop
4962 forever in llog_process(). Validate the llog record length
4963 and skip the remainder of the block on error.
4966 Frequency : occasional (liblustre only)
4968 Description: liblustre could not open files whose last component is a symlink
4969 Details : sysio_path_walk() would incorrectly pass the open intent to
4970 intermediate path components.
4973 Frequency : rare (liblustre only with non-standard tuning)
4974 Bugzilla : 7201 (7350)
4975 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4976 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4977 after it has been released, only do this for referenced locks
4979 Severity : enhancement
4981 Description: specifying an (invalid) directory default stripe_size of -1
4982 would reset the directory default striping
4983 Details : stripe_size -1 was used internally to signal directory stripe
4984 removal, now use "all default" to signal dir stripe removal
4985 as a directory striping of "all default" is not useful
4988 Frequency : common for large clusters running liblustre clients
4990 Description: doing an ls when liblustre clients are running is slow
4991 Details : sending a glimpse AST to a liblustre client waits for every AST
4992 to time out, as liblustre clients will not respond. Since they
4993 cannot cache data we refresh the OST lock LVB from disk instead.
4995 Severity : enhancement
4997 Description: doing an ls at the same time as file IO can be slow
4998 Details : enqueue and other "small" requests can be blocked behind many
4999 large IO requests. Create a new OST IO portal for non-IO
5000 requests so they can be processed faster.
5003 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
5005 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
5006 created file with mode 000
5007 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
5008 allow an ftruncate() as a non-root user to a file with mode 000.
5009 Lustre now respects this flag to disable mode checks when
5010 truncating a file owned by the user
5013 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
5016 Description: Revoking locks from clients that went dead or catatonic might take
5018 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
5019 cancellation of such locks instant on servers without waiting for
5020 any reply from clients. Clients drops these locks when cancel
5021 notification from server is received without replying.
5024 Frequency : liblustre-only, when liblustre client dies or becomes busy
5026 Description: Doing ls on Linux clients can take a long time with active
5028 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
5029 granting such locks to it in the first place if possible. Locks
5030 are taken by proxy on the OST during the read or write and
5031 dropped immediately afterward. Add connect flags handling, do
5032 not grant locks to liblustre clients for glimpse ASTs.
5034 Severity : enhancement
5036 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
5037 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5038 sequential reads and behaves badly with applications doing random
5039 reads. Improve it by reading ahead at least read region, and
5040 avoiding excessive large RPC for small reads.
5042 Severity : enhancement
5044 Description: Creating more than 1000 files for a single job may cause a load
5045 imbalance on the OSTs if there are also a large number of OSTs.
5046 Details : qos_prep_create() uses an OST index reseed value that is an
5047 even multiple of the number of available OSTs so that if the
5048 reseed happens in the middle of the object allocation it will
5049 still utilize the OSTs as uniformly as possible.
5054 Description: OST or MDS may oops in ping_evictor_main()
5055 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5056 but doesn't restart at beginning of obd_exports_timed list
5059 Severity : enhancement
5061 Description: improve by-nid export eviction on the MDS and OST
5062 Details : allow multiple exports with the same NID to be evicted at one
5063 time without re-searching the exports list.
5066 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5068 Description: MDS may oops in groups_free()
5069 Details : in rare race conditions a newly allocated group_info struct is
5070 freed again, and this can be NULL. The 2.4 compatibility code
5071 for groups_free() checked for a NULL pointer, but 2.6 did not.
5074 Frequency : common for liblustre clients doing little filesystem IO
5075 Bugzilla : 9352, 7313
5076 Description: server may evict liblustre clients accessing contended locks
5077 Details : if a client is granted a lock or receives a completion AST
5078 with a blocking AST already set it would not reply to the AST
5079 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5082 Frequency : lfs setstripe, only systems with more than 160 OSTs
5084 Description: unable to set striping with a starting offset beyond OST 160
5085 Details : llapi_create_file() incorrectly limited the starting stripe
5086 index to the maximum single-file stripe count.
5089 Frequency : LDAP users only
5091 Description: lconf did not handle in-kernel recovery with LDAP properly
5092 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5094 Severity : enhancement
5096 Description: bind OST threads to NUMA nodes to improve performance
5097 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5098 node and do their allocations there to localize memory access
5100 Severity : enhancement
5102 Description: llmount can determine client NID directly from Myrinet (GM)
5103 Details : the client NID code from gmnalnid was moved directly into
5104 llmount, removing the need to use this or specifying the
5105 client NID explicitly when mounting GM clients with zeroconf
5108 Frequency : if client is started with down MDS
5110 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5111 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5112 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5113 ptlrpc_import_delayed_req() because we didn't check if the
5114 request was interrupted, and we also didn't break out of the
5115 event loop if there was no timeout
5120 Description: data loss during non-page-aligned writes to a single file from
5121 both multiple nodes and multiple threads on one node at same time
5122 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5123 inconsistency led to false short-reads, that were cached and later
5124 used by ->prepare_write() to fill in partially written page,
5125 leading to data loss.
5128 Frequency : always, if lconf --abort_recovery used
5130 Description: lconf --abort_recovery fails with 'Operation not supported'
5131 Details : lconf was attempting to abort recovery on the MDT device and not
5134 Severity : enhancement
5136 Description: remove cleanup logs
5137 Details : replace lconf-generated cleanup logs with lustre internal
5138 cleanup routines. Eliminates the need for client-cleanup and
5141 Severity : enhancement
5143 Description: add support for EAs (user and system) on lustre filesystems
5144 Details : it is now possible to store extended attributes in the Lustre
5145 client filesystem, and with the user_xattr mount option it
5146 is possible to allow users to store EAs on their files also
5148 Severity : enhancement
5150 Description: Add possibility (config option) to show minimal available OST free
5152 Details : When compiled with --enable-mindf configure option, statfs(2)
5153 (and so, df) will return least minimal free space available from
5154 all OSTs as amount of free space on FS, instead of summary of
5155 free spaces of all OSTs.
5157 Severity : enhancement
5159 Description: do not expand extent locks acquired on OST-side
5160 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5161 by server: they are not cached anyway.
5164 Frequency : when mmap is used/binaries executed from Lustre
5166 Description: Unmmap pages before throwing them away from read cache.
5167 Details : llap_shrink cache now attempts to unmap pages before discarding
5168 them (if unmapping failed - do not discard). SLES9 kernel has
5169 extra checks that trigger if this unmapping is not done first.
5174 Description: lconf didn't resolve symlinks before checking to see whether a
5175 given mountpoint was already in use
5178 Frequency : when migrating failover services
5179 Bugzilla : 6395, 9514
5180 Description: When migrating a subset of services from a node (e.g. failback
5181 from a failover service node) the remaining services would
5182 time out and evict clients.
5183 Details : lconf --force (implied by --failover) sets the global obd_timeout
5184 to 5 seconds in order to quickly disconnect, but this caused
5185 other RPCs to time out too quickly. Do not change the global
5186 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5188 Severity : enhancement
5189 Frequency : if MDS is started with down OST
5190 Bugzilla : 9439,5706
5191 Description: Allow startup/shutdown of an MDS without depending on the
5192 availability of the OSTs.
5193 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5194 Add appropriate locking and lov-osc refcounts for safe
5195 cleaning. Add osc abort_inflight calls in case the
5196 synchronize never started.
5199 Frequency : occasional (Cray XT3 only)
5201 Description: root not authorized to access files in CRAY_PORTALS environment
5202 Details : The client process capabilities were not honoured on the MDS in
5203 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5204 been accessed by an authorized user then root was able to access
5205 the file on the local client also. The root user capabilities
5206 are now allowed on the MDS, as this environment has secure UID.
5209 Frequency : occasional
5211 Description: ldiskfs "too long searching" message happens too often
5212 Details : A debugging message (otherwise harmless) prints too often on
5213 the OST console. This has been reduced to only happen when
5214 there are fragmentation problems on the filesystem.
5219 Description: Division by zero in statfs when all OSCs are inactive
5220 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5221 lov_statfs divides by value returned by lov_get_stripecnt().
5225 Bugzilla : 9489, 3273
5226 Description: First write from each client to each OST was only 4kB in size,
5227 to initialize client writeback cache, which caused sub-optimal
5228 RPCs and poor layout on disk for the first written file.
5229 Details : Clients now request an initial cache grant at (re)connect time
5230 and so that they can start streaming writes to the cache right
5231 away and always do full-sized RPCs if there is enough data.
5232 If the OST is rebooted the client also re-establishes its grant
5233 so that client cached writes will be honoured under the grant.
5238 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5239 Details : Now I/O RPCs go to different portal number and (presumably) fast
5240 lock requests (and glimses) and other RPCs get their own service
5241 threads pool that should be able to service those RPCs
5244 Severity : enhancement
5246 Description: Ability to exchange lustre version between client and servers and
5247 issue warnings at client side if client is too old. Also for
5248 liblustre clients there is ability to refuse connection of too old
5250 Details : New 'version' field is added to connect data structure that is
5251 filled with version info. That info is later checked by server and
5255 Frequency : rare, liblustre only.
5256 Bugzilla : 9296, 9581
5257 Description: Two simultaneous writes from liblustre at offset within same page
5258 might proceed at the same time overwriting eachother with stale
5260 Details : I/O lock withing llu_file_prwv was released too early, before data
5261 actually was hitting the wire. Extended lock-holding time until
5262 server acknowledges receiving data.
5265 Frequency : extremely rare. Never observed in practice.
5267 Description: avoid generating lustre_handle cookie of 0.
5268 Details : class_handle_hash() generates handle cookies by incrementing
5269 global counter, and can hit 0 occasionaly (this is unlikely, but
5270 not impossible, because initial value of cookie counter is
5271 selected randonly). Value of 0 is used as a sentinel meaning
5272 "unassigned handle" --- avoid it. Also coalesce two critical
5273 sections in this function into one.
5275 Severity : enhancement
5277 Description: allow liblustre clients to delegate truncate locking to OST
5278 Details : To avoid overhead of locking, liblustre client instructs OST to
5279 take extent lock in ost_punch() on client's behalf. New connection
5280 flag is added to handle backward compatibility.
5282 Severity : enhancement
5283 Bugzilla : 4928, 7341, 9758
5284 Description: allow number of OST service threads to be specified
5285 Details : a module parameter allows the number of OST service threads
5286 to be specified via "options ost ost_num_threads={N}" in the
5287 OSS's /etc/modules.conf or /etc/modprobe.conf.
5291 Bugzilla : 6146, 9635, 9895
5292 Description: servers crash with bad pointer in target_handle_connect()
5293 Details : In rare cases when a client is reconnecting it was possible that
5294 the connection request was the last reference for that export.
5295 We would temporarily drop the export reference and get a new
5296 one, but this may have been the last reference and the export
5297 was just destroyed. Get new reference before dropping old one.
5299 Severity : enhancement
5300 Frequency : if client is started with failover MDS
5302 Description: Allow multiple MDS hostnames in the mount command
5303 Details : Try to read the configuration from all specified MDS
5304 hostnames during a client mount in case the "primary"
5307 Severity : enhancement
5309 Description: Stop sending data to evicted clients as soon as possible.
5310 Details : Check if the client we are about to send or are sending data to
5311 was evicted already. (Check is done every second of waiting,
5312 for which l_wait_event interface was extended to allow checking
5313 of exit condition at specified intervals).
5316 Frequency : rare, normally only when NFS exporting is done from client
5318 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5320 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5323 Frequency : always, if srand() is called before liblustre initialization
5325 Description: Liblustre uses system PRNG disturbing its usage by user application
5326 Details : Introduce internal to lustre fast and high-quality PRNG for
5327 lustre usage and make liblustre and some other places in generic
5328 lustre code to use it.
5330 Severity : enhancement
5331 Bugzilla : 9477, 9557, 9870
5332 Description: Verify that the MDS configuration logs are updated when xml is
5333 Details : Check if the .xml configuration logs are newer than the config
5334 logs stored on the MDS and report an error if this is the case.
5335 Request --write-conf, or allow starting with --old_conf.
5337 Severity : enhancement
5339 Description: Handle symlinks in the path when checking if Lustre is mounted.
5340 Details : Resolve intermediate symlinks when checking if a client has
5341 mounted a filesystem to avoid duplicate client mounts.
5346 Description: lconf can hit an error exception but still return success.
5347 Details : The lconf command catches the Command error exception at the top
5348 level script context and will exit with the associated exit
5349 status, but doesn't ensure that this exit status is non-zero.
5354 Description: failure of ptlrpc thread startup can cause oops
5355 Details : Starting a ptlrpc service thread can fail if there are a large
5356 number of threads or the server memory is very fragmented.
5357 Handle this without oopsing.
5360 Frequency : always, only if liblustre and non-default acceptor port was used
5362 Description: liblustre cannot connect to servers with non-default acceptor port
5363 Details : tcpnal_set_default_params() was not called and was therefore
5364 ignoring the environment varaible TCPNAL_PORT, as well as other
5365 TCPNAL_ environment variables
5370 Description: two objects could be created on the same OST for a single file
5371 Details : If an OST is down, in some cases it was possible to create two
5372 objects on a single OST for a single file. No problems other
5373 than potential performance impact and spurious error messages.
5377 Bugzilla : 5681, 9562
5378 Description: Client may oops in ll_unhash_aliases
5379 Details : Client dcache may become inconsistent in race condition.
5380 In some cases "getcwd" can fail if the current directory is
5386 Description: Inode refcounting problems in NFS export code
5387 Details : link_raw functions used to call d_instantiate without obtaining
5388 extra inode reference first.
5392 Bugzilla : 9942, 9903
5393 Description: Referencing freed requests leading to crash, memleaks with NFS.
5394 Details : We used to require that call to ll_revalidate_it was always
5395 followed by ll_lookup_it. Also with revalidate_special() it is
5396 possible to call ll_revalidate_it() twice for the same dentry
5397 even if first occurence returned success. This fix changes semantic
5398 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5399 reference on a request referred from the intent.
5400 ll_intent_release() then releases such a request.
5403 Frequency : rare, normally benchmark loads only
5405 Description: unlinked inodes were kept in memory on the client
5406 Details : If a client is repeatedly creating and unlinking files it
5407 can accumulate a lot of stale inodes in the inode slab cache.
5408 If there is no other client load running this can cause the
5409 client node to run out of memory. Instead flush old inodes
5410 from client cache that have the same inode number as a new inode.
5413 Frequency : SLES9 2.6.5 kernel and long filenames only
5414 Bugzilla : 9969, 10379
5415 Description: utime reports stale NFS file handle
5416 Details : SLES9 uses out-of-dentry names in some cases, which confused
5417 the lustre dentry revalidation. Change it to always use the
5421 Frequency : rare, unless heavy write-truncate concurrency is continuous
5422 Bugzilla : 4180, 6984, 7171, 9963, 9331
5423 Description: OST becomes very slow and/or deadlocked during object unlink
5424 Details : filter_destroy() was holding onto the parent directory lock
5425 while truncating+unlinking objects. For very large objects this
5426 may block other threads for a long time and slow overall OST
5427 responsiveness. It may also be possible to get a lock ordering
5428 deadlock in this case, or run out of journal credits because of
5429 the combined truncate+unlink. Solution is to do object truncate
5430 first in one transaction without parent lock, and then do the
5431 final unlink in a new transaction with the parent lock. This
5432 reduces the lock hold time dramatically.
5435 Frequency : rare, 2.4 kernels only
5437 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5438 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5439 before unmounting their filesystem in order to avoid deadlock.
5440 The kernel_locked() function in 2.4 kernels only checks whether
5441 the kernel lock is held, not whether it is this process that is
5442 holding it as 2.6 kernels do.
5447 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5448 Details : The client ptlrpc code may be trying to reconnect to a down
5449 server before a previous connection attempt has timed out.
5450 Increase the reconnect interval to be longer than the connection
5451 timeout interval to avoid sending duplicate connections to
5455 Frequency : echo_client brw_test command
5457 Description: fix echo_client to work with OST preallocated code
5458 Details : OST preallocation code (5137) didn't take echo_client IO path
5459 into account: echo_client calls filter methods outside of any
5460 OST thread and, hence, there is no per-thread preallocated
5461 pages and buffers to use. Solution: hijack pga pages for IO. As
5462 a byproduct, this avoids unnecessary data copying.
5466 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5467 Description: Client can oops in mdc_commit_close() after open replay
5468 Details : It was possible for the MDS to return an open request with no
5469 transaction number in mds_finish_transno() if the client was
5470 evicted, but without actually returning an error. Clients
5471 would later try to replay that open and may trip an assertion
5472 Simplify the client close codepath, and always return an error
5473 from the MDS in case the open is not successful.
5476 Frequency : rare, 2.6 OSTs only
5478 Description: OST may deadlock under high load on fragmented files
5479 Details : If there was a heavy load and highly-fragmented OST filesystems
5480 it was possible to have all the OST threads deadlock waiting on
5481 allocation of biovecs, because the biovecs were not released
5482 until the entire RPC IO was completed. Instead, release biovecs
5483 as soon as they are complete to ensure forward IO progress.
5485 Severity : enhancement
5487 Description: Support for specifying external journal device at mount
5488 Details : If an OST or MDS device is formatted with an external journal
5489 device, this device major/minor is stored in the ext3 superblock
5490 and may not be valid for failover. Allow detecting and
5491 specifying the external journal at mount time.
5496 Description: Mounting an MDS with pending unlinked files may cause oops
5497 Details : target_finish_recovery() calls mds_postrecov() which returned
5498 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5499 considers this an error and immediately begins cleaning up the
5500 lov, just after starting the mds_lov process
5502 Severity : enhancement
5504 Description: Implement 'lfs df' to report actual free space on per-OST basis
5505 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5506 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5507 usage of objects; '-h' to report in human readable format.
5509 ------------------------------------------------------------------------------
5511 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5518 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5519 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5520 at startup during orphan unlinking. Instead, skip buddy update
5521 before setup, as it will be regenerated after recovery is complete.
5526 Description: If an OST is inactive, its locks might reference stale inodes.
5527 Details : lov_change_cbdata() must iterate over all namespaces, even if
5528 they are inactive to clear inode references from the lock.
5530 Severity : enhancement
5531 Frequency : occasional, if non-standard max_dirty_mb used
5533 Description: Client will block write RPCs if not enough grant
5534 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5535 then the client will block writes while waiting for another RPC
5536 to complete instead of consuming its dirty limit. With change
5537 we get improved performance when max_dirty_mb is small.
5539 Severity : enhancement
5540 Bugzilla : 3389, 6253
5541 Description: Add support for supplementary groups on the MDS.
5542 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5543 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5544 lookups for user supplementary groups into a cache.
5548 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5549 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5550 for liblustre O_EXCL is enforced by MDS.
5554 Description: readdir never returns NULL in liblustre.
5555 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5556 dentry in struct dirent.
5560 Frequency : liblustre only, depends on application IO pattern
5561 Description: liblustre clients evicted if not contacting servers
5562 Details : Don't put liblustre clients into the ping_evictor list, so
5563 they will not be evicted by the pinger ever.
5565 Severity : enhancement
5567 Description: Add ability to evict clients by NID from MDS.
5568 Details : By echoing "nid:$NID" string into
5569 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5570 $NID would be instantly evicted from this MDS and from all active
5571 OSTs connected to it.
5575 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5576 Details : lookup_it_finish() used to query file size from OSTs that was not
5581 Description: service threads change working directory to that of init
5582 Details : Starting lustre service threads may pin the working directory
5583 of the parent thread, making that filesystem busy. Threads
5584 now change to the working directory of init to avoid this.
5588 Frequency : during shutdown only
5589 Description: shutdown with a failed MDS or OST can cause unmount to hang
5590 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5595 Frequency : 2.6 only
5596 Description: chmod/chown may include an extra supplementary group
5597 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5598 supplementary group and if none is specified this is used.
5601 Bugzilla : 5479 (6816)
5603 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5604 Details : If lookup is in progress on a file that is unlinked we might try
5605 to revalidate the inode and fail in revalidate after lookup is
5606 complete and ll_file_open() enqueues the open again but
5607 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5610 Frequency : always, if lconf --abort_recovery used
5612 Description: lconf --abort_recovery fails with 'Operation not supported'
5613 Details : lconf was attempting to abort recovery on the MDT device and not
5616 ------------------------------------------------------------------------------
5618 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5623 Frequency : rare (only unsupported configurations with a node running as an
5625 Bugzilla : 6514, 5137
5626 Description: Mounting a Lustre file system on a node running as an OST could
5628 Details : OSTs now preallocates memory needed to write out data at
5629 startup, instead of when needed, to avoid having to
5630 allocate memory in possibly low memory situations.
5631 Specifically, if the file system is mounted on on OST,
5632 memory pressure could force it to try to write out data,
5633 which it needed to allocate memory to do. Due to the low
5634 memory, it would be unable to do so and the node would
5635 become unresponsive.
5637 Severity : enhancement
5639 Description: Addition of lconf --service command line option
5640 Details : lconf now accepts a '--service <arg>' option, which is
5641 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5643 Severity : enhancement
5645 Description: Failover mode is now the default for OSTs.
5646 Details : By default, OSTs will now run in failover mode. To return to
5647 the old behaviour, add '--failout' to the lmc line for OSTs.
5649 Severity : enhancement
5651 Description: Health checks are now provided for MDS and OSTs
5652 Details : Additional detailed health check information on MSD and OSTs
5653 is now provided through the procfs health_check value.
5656 Frequency : occasional, depends on IO load
5658 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5659 after numerous create/delete cycles
5660 Details : The ext3 inode allocation policy would not allocate new inodes
5661 very well on the OSTs because there are no new directories
5662 being created. Instead we look for groups with free space if
5663 the parent directories are nearly full.
5668 Description: Network or server problems during mount may cause partially
5669 mounted clients instead of returning an error.
5670 Details : The config llog parsing code may overwrite the error return
5671 code during mount error handling, returning success instead
5677 Description: MDS can fail to allocate large reply buffers
5678 Details : After long uptimes the MDS can fail to allocate large reply
5679 buffers (e.g. zconf client mount config records) due to memory
5680 fragmentation or consumption by the buffer cache. Preallocate
5681 some large reply buffers so that these replies can be sent even
5682 under memory pressure.
5686 Frequency : rare (liblustre)
5687 Description: fsx running with liblustre complained that using truncate() to
5688 extend the file doesn't work. This patch corrects that issue.
5689 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5690 ATTR_SIZE and lsm use in llu_setattr_raw.
5694 Frequency : rare, only 2.6 kernels
5695 Description: Unusual file access patterns on the MDS may result in inode
5696 data being lost in very rare circumstances.
5697 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5698 code in the 2.6 kernel caused the inode and/or EA data not to
5699 be read from disk, causing single-file corruption.
5703 Frequency : rare, only 2.6 filesystems using extents
5704 Description: Heavy concurrent write and delete load may cause data corruption.
5705 Details : It was possible under high-load situations to have an extent
5706 metadata block in the block device cache from a just-unlinked
5707 file overwrite a newly-allocated data block. We now unmap any
5708 metadata buffers that alias just-allocated data blocks.
5712 Frequency : filesystems with default stripe_count larger than 77
5713 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5714 Details : lconf specifies an inode size of 4096 bytes when the default
5715 stripe_count is larger than 77. This conflicts with the default
5716 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5717 this case to avoid pinning too much memory for large EAs.
5719 ------------------------------------------------------------------------------
5721 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5726 Frequency : rare (extremely heavy IO load with hundreds of clients)
5728 Description: Client is evicted, gets IO error writing to file
5729 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5730 caused clients to be evicted for AST timeouts. The fixes in
5731 bug 5192 mean we no longer need to have such short AST timeouts
5732 so ldlm_timeout has been increased.
5735 Frequency : occasional during --force or --failover shutdown under load
5736 Bugzilla : 5949, 4834
5737 Description: Server oops/LBUG if stopped with --force or --failover under load
5738 Details : a collection of import/export refcount and cleanup ordering
5739 issues fixed for safer force cleanup
5742 Frequency : only filesystems larger than 120 OSTs
5743 Bugzilla : 5990, 6223
5744 Description: lfs getstripe would oops on a very large filesystem
5745 Details : lov_getconfig used kfree on vmalloc'd memory
5748 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5749 Bugzilla : 6242, 6243
5750 Description: reading from files that had been truncated to a non-zero size
5751 but never opened returned no data
5752 Details : ll_file_read() reads zeros from no-object files to EOF
5757 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5758 Details : There's little harm in aborting MDS/OSS recovery and letting it
5759 try again, so I removed the LASSERT and return an error instead.
5761 Severity : enhancement
5763 Description: New debugging infrastructure for tracking down data corruption
5764 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5765 (b) cover more of the client-side code path, and (c) try to narrow
5766 down where problems occurred
5770 Bugzilla : 3819, 4364, 4397, 6313
5771 Description: Racing close and eviction MDS could cause assertion in mds_close
5772 Details : It was possible to get multiple mfd references during close and
5773 client eviction, leading to one thread referencing a freed mfd.
5775 Severity: : enhancement
5776 Bugzilla : 3262, 6359
5777 Description: Attempts to reconnect to servers are now more aggressive.
5778 Details : This builds on the enhanced upcall-less recovery that was added
5779 in 1.4.2. When trying to reconnect to servers, clients will
5780 now try each server in the failover group every 10 seconds. By
5781 default, clients would previously try one server every 25 seconds.
5786 Description: After recovery, certain operations trigger a failed
5787 assertion on a client.
5788 Details : Failing over an mds, using lconf -d --failover, while a
5789 client was doing a readdir() call would cause the client to
5790 LBUG after recovery completed and the readdir() was resent.
5792 Severity : enhancement
5794 Description: Default groups are now added by lconf
5795 Details : You can now run lconf --group <servicename> without having to
5796 manually add groups with lmc.
5799 Frequency : occasional
5801 Description: Nodes with an elan id of 0 trigger a failed assertion
5804 Frequency : always when accessing e.g. tty/console device nodes
5806 Description: tty and some other devices nodes cannot be used on lustre
5807 Details : file's private_data field is used by device data and lustre
5808 values in there got lost. New field was added to struct file to
5809 store fs-specific private data.
5812 Frequency : when exporting Lustre via NFS
5814 Description: NFSD failed occasionally when looking up a path component
5815 Details : NFSD is looking up ".." which was broken in ext3 directories
5816 that had grown large enough to become hashed.
5819 Frequency : Clusters with multiple interfaces not on the same subnet
5821 Description: Nodes will repeatedly try to reconnect to an interface which it
5822 cannot reach and report an error to the log.
5823 Details : Extra peer list entries will be created by lconf with some peers
5824 unreachable. lconf now validates the peer before adding it.
5827 Frequency : Only if a default stripe is set on the filesystem root.
5829 Description: Setting a default stripe on the filesystem root prevented the
5830 filesystem from being remounted.
5831 Details : The client was sending extra request flags in the root getattr
5832 request and did not allocate a reply buffer for the dir EA.
5835 Frequency : occasional, higher if lots of files are accessed by one client
5836 Bugzilla : 6159, 6097
5837 Description: Client trips assertion regarding lsm mismatch/magic
5838 Details : While revalidating inodes the VFS looks up inodes with ifind()
5839 and in rare cases can find an inode that is being freed.
5840 The ll_test_inode() code will free the lsm during ifind()
5841 when it finds an existing inode and then the VFS later attaches
5842 this free lsm to a new inode.
5846 Bugzilla : 6422, 7030
5847 Description: MDS deadlock between mkdir and client eviction
5848 Details : Creating a new file via mkdir or mknod (starting a transaction
5849 and getting the ns lock) can deadlock with client eviction
5850 (gets ns lock and trying to finish a synchronous transaction).
5853 Frequency : occasional
5854 Description: While starting a server, the fsfilt_ext3 module could not be
5856 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5857 kernels. Previously, lconf would still use the ext3 name
5858 when trying to load modules. Now, it will correctly use
5859 ext3 on 2.4 and ldiskfs on 2.6.
5861 Severity : enhancement
5862 Description: The default stripe count has been changed to 1
5863 Details : The interpretation of the default stripe count (0, to lfs
5864 or lmc) has been changed to mean striping across a single
5865 OST, rather than all available. For general usage we have
5866 found a stripe count of 1 or 2 works best.
5868 Severity : enhancement
5869 Description: Add support for compiling against Cray portals.
5870 Details : Conditional compiling for some areas that are different
5874 Frequency : occasional
5875 Bugzilla : 6409, 6834
5876 Description: Creating files with an explicit stripe count may lead to
5877 a failed assertion on the MDS
5878 Details : If some OSTs are full or unavailable, creating files may
5879 trigger a failed assertion on the MDS. Now, Lustre will
5880 try to use other servers or return an error to the
5884 Frequency : occasional
5886 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5887 caused lock timeout during readahead (since 1.4.2).
5888 Details : Processes doing readahead might match a lock that hasn't been
5889 granted yet if there are overlapping and conflicting lock
5890 requests. The readahead process waits on ungranted lock
5891 (original lock is CBPENDING), while OST waits for that process
5892 to cancel CBPENDING read lock and eventually evicts client.
5894 Severity : enhancement
5896 Description: Initial enabling of flock support for clients
5897 Details : Implements fcntl advisory locking and file status functions.
5898 This feature is provided as an optional mount flag (default
5899 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5900 locking are implemented yet, and those that are are not guaranteed
5901 to be completely correct in production environments.
5902 mount -t lustre -o [flock|noflock] ...
5905 Frequency : occasional
5907 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5908 assertion in the ext3 JBD (journaling) layer.
5909 Details : The b_committed_data struct is protected by the big kernel lock
5910 in 2.4 kernels, serializing journal_commit_transaction() and
5911 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5912 access to this struct. In 2.6 kernels there is finer grained
5913 locking to improve SMP performance of the JBD layer.
5917 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5922 Description: merge and cleanup kernel patches.
5925 -----------------------------------------------------------------------------
5927 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5929 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5930 versions of Lustre. Please update all servers and clients to
5931 version 1.4.2 or later at the same time. You must also run
5932 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5933 to update the configuration logs.
5935 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5936 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5937 - swab llog records of type '0' so we get proper header size/idx (5861)
5938 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5939 - fix rename of one directory over another leaking an inode (5953)
5940 - avoid SetPageDirty on 2.6 (5981)
5941 - don't re-add just-being-destroyed locks to the waiting list (5653)
5942 - when creating new directories, inherit the parent's custom
5943 striping settings if present parent (3048)
5944 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5945 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5946 - don't hold client locks on temporary worklist from l_lru (5666)
5947 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5948 - automatically evict dead clients (5921)
5949 - Update file size properly in create+truncate+fstat case (6196)
5950 - Do not unhash mountpoint dentries, do not allow removal of
5952 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5953 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5954 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5955 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5956 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5957 - Correctly handle reads of files with no objects (6243)
5958 - lctl recover will also mark a device active if deactivate used (5933)
5960 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5961 - allow --write-conf on an MDS with different nettype than client (5619)
5962 - don't write config llogs to MDS for mounts not from that MDS (5617)
5963 - lconf should create multiple TCP connections from a client (5201)
5964 - init scripts are now turned off by default; run chkconfig --on
5965 lustre and chkconfig --on lustrefs to use them
5966 - upcalls are no longer needed for clients to recover to failover
5968 - add --abort-recovery option to lconf to abort recovery on device
5970 - add support for an arbitrary number of OSTs (3026)
5971 - Quota support protocol changes.
5972 - forward compatibility changes to wire structs (6007)
5973 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5974 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5975 - improved "lustre status" script
5976 - initialize blocksize for non-regular files (6062)
5977 - added --disable-server and --disable-client configure options (5782)
5978 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5979 - Vanilla 2.4.29 support
5980 - increase maximum number of obd devices to 520 (6242)
5981 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5982 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5983 SLES 9 2.6.5 kernels
5984 - limit stripes per file to 160 (the maximum EA size) (6093)
5986 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5989 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5990 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5991 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5992 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5993 - small ext3 extents cleanups and fixes (5733)
5994 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5995 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5996 - avoid needless client->OST connect, fix handle mismatch (5317)
5997 - fix DLM error path that led to out-of-sync client, long delays (5779)
5998 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5999 - fix several locking issues related to i_size (5492,5624,5654,5672)
6000 - don't move pending lock onto export if it is already evicted (5683)
6001 - fix kernel oops when creating .foo in unlinked directory (5548)
6002 - fix deadlock in obdfilter statistics vs. object create (5811)
6003 - use time_{before,after} to avoid timer jiffies wrap (5882)
6004 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
6005 - Do not leak request if server was not able to process it (5154)
6006 - If mds_open unable to find parent dir, make that negative lookup(5154)
6007 - don't create new directories with extent-mapping (5909, 5936)
6009 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
6010 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
6011 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
6012 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
6013 - support for RHEL4 kernel on i686 (5773)
6014 - provide error messages when incompatible logs are encountered (5898)
6016 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
6017 * version 1.4.0.10 (1.4.1 release candidate 1)
6019 - don't keep a lock reference when lock is not granted (4238)
6020 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6021 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
6022 - drop import inflight refcount on signal_completed_replay error (5255)
6023 - unlock page after async write error during send (3677)
6024 - handle missing objects in filter_preprw_read properly (5265)
6025 - no transno return for symlink open, don't save no-trasno open (3440)
6026 - don't try to complete elan receive that already failed (4012)
6027 - free RPC server reply state on error (5406)
6028 - clean up thread from ptlrpc_start_thread() on error (5160)
6029 - readahead could read extra page into cache that wasn't ejected (5388)
6030 - prevent races in class_attach/setup/cleanup/detach (5260)
6031 - don't dereference de->d_inode after l_dput of de (5458)
6032 - use "int" for stripe value returned from lock_to_stripe (5544)
6033 - mballoc allocation and error-checking fixes in 2.6 (5504)
6034 - block device patches to fix I/O request sizes in 2.6 (5482)
6035 - look up hostnames for IB nals (5602)
6036 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
6037 - don't start multiple acceptors for the same port (5277)
6038 - fix incorrect LASSERT in mds_getattr_name (5635)
6039 - export a proc file for general "ping" checking (5628)
6040 - fix "lfs check" to not block when the MDS is down (5628)
6042 - service request history (4965)
6043 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6044 - create an "evict_client" /proc entry on OSTs, like the MDS has
6045 - fix mount usage message, return errors per mount(8) (5168)
6046 - change grep [] to grep "[]" in tests so they work in more UMLs
6047 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6048 - remove ancient LOV_MAGIC_V0 EA support (5047)
6049 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6050 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6051 - allow readahead allocations to fail when low on memory (5383)
6052 - mmap locking landed again, after considerable improvement (2828)
6053 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6055 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6058 - send OST transaction number in read/write reply to free req (4966)
6059 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6060 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6061 - the watchdog thread now runs as interruptible (5246)
6062 - flock/lockf fixes (but it's still disabled, pending 5135)
6063 - don't use EXT3 constants in llite code (5094)
6064 - memory shortage at startup could cause assertion (5176)
6066 - reorganization of lov code
6067 - single portals codebase
6069 - add extents/mballoc support (5025)
6070 - direct I/O reads in the obdfilter (4048)
6071 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6073 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6076 - send OST transaction number in read/write reply to free req (4966)
6077 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6078 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6079 - the watchdog thread now runs as interruptible (5246)
6080 - handle missing objects in filter_preprw_read properly (5265)
6081 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6082 - drop import inflight refcount on signal_completed_replay error (5255)
6083 - unlock page after async write error during send (3677)
6084 - return original error code on reconstructed replies (3761)
6085 - no transno return for symlink open, don't save no-trasno open (3440)
6087 - add pid to ldlm debugging output (4922)
6088 - bump the watchdog timeouts -- we can't handle 30sec yet
6089 - extra debugging for orphan dentry/inode bug (5259)
6091 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6094 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6095 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6096 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6097 - fix NULL dereference in /proc/sys/portals/routes (4827)
6098 - allow failed mdc_close() operations to be interrupted (4561)
6099 - stop precreate on OST before MDS would time out on it (4778)
6100 - don't send partial-page writes before EOF from client (4410)
6101 - discard client grant for sub-page writes on large-page clients (4520)
6102 - don't free dentries not owned by NFS code, check generation (4806)
6103 - fix lsm leak if mds_create_objects() fails (4801)
6104 - limit debug_daemon file size, always print CERROR messages (4789)
6105 - use transno after validating reply (3892)
6106 - process timed out requests if import state changes (3754)
6107 - update mtime on OST during writes, return in glimpse (4829)
6108 - add mkfsoptions to LDAP (4679)
6109 - use ->max_readahead method instead of zapping global ra (5039)
6110 - don't interrupt __l_wait_event() during strace
6112 - add software watchdogs to catch hung threads quickly (4941)
6113 - make lustrefs init script start after nfs is mounted
6114 - fix CWARN/ERROR duplication (4930)
6115 - return async write errors to application if possible (2248)
6116 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6117 - print NAL number in %x format (4645)
6118 - update barely-supported suse-2.4.21-171 series (4842)
6119 - support for sles 9 %post scripts
6120 - support for building 2.6 kernel-source packages
6121 - support for sles km_* packages
6123 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6126 - ignore -ENOENT errors in osc_destroy (3639)
6127 - notify osc create thread that OSC is being cleaned up (4600)
6128 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6129 - reconstruct ost_handle() like mds_handle() (4657)
6130 - create a new thread to do import eviction to avoid deadlock (3969)
6131 - let lconf resolve symlinked-to devices (4629)
6132 - don't unlink "objects" from directory with default EA (4554)
6133 - hold socknal file ref over connect in case target is down (4394)
6134 - allow more than 32000 subdirectories in a single directory (3244)
6135 - fix blocks count for O_DIRECT writes (3751)
6136 - OST returns ENOSPC from object create when no space left (4539)
6137 - don't send truncate RPC if file size isn't changing (4410)
6138 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6139 - bind to privileged port in socknal and tcpnal (3689)
6141 - rate limit CERROR/CWARN console message to avoid overload (4519)
6142 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6143 - basic mmap support (3918)
6144 - kernel patch series update from b1_4 (4711)
6146 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6149 - avoid crash during MDS cleanup with OST shut down (2775)
6150 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6151 - don't use bad inodes on the MDS (3744)
6152 - dynamic object preallocation to improve recovery speed (4236)
6153 - don't hold spinlock over lock dumping or change debug flags (4401)
6154 - don't zero obd_dev when it is force cleaned (3651)
6155 - print grants to console if they go negative (4431)
6156 - "lctl deactivate" will stop automatic recovery attempts (3406)
6157 - look for existing locks in ldlm_handle_enqueue() (3764)
6158 - don't resolve lock handle twice in recovery avoiding race (4401)
6159 - revalidate should check working dir is a directory (4134)
6161 - don't always mark "slow" obdfilter messages as errors (4418)
6163 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6166 - don't close LustreDB during write_conf until it is done (3860)
6167 - fix typo in lconf for_each_profile (3821)
6168 - allow dumping logs from multiple threads at one time (3820)
6169 - don't allow multiple threads in OSC recovery (3812)
6170 - fix debug_size parameters (3864)
6171 - fix mds_postrecov to initialize import for llog ctxt (3121)
6172 - replace config semaphore with spinlock (3306)
6173 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6174 - don't allow enqueue to complete on a destroyed export (3822)
6175 - down write_lock before checking llog header bitmap (3825)
6176 - recover from lock replay timeout (3764)
6177 - up llog sem before sending rpc (3652)
6178 - reduce ns lock hold times when setting kms (3267)
6179 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6180 - fix NULL deref and obd_dev leak on setup error (3312)
6181 - replace some LBUG about llog ops with error handling (3841)
6182 - don't match INVALID dentries from d_lookup and spin (3784)
6183 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6184 - fix invalid assertion in ptlrpc_set_wait (3880)
6186 - add libwrap support for the TCP acceptor (3996)
6187 - add /proc/sys/portals/routes for non-root route listing (3994)
6188 - allow setting MDS UUID in .xml (2580)
6189 - print the stack of a process that LBUGs (4228)
6191 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6194 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6195 - make sure to unset replay flag from failed open requests (3440)
6196 - if default stripe count is 0, use OST count for inode size (3636)
6197 - update parent mtime/ctime on client for create/unlink (2611)
6198 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6199 - free recovery state on server during a forced cleanup (3571)
6200 - unregister_reply for resent reqs (3063)
6201 - loop back devices mounting and status check on 2.6 (3563)
6202 - fix resource-creation race that can provoke i_size == 0 (3513)
6203 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6204 - more debugging for page-accounting assertion (3746)
6205 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6206 - avoid deadlock after precreation failure (3758)
6207 - fix race and lock order deadlock in orphan handling (3450, 3750)
6208 - add validity checks when grabbing inodes from l_ast_data (3599)
6210 - add /proc/.../recovery_status to obdfilter (3428)
6211 - lightweight CDEBUG infrastructure, debug daemon (3668)
6212 - change default OSC RPC parameters to be better on small clusters
6213 - turn off OST read cache for files smaller than 32MB
6214 - install man pages and include them in rpms (3100)
6215 - add new init script for (un)mounting lustre filesystems (2593)
6216 - run chkconfig in %post for init scripts (3701)
6217 - drop scimac NAL (unmaintained)
6219 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6222 - clean kiobufs before and after use (3485)
6223 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6224 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6225 - change init script to fail more clearly if not run as root (1528)
6226 - allow clients to reconnect during replay (1742)
6227 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6228 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6229 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6230 - fix cond_resched (3554)
6231 - only evict unfinished clients after recovery (3515)
6232 - allow bulk resend, prevent data loss (3570)
6233 - dynamic ptlrpc request buffer allocation (2102)
6234 - don't allow unlinking open directory if it isn't empty (2904)
6235 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6236 - remove extraneous obd dereference causing LASSERT failure (3334)
6237 - don't use get_cycles() when creating temp. files on the mds (3156)
6238 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6239 - handle EEXIST for set-stripe, set proper directory name (3336)
6241 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6242 - fix ksocknal_fmb_callback() error messages (2918)
6244 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6247 - don't copy lvb into (possibly NULL) reply on error (2983)
6248 - don't deref dentry after dput, don't free lvb on error (2922)
6249 - use the kms to determine writeback rpc length (2947)
6250 - increment oti_logcookies when osc is inactive (2948)
6251 - update client's i_blocks count via lvb messages (2543)
6252 - handle intent open/close of special files properly (1557)
6253 - mount MDS with errors=remount-ro, like obdfilter (2009)
6254 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6255 - don't use cancelling-locks' kms values (2947)
6256 - use highest lock extent for kms, not last one (2925)
6257 - don't dereference ERR_PTR() dentry in error handling path (3107)
6258 - fix thread race in portals_debug_dumplog() (3122)
6259 - create lprocfs device entries at setup instead of at attach (1519)
6260 - common AST error handler, don't evict client on completion race (3145)
6261 - zero nameidata in detach_mnt in 2.6 (3118)
6262 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6263 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6264 - initialize RPC timeout timer earlier for 2.6 (3219)
6265 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6266 - print nal/nid for unknown nid (3258)
6267 - additional checks for oscc recovery before doing precreate (3284)
6268 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6269 - don't crash in mdc_close for bad permissions on open (3285)
6270 - zero i_rdev for non-device files (3147)
6271 - clear page->private before handing to FS, better assertion (3119)
6272 - tune the read pipeline (3236)
6273 - fix incorrect decref of invalidated dentry (2350)
6274 - provide read-ahead stats and refine rpc in flight stats (3328)
6275 - don't hold journal transaction open across create RPC (3313)
6276 - update atime on MDS at close time (3265)
6277 - close LDAP connection when recovering to avoid server load (3315)
6278 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6279 - don't leak open file on MDS after open resend (3325)
6280 - serialize filter_precreate and filter_destroy_precreated (3329)
6281 - loop device shouldn't call sync_dev() for nul device (3092)
6282 - clear page cache after eviction (2766)
6283 - resynchronize MDS->OST in background (2824)
6284 - refuse to mount the same filesystem twice on same mountpoint (3394)
6285 - allow llmount to create routes for mounting behind routers (3320)
6286 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6287 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6288 - fix rare problem with rename on htree directories (3417)
6290 - allow default OST striping configuration per directory (1414)
6291 - fix compilation for qswnal for 2.6 kernels (3125)
6292 - increase maximum number of MDS request buffers for large systems
6293 - change liblustreapi to be useful for external progs like lfsck (3098)
6294 - increase local configuration timeout for slow disks (3353)
6295 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6297 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6300 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6301 - don't overwrite extent policy data in reply if lock was blocked (2901)
6302 - drop filter export grants atomically with removal from device (2663)
6303 - del obd_self_export from work_list in class_disconnect_exports (2908)
6304 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6305 - swab reply message in mdc_close, other PPC fixes (2464)
6306 - fix destroying of named logs (2325)
6307 - overwrite old logs when running lconf --write_conf (2264)
6308 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6309 - fix race in target_handle_connect (2898)
6310 - mds_reint_create() should take same inode create lock (2926)
6311 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6312 - don't close files for self_export to avoid uninitialized obd (2936)
6313 - allow MDS with the same name as client node (2939)
6314 - hold dentry reference for closed log files for unlink (2325)
6315 - reserve space for all logs during transactions (2059)
6316 - don't evict page beyond end of stripe extent (2925)
6317 - don't oops on a deleted current working directory (2399)
6318 - handle hard links to targets without a parent properly (2517)
6319 - don't dereference NULL lock when racing during eviction (2867)
6320 - don't grow lock extents when lots of conflicting locks (2919)
6322 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6325 - account for cache space usage on clients to avoid data loss (974)
6326 - lfsck support in lustre kernel code (2349)
6327 - reduce journal credits needed for BRW writes (2370)
6328 - orphan handling to avoid losing space on client/server crashes
6329 - ptlrpcd can be blocked, stopping ALL progress (2477)
6330 - use lock value blocks to assist in proper KMS, faster stat (1021)
6331 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6332 - recovery for initial connections (2355)
6333 - fixes for mds_cleanup_orphans (1934)
6334 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6335 - block all file creations until orphan recovery completes (1901)
6336 - client remove rq_connection from request struct (2423)
6337 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6338 - recovery timer race (2670)
6339 - mdc_close recovey bug (2532)
6340 - ptlrpc cleanup bug (2710)
6341 - mds timeout on local locks (2588)
6342 - namespace lock held during RPCs (2431)
6343 - handle interrupted sync write properly (2503)
6344 - don't try to handle a message that hasn't been replied to (2699)
6345 - client assert failure during cleanup after abort recovery (2701)
6346 - leak mdc device after failed mount (2712)
6347 - ptlrpc_check_set allows timedout requests to complete (2714)
6348 - wait for inflight reqs when ptlrpcd finishes (2710)
6349 - make sure unregistered services are removed from the srv_list
6350 - reset bulk XID's when resending them (caught by 1138 test)
6351 - unregister_bulk after timeout
6352 - fix lconf error (2694)
6353 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6354 - readahead locks pages, leaves pending causing memory pressure (2673)
6355 - increase OST request buffers to 4096 on large machines (2729)
6356 - fix up permission of existing directories in simple_mkdir (2661)
6357 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6358 - don't assign transno to errored transactions (2742)
6359 - don't delete objects on OST if given a bogus objid from MDS (2751)
6360 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6361 - if rq_no_resend, then timeout request after recovery (2432)
6362 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6363 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6364 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6365 - MDS recovery shouldn't delete objects that it has given out (2730)
6366 - if enqueue arrives after completion, don't clobber LVB (2819)
6367 - don't unlock pages twice when trigger_group_io returns error (2814)
6368 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6369 - don't write pages to disk if there was an error (1450)
6370 - don't ping imports that have recovery disabled (2676)
6371 - take buffered bytes into account when balancing socknal conn (2817)
6372 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6373 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6374 - remove little-endian swabbing of llog records (1987)
6375 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6376 - retry reposting request buffers if they fail (1191)
6377 - grow extent at grant time to avoid granting a revoked lock (2809)
6378 - lock revoke doesn't evict page if covered by a second lock (2765)
6379 - disable VM readahead to avoid reading outside lock extents (2805)
6381 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6382 - updated kernel patches for hp-2.4.20 kernel (2681)
6384 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6387 - fix truncated write corruption (2366)
6388 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6390 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6391 - ensure len(uuid) < 37 in lmc (1171)
6392 - fix ia64 OOPS in llog_test (2255)
6393 - zero end of page at obdfilter for partial page writes (2648)
6394 - don't leave stale dentries around after renames (bug 2428)
6395 - fix timeouts when evicting a client with a single lock held (2642)
6396 - set deadline for the initial HELLO message to drain (2634)
6397 - print out dotted-quad IP addresses in the socknal (2302)
6399 - additional debugging for MDS client eviction problem (2443)
6400 - fix mkfsoptions support for osts (2603, 2604)
6402 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6405 - add series for the vanilla 2.6.0 kernel
6406 - add series for the vanilla 2.4.24 kernel
6407 - add series for a cray x86/64 UL kernel drop
6408 - fix xattr patches for the vanilla 2.4.19 series
6410 - generate true UUIDs in lmc (1171)
6411 - have portals stack dumping break in UML (2466)
6412 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6413 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6414 - wake callback waiting threads on client eviction (2460)
6415 - Add --ptldebug and --subsystem to lmc (1719)
6416 - update assertion to allow safe interrupt allocation
6417 - set rq_no_resend for cancel requests (2432)
6418 - recalculate ptlrpcd timeout after resend (2494)
6419 - call vfs_rmdir when removing pending directories (2368)
6420 - fix renaming a file to itself (2429)
6421 - lmc creates a default one-stripe lov (2454)
6422 - expand procfs space to handle large clusters (2326)
6423 - increase UML stack to avoid overflow
6424 - update lconf's list of debug and subsystem masks
6425 - fix lfs find --obd (2510)
6426 - /proc tunable for disabling filter read caching (2591)
6427 - stop rpm packages from altering slapd.conf (2301)
6428 - disable nagle in the socknal under 0conf (2578)
6429 - choose mds inode size based on stripe count (2572)
6430 - fix kernel-source rpm problems (2516)
6432 - add --disable-doc to avoid pdf generation (2421)
6433 - update documentation, tests, type-os, comments
6434 - avoid format warnings on ia64
6435 - remove the TOE NAL
6436 - tiny code cleanups by removing unused fields
6438 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6441 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6442 - remove the most blatant lies from BUILDING file (2371)
6443 - change default debug level to reasonable production setting
6444 - reduce client side cache size to reduce cache flush time
6445 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6446 - make TCP zerocopy and pinger support enabled by default (2476)
6447 - sync writes completed after process exits caused crashes (2319)
6448 - maintain correct mount count on the MDS (2356)
6449 - backout 1557, because 2316 wasn't really fixed
6450 - better file I/O statistics gathering in /proc
6451 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6452 - another kernel patch to fix zero-copy TCP function export
6453 - don't take duplicate lock when processing re-sent getattr (2420)
6454 - lctl uses obd_self_export instead of creating new conn (2353)
6455 - MDS/OST recovery case which requires object creation asserted (2425)
6456 - move lfs from /usr/sbin to /usr/bin in packages
6457 - fix race between mds_client_add and mds_client_free (2417)
6458 - use kmalloc instead of slabs in portals (2430)
6459 - don't create duplicate records when a failover MDS is present (2442)
6460 - remove unnecessary mount age check (2332)
6461 - don't remove directory inodes from locks prematurely (2451)
6462 - don't break if MDS service name is the same as hostname (2103)
6463 - fix races in client write RPC generation when cache full (2482)
6465 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6468 - remove now-unused request->rq_obd (278)
6469 - if an allocation fails, print out how much memory we've used (1933)
6470 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6471 - add the "configurable stack size" patch to most series files (1256)
6472 - ability to write large log records, for 100+ OST configs (2306)
6473 - fix NULL deref when filter_prep fails (2314)
6474 - fix operator precedence error in filter_sync
6475 - dynamic allocation of socknal TX descriptors (2315)
6476 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6477 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6478 - ensure that CWARN messages in Portals always get to the syslog
6479 - __init/__exit are not for prototype decls (ldlm_init/exit)
6480 - x86-64 compile warning fixes
6481 - fix gateway LMC keyword conflict (2318)
6482 - fix MDS lock inversions in getattr/reint paths (1844)
6483 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6484 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6485 - DLM race condition prevented some lock evictions (2328)
6486 - ENOMEM detection and retry on socknal sends (2230)
6487 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6488 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6489 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6490 - better error messages when a client is rejected during recovery (1505)
6491 - avoid cancelling locks which were never granted, after failure (2330)
6492 - fix i_sem/journal inversion in mds_client_add (2333)
6493 - fix truncate/getattr lock cycle deadlock (2334)
6494 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6495 - fix two rare exit paths which could leak an l_lock() ref (2321)
6496 - fencepost error in MDS/OST orphan recovery (2226)
6497 - make log record alignment 8 bytes (1988)
6498 - lstripe now fails when requested offset > ost_count (2237)
6499 - ensure that all kernel series have a complete list.h (1607)
6500 - fix crashes in special-file operations (2316)
6501 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6503 - allow configurable automake binary, for testing new versions
6504 - small update to the lfs documentation
6506 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6508 * fix negative export reference count in fsfilt_sync (2312)
6510 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6511 * release candidate 0.9.1
6513 - orphans are moved into the PENDING directory for possible recovery
6514 - replayed opens now open by fid for orphan/rename safety (1042)
6515 - last close of an orphan inode generates a transno (683)
6516 - chdir() and mount() now pin the directory entry (1020)
6517 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6518 - discard very old requests without processing them (1502)
6519 - remove some common, well-understood CERRORs (1505)
6520 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6521 - clear "grant" flags in OST replies until OST grant code lands (1644)
6522 - fix read performance by not clobbering i_blksize on client (1598)
6523 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6524 - make LRU size a /proc tunable, clears locks when reduced (707)
6525 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6526 - server locks take references on exports now (1558)
6527 - build fixes for 2.4.20-rh trees (1663)
6528 - return an error from lov_create if all OSCs are inactive (1751)
6529 - fix import levels when a reconnect happens without a timeout (1597)
6530 - exit early from mds_open if we get a lookup error (1749)
6531 - partial page read at EOF wouldn't wait for disk before sending (1642)
6532 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6533 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6534 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6535 - write/truncate lock inversion (1639)
6536 - Don't auto-load obdclass, portals modules during cleanup (1495)
6537 - fix timestamps from jumping to "now" (1763)
6538 - extra journal assertions (1648)
6539 - add an extra multiunlink test (1771)
6540 - fix read_record/write_record API (1776)
6541 - fix leak of offset_extent, possible incorrect i_size later (1772)
6542 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6543 - Debugging for the kqswnal_get_idle_tx problems (1820)
6544 - Allow recovery to be attempted multiple times (1536)
6545 - Write out MDS last_rcvd file after it is first created (1600)
6546 - Fix tx_descriptor leak in failed transmit situations (1827)
6547 - ext3 journaling fixes for assertion failure after IO error (1871)
6548 - class_export_put() on freed export after completion AST error (1896)
6549 - Fix revalidate looping in VFS (1322)
6550 - Don't access a freed export during MDS_REINT timeout (1521)
6551 - Add open-unlink recovery support on the MDS (1673,1764)
6552 - Return an error if no MDS data was read from last_rcvd (1946)
6553 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6554 - Don't setup a disk device that doesn't match exported UUID (317)
6555 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6556 - avoid committing NULL handle in force close
6557 - local.sh is now a one-stripe LOV configuration
6558 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6559 - fix ext3 htree duplicate directory entry corruption (1516)
6560 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6561 - update to new LOV EA format (2097)
6562 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6564 2003-06-15 Phil Schwan <phil@clusterfs.com>
6567 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6568 - per-import/export recovery handling (958, 931, 959)
6569 - multiple last-rcvd slots, for serving multiple FSes (949)
6570 - connections are again shared between multiple imp/exports (963, 964)
6571 - "umount -f" would hang if any requests needed to be sent (393, 978)
6572 - avoid pinning large req buffer by copying for queued messages (989)
6573 - add "uuid" to "lctl device" command to help upcalls (991)
6574 - "open" RPCs with transnos would confuse recovery counters (1037)
6575 - do proper endian conversion of all wire messages (288, 340, 891)
6576 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6577 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6578 - fix ptlrpc_connection leak in target_handle_connect (1174)
6579 - fix import refcounting bug in OST and MDS cleanup (1134)
6580 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6581 - fix very unlikely obd_types race condition (501)
6582 - remove osc_open hack for echo_client (1187)
6583 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6584 - a failure in read_inode2 leads to deadlock (1139)
6585 - cancel ack-locks as soon as transaction is committed (1072)
6586 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6587 - make sure to commitrw after any preprw to avoid deadlock (1162)
6588 - failing to execute a file in a lustre FS would lock inode (1203)
6589 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6590 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6591 - obd_uuid2tgt would walk off the end of the list (1255)
6592 - on IA64 the osc would give portals incorrect bulk size (1258)
6593 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6594 - fix lock inversion caused by new llite matching code (1282)
6595 - limit the number of dirty pages on a client to 10MB (1286)
6596 - timed out locks were not being corrected cancelled (1289)
6597 - fix O_DIRECT above 4GB on IA-32 (1292)
6598 * major user-visible changes
6599 - fail out/fail over policy now controlled by the upcall (993)
6601 - add OBD_PING to check server availability and failure (954)
6602 - lustre messages are now sent in sending host order (288, 340, 891)
6603 - add eadatalen to MDS getattr reply (340)
6604 - OST read replies may contain second buffer, with per-page status (593)
6606 2003-03-11 Phil Schwan <phil@clusterfs.com>
6609 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6610 - failed open()s could cause deadlock; fixed (867, 869)
6611 - stop cancelling OST locks when files are closed (481)
6612 - overlapping XID spaces caused network corruption (851, 853)
6613 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6614 - setattr_raw added, to do single-RPC, server-side setattrs
6615 - lmc/lconf syntax change for OST UUIDs
6616 - fix crashy race condition between ptlrpc_free_req and osc_close
6617 - don't use request in mdc_enqueue if we hit a timeout (889)
6618 - don't set the inode i_size for regular files from the MDS (896)
6619 - handle out of order completion AST (842)
6620 - don't LBUG if a lock request times out after receiving AST (913)
6621 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6622 - if a bad lock AST arrives, send an error instead of dropping entirely
6623 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6624 - fix leak in bulk IO when only partially completed (899, 900, 926)
6625 - fix O_DIRECT for ia64 (55)
6626 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6627 - C-z after timeout could hang a process forever; fixed (977)
6629 - client-side I/O cache (678, 924, 929, 941, 970)
6631 - READPAGE and SETATTRs which don't take server-side locks get
6634 2003-02-11 Phil Schwan <phil@clusterfs.com>
6637 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6638 - Fix fsfilt_extN_readpage() to read a full page of directory
6639 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6640 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6641 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6642 - Most intent operations were removed, in favour of a new RPC mode
6643 that does a single RPC to the server and bypasses most of the VFS
6644 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6645 - Aggressively cancel local locks on DLM servers
6646 - mds_reint_unlink sends EA to the client if it's the last nlink.
6647 client uses that EA to unlink OST objects.
6648 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6649 - recursive symlinks were fixed (439)
6650 - fixed NULL deref in DEBUG_REQ
6651 - filter_update_lastobjid no longer calls sync, which annoyed extN
6652 - fixed multi-client small-writes to a single file problem (445)
6653 - fixed mtime updates during file writes (607)
6654 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6655 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6656 - fixed Linux OST size reporting problem (444, 656)
6657 - OST now updates object mtime with writes or setattr (607, 619)
6658 - client verifies file size before zeroing page past EOF (445)
6659 - OST now writes last allocated objid to disk with allocation (108)
6660 - LOV on echo now works (409)
6662 - mds_reint_unlink sends a new buffer, with the EA included. this
6663 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6664 set if a regular file was being unlinked, and it was the last link
6665 - use PtlGet from the target for bulk writes (315)
6666 - OST now updates object mtime with writes or setattr (607, 619)
6667 - LDLM now has a grant-time callback to revalidate locked items, if
6669 - Many MDS operations were reorganized to combat race conditions
6671 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6672 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6674 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6677 - Fully reactivate OST imports after reconnection (512, others)
6678 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6679 - More graceful error handling for truncating on dead OST (515)
6680 - Don't error out unless we're actually accessing dead stripes (474)
6681 - Fix garbage sizes when stripes are missing (410)
6682 - LRU counters were broken, causing constant lock purge (433, 432)
6683 - garbage on read from stripes with failed OSTs (441)
6684 - mark OSCs as active before reconnecting during recovery (438)
6685 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6686 - lfind did not preserve OST order in output (443)
6687 - symlinks cause hung clients, incorrect data (439)
6688 - stop dereferencing request after dropping refcount (457)
6689 - don't LASSERT(spin_is_locked) on non-SMP (455)
6690 - fixes for many rename() bugs
6691 - fstat didn't correctly synchronize attributes (399)
6692 - server must handle lock cancellation during blocking AST prep (487)
6693 - bulk descriptors were free()d too soon (511)
6694 - fix paths in lconf, which would load incorrect modules (451, 507)
6695 - fix confusing lconf 'host not found' error message (386)
6696 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6697 - fix race condition in mdc_blocking_ast() for inode access (526)
6698 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6699 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6700 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6701 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6702 - fix lstripe in several ways (kernel side) (527)
6703 - fix request leak in ldlm_cli_enqueue (262)
6704 - incorrect OSC was marked inactive after OST failure
6705 - call mds_fs_cleanup before unmounting filesystem (524)
6706 - fix races between taking ns_lock and ldlm_lock_change_resource
6707 - fix races updating LOV export open file list
6708 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6709 - fix recovery NULL deref in ldlm_cli_cancel_unused
6710 - fix some DLM races by using new hash table for lock handles (419)
6711 - permit the client to specify desired inodes, at replay
6712 - duplicate requests when we queue them for replay reintegration
6713 - fix last_rcvd offset calculation
6714 - sync after each recovered transaction, so we always make progress
6715 - never, not always, ERESTART requests without transnos
6716 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6718 - skip replay if the MDS says that the client is already connected
6719 - don't check for a recovery-enabled export to match lctl's UUID
6720 - don't INC_USE_COUNT for phantom exports
6721 - don't crash when cleaning up phantom exports (567)
6722 - don't double-finish or set replay data for errored mdc_open requests
6723 - abort requests when they time out, so we don't get old replies
6724 - send/receive replies for AST messages again
6725 - if the client says that it doesn't have the lock, cancel it on the
6727 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6728 mark it as destroyed, it will all work out when decref is called
6729 - fix module use counts (22, 581)
6731 - ASTs now expect a reply (server cancels lock on error reply)
6733 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6736 - fix many simultaneous client startup (392)
6737 - fix dentry->d_it clobbering
6738 - credentials weren't being shipped for readdir/getattr operations
6739 - remove invalid assertions triggered during some concurrent MD
6741 - proper Lustre versions added (336, 389)
6742 - fix memory leak for create error case (398)
6743 - fix LOV locking bug that would get cli/srv out of sync
6744 - fix echo client over LOV (409)
6745 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6746 - fix double-O_EXCL intent crash (424)
6747 - avoid sending multiple lock CANCELs (352)
6749 - MDS can do multi-client recovery (modulo bugs in new code)
6751 - many updates, edits, cleanups
6753 2002-11-18 Phil Schwan <phil@clusterfs.com>
6756 - fix null d_it dereference (346)
6757 - fix full OST/dbench hang (333)
6758 - fix permission problem with file removal (286)
6759 - fix removal of OSCs from LOV when they fail
6760 - fix NULL deref during bulk timeout (214)
6761 - fix problems related to multiple filesystems on one MDS (241)
6762 - fixed serious subtle metadata locking bugs
6763 - free locks on clients when inodes are removed due to memory
6765 - fix inode pointer in lock data (285)
6766 - partial support for multiple MDS on a single host (241)
6767 - data locks weren't cancelled at clear_inode time (290, 311)
6768 - intent locks could lead to unbounded lock growth (205)
6769 - added a maximum lock count, an LRU list, and a flusher
6770 - fix multiple rename (365)
6771 - properly abstracted the echo client
6772 - OSC locked 1 byte too many; fixed
6773 - rewrote brw callback code:
6774 - fixed recovery bugs related to LOVs (306)
6775 - fixed too-many-pages-in-one-write crash (191)
6776 - fixed (again) crash in sync_io_timeout (214)
6777 - probably fixed callback-related race (385)
6779 - Add capability to MDS protocol
6780 - LDLM cancellations and callbacks on different portals
6782 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6785 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6786 - defer open object destruction to close time (601981, 138)
6787 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6788 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6789 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6790 - don't call obd_setattr in ll_file_release for destroyed objects
6791 * protocol change to lustre_msg: move |version| and add |flags|
6792 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6793 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6794 * added replay of create, unlink, link and rename operations during
6795 MDS failover; recovery should be much more robust now
6796 * remove failed OSCs from LOVs (only lov_create uses this so far)
6797 * the lustre-HOWTO was brought (more) up to date (582544)
6799 2002-10-23 Phil Schwan <phil@clusterfs.com>
6802 - in-use dentries weren't being reused properly (617851)
6803 - prevent multiple LDLM setup (599178)
6804 - fix LOV size calculations for truncate (617853)
6805 - fix client handling of MDS intent errors (POSIX)
6806 - fix permission bug in lovstripe.c test (624321)
6807 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6808 - truncate past end of file could corrupt data
6809 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6810 - a race in recovery could return ETIMEDOUT to apps (623947)
6811 - building outside the source directory was fixed
6812 * the lustre-HOWTO was brought (more) up to date (582544)
6813 * major progress was made on recovery functionality
6815 2002-10-10 Phil Schwan <phil@clusterfs.com>
6818 - recovery deadlock fix
6819 - rm -rf causes LBUG fix (617817)
6820 - file open by multiple tasks fix (618962)
6821 - directory permissions bugs (602707 and 620007)
6822 - journal_stop fixed with locking (611313)
6823 - O_APPEND failures resolved (618273, perhaps 614459)
6824 - lconf PATH fix (619770)
6825 - IA64 build fix (621450)
6826 - RPC buffer sizes scale with amount of memory
6828 2002-10-01 Phil Schwan <phil@clusterfs.com>
6831 - locks would be cancelled without throwing away data pages,
6832 resulting in inconsistent data (605627)
6833 - inode attributes were not always being refreshed (605627, 612449)
6834 - lconf now continues to cleanup after lctl reports an error
6835 - MDS now enforces user permissions (602707)
6836 - lprocfs cleanup fixed, but not yet enabled (614157)
6837 - fixed infinite server hang, should a client not respond to an AST
6838 - avoid going into recovery if user calls readlink() with a buffer
6839 that's too small (613941)
6840 - AST RPCs no longer require replies (614867) -- this may be changed
6841 - don't crash server if client sends an IOV that's too big (611336)
6842 - fixed lock conversion deadlock (611892)
6843 - fixed the following of symlinks (614622)
6844 * recovery: the server can remove locks from a client that dies, other
6845 clients can make progress
6846 * more extN patch fixes
6847 * compile-time configurable ptlrpc buffer allocations
6849 - collaborative read cache document
6850 - Lustre Lite Performance CDR document-in-progress
6852 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6855 - fix typo in patch-2.4.18
6857 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6860 - clear ptlrpc request each time in handle_incoming_request()
6861 - unlink of files now destroys the object on the OST
6863 2002-09-19 Peter Braam <braam@clusterfs.com>
6865 * add hard link support
6866 * change obdfile creation method
6867 * kernel patch changed
6869 2002-09-19 Peter Braam <braam@clusterfs.com>
6872 - stack overflow bug in extN fixed
6874 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6876 * documentation updates
6877 - add man pages for config tools
6878 - update tests/README to describe testing with new config tools
6879 - finish metadata API descriptions
6880 * bug fixes and cleanups
6881 - statfs workaround for 16TB limit
6882 - LOV stripe allocation improved, can stripe on subset of OSTs
6883 - LOV file size/IO offset was wrong for files > 4GB in size
6884 - object EA data was being dropped, caused files to be unreadable
6885 - memory overflow with non-LOV OST caused memory corruption
6886 - fixed regression tests to work with new config tools, obdfilter
6887 - fixed bug when directory size became larger than 1 block
6888 - fixed bug (for single client case) when PWD was deleted
6889 - invalidate local directory pages when doing intent-based ops
6890 - avoid LDLM oops when lock callback contained bad data
6892 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6894 * documentation updates
6895 * bug fixes and cleanups
6896 - configuration tools
6899 - 64-bit compile warnings
6900 - 64-bit internal statfs data
6902 * test_brw on persistent OST devices
6904 * lprocfs (disabled)
6906 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6908 * documentation updates
6909 * bug fixes and cleanups
6910 * configuration tools
6912 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6915 * many small fixes to 0_5_4
6916 * io/network handling
6917 * thinkos in MDS operations
6919 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6922 * crucial basic fixes to 0.5.3
6923 * IOR, Iozone work over Elan
6926 2002-08-07 Phil Schwan <phil@clusterfs.com>
6927 * version 0_5_3, our first alpha
6928 * we use the new Portals iovs
6929 * documentation updates
6930 * bug fixes and cleanups
6931 * small changes in the DLM wire protocol
6933 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6934 * version 0_5_1 with some initial stability,
6935 * locking on MD and file I/O.
6936 * documentation updates
6937 * several bug fixes since 0.5.0
6938 * small changes in wire protocol
6940 2002-07-18 Phil Schwan <phil@clusterfs.com>
6942 * delivered as Lustre Light Alpha
6943 * fixed a crash after handling invalid MDS requests
6944 * fixed directory pages for architectures with non-4k pages sizes
6946 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6947 * release version v0_4_4
6948 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6950 * Fixes a number of interruption problems with OST operations.
6951 * Update documentation for portals header changes
6952 * Move all wire protocol structs/defines to lustre_idl.h
6953 * Fixes symlink length bug.
6954 * Add tcpdump to repository.
6956 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6957 * release version v0_4_3
6958 * Fixes statfs for inodes on extN.
6959 * Fixes bug in runtests which would delete /etc/hosts.
6960 * Use 64-bit object IDs wherever possible (not into VFS though)
6961 Remove ost_get_info, which is unused by lustre, and out of date.
6963 2002-07-03 Peter Braam <braam@clusterfs.com>
6964 * release version v0_4_2 Fixes a lookup error (type not passed)
6965 * move forward to head of Portals
6966 * move forward to latest Lustre kernel
6968 2002-06-25 Peter Braam <braam@clusterfs.com>
6969 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob