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 5.14.0-162.23.1.el9 (RHEL9.1)
7 4.18.0-477.el8 (RHEL8.8)
8 * Other server kernels known to build and work at some point (others may also work):
9 3.10.0-862.14.4.el7 (RHEL7.5)
10 3.10.0-957.27.2.el7 (RHEL7.6)
11 3.10.0-1062.18.1.el7 (RHEL7.7)
12 3.10.0-1127.19.1.el7 (RHEL7.8)
13 3.10.0-1160.88.1.el7 (RHEL7.9)
14 4.18.0-80.11.2.el8 (RHEL8.0)
15 4.18.0-147.8.1.el8 (RHEL8.1)
16 4.18.0-193.28.1.el8 (RHEL8.2)
17 4.18.0-240.22.1.el8 (RHEL8.3)
18 4.18.0-305.25.1.el8 (RHEL8.4)
19 4.18.0-348.23.1.el8 (RHEL8.5)
20 4.18.0-372.32.1.el8 (RHEL8.6)
21 4.18.0-425.10.1.el8 (RHEL8.7)
22 4.4.120-92.70 (SLES12 SP2)
23 4.4.180-94.100 (SLES12 SP3)
24 4.4.0-131 (Ubuntu 16.04)
25 4.15.0-32 (Ubuntu 18.04)
26 5.4.0-48 (Ubuntu 20.04)
27 vanilla linux 5.4.0 (ZFS + ldiskfs)
28 vanilla linux 5.4.21 (ZFS + ldiskfs)
29 vanilla linux 5.4.136 (ZFS + ldiskfs)
30 5.10.0-60.79.0.103.oe2203 (openEuler 22.03 LTS)
31 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
32 * Client primary kernels built and tested during release cycle:
33 5.14.0-283.el9 (RHEL9.2)
34 5.14.0-162.23.1.el9 (RHEL9.1)
35 4.18.0-477.el8 (RHEL8.8)
36 5.4.0-37 (Ubuntu 20.04)
37 5.14.21-150400.24.46 (SLES15 SP4)
38 * Other clients known to build on these kernels at some point (others may also work):
39 3.10.0-862.14.4.el7 (RHEL7.5)
40 3.10.0-957.27.2.el7 (RHEL7.6)
41 4.14.0-49.13.1.el7a (RHEL7.5)
42 4.14.0-115.2.2.el7a (RHEL7.6)
43 3.10.0-1062.18.1.el7 (RHEL7.7)
44 3.10.0-1127.19.1.el7 (RHEL7.8)
45 3.10.0-1160.88.1.el7 (RHEL7.9)
46 4.18.0-80.11.2.el8 (RHEL8.0)
47 4.18.0-147.8.1.el8 (RHEL8.1)
48 4.18.0-193.28.1.el8 (RHEL8.2)
49 4.18.0-240.22.1.el8 (RHEL8.3)
50 4.18.0-305.25.1.el8 (RHEL8.4)
51 4.18.0-348.23.1.el8 (RHEL8.5)
52 4.18.0-372.32.1.el8 (RHEL8.6)
53 4.18.0-425.10.1.el8 (RHEL8.7)
54 5.14.0-70.30.1.el9 (RHEL9.0)
55 4.4.120-92.70 (SLES12 SP2)
56 4.4.180-94.100 (SLES12 SP3)
57 4.12.14-95.48 (SLES12 SP4)
58 4.12.14-122.133 (SLES12 SP5)
59 4.12.14-197.75 (SLES15 SP1)
60 5.3.18-24.96 (SLES15 SP2)
61 5.3.18-150300.59.93 (SLES15 SP3)
62 4.4.0-131 (Ubuntu 16.04)
63 4.15.0-48 (Ubuntu 18.04)
64 5.8.0-53 (Ubuntu 20.04.2 HWE)
65 5.11.0-31 (Ubuntu 20.04.3 HWE)
66 5.11.0 (vanilla kernel.org)
67 5.10.0 (openEuler 22.03 LTS)
68 * Recommended e2fsprogs version: 1.46.6-wc1 or newer
69 * Recommended ZFS version: 2.1.5
70 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
71 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
72 stack overflow. For more information, please refer to bugzilla 17630.
73 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
74 the re-exporting nodes to be mounted with "32bitapi" mount option
76 --------------------------------------------------------------------------------
81 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
82 for currently supported client and server kernel versions.
83 * Server primary kernels built and tested during release cycle:
84 4.18.0-348.2.1.el8 (RHEL8.5)
85 * Other server kernels known to build and work at some point (others may also work):
86 3.10.0-862.14.4.el7 (RHEL7.5)
87 3.10.0-957.27.2.el7 (RHEL7.6)
88 3.10.0-1062.18.1.el7 (RHEL7.7)
89 3.10.0-1127.19.1.el7 (RHEL7.8)
90 3.10.0-1160.49.1.el7 (RHEL7.9)
91 4.18.0-80.11.2.el8 (RHEL8.0)
92 4.18.0-147.8.1.el8 (RHEL8.1)
93 4.18.0-193.28.1.el8 (RHEL8.2)
94 4.18.0-240.22.1.el8 (RHEL8.3)
95 4.18.0-305.25.1.el8 (RHEL8.4)
96 4.4.120-92.70 (SLES12 SP2)
97 4.4.180-94.100 (SLES12 SP3)
98 4.4.0-131 (Ubuntu 16.04)
99 4.15.0-32 (Ubuntu 18.04)
100 5.4.0-48 (Ubuntu 20.04)
101 vanilla linux 5.4.0 (ZFS + ldiskfs)
102 vanilla linux 5.4.21 (ZFS + ldiskfs)
103 vanilla linux 5.4.136 (ZFS + ldiskfs)
104 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
105 * Client primary kernels built and tested during release cycle:
106 4.18.0-348.2.1.el8 (RHEL8.5)
107 5.4.0-37 (Ubuntu 20.04)
108 5.3.18-59.27 (SLES15 SP3)
109 * Other clients known to build on these kernels at some point (others may also work):
110 3.10.0-862.14.4.el7 (RHEL7.5)
111 3.10.0-957.27.2.el7 (RHEL7.6)
112 4.14.0-49.13.1.el7a (RHEL7.5)
113 4.14.0-115.2.2.el7a (RHEL7.6)
114 3.10.0-1062.18.1.el7 (RHEL7.7)
115 3.10.0-1127.19.1.el7 (RHEL7.8)
116 3.10.0-1160.49.1.el7 (RHEL7.9)
117 4.18.0-80.11.2.el8 (RHEL8.0)
118 4.18.0-147.8.1.el8 (RHEL8.1)
119 4.18.0-193.28.1.el8 (RHEL8.2)
120 4.18.0-240.22.1.el8 (RHEL8.3)
121 4.18.0-305.25.1.el8 (RHEL8.4)
122 4.4.120-92.70 (SLES12 SP2)
123 4.4.180-94.100 (SLES12 SP3)
124 4.12.14-95.48 (SLES12 SP4)
125 4.12.14-122.91 (SLES12 SP5)
126 4.12.14-197.75 (SLES15 SP1)
127 5.3.18-24.96 (SLES15 SP2)
128 4.4.0-131 (Ubuntu 16.04)
129 4.15.0-48 (Ubuntu 18.04)
130 5.8.0-53 (Ubuntu 20.04.2 HWE)
131 5.11.0-31 (Ubuntu 20.04.3 HWE)
132 5.11.0 (vanilla kernel.org)
133 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
134 * Recommended ZFS version: 2.0.7
135 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
136 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
137 stack overflow. For more information, please refer to bugzilla 17630.
138 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
139 the re-exporting nodes to be mounted with "32bitapi" mount option
141 --------------------------------------------------------------------------------
145 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
146 for currently supported client and server kernel versions.
147 * Server primary kernels built and tested during release cycle:
148 3.10.0-1062.18.1.el7 (RHEL7.7)
149 3.10.0-1127.19.1.el7 (RHEL7.8)
150 3.10.0-1160.6.1.el7 (RHEL7.9)
151 * Other server kernels known to build and work at some point (others may also work):
152 3.10.0-862.14.4.el7 (RHEL7.5)
153 3.10.0-957.27.2.el7 (RHEL7.6)
154 4.18.0-80.11.2.el8 (RHEL8.0)
155 4.18.0-147.8.1.el8 (RHEL8.1)
156 4.18.0-193.28.1.el8 (RHEL8.2)
157 4.18.0-240.1.1.el8 (RHEL8.3)
158 4.4.120-92.70 (SLES12 SP2)
159 4.4.180-94.100 (SLES12 SP3)
160 4.4.0-131 (Ubuntu 16.04)
161 4.15.0-32 (Ubuntu 18.04)
162 5.4.0-48 (Ubuntu 20.04)
163 vanilla linux 5.4.0 (ZFS + ldiskfs)
164 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
165 * Client primary kernels built and tested during release cycle:
166 3.10.0-1062.18.1.el7 (RHEL7.7)
167 3.10.0-1127.19.1.el7 (RHEL7.8)
168 3.10.0-1160.6.1.el7 (RHEL7.9)
169 4.12.14-95.48 (SLES12 SP4)
170 4.12.14-122.20 (SLES12 SP5)
171 4.12.14-197.75 (SLES15 SP1)
172 5.3.18-24.24 (SLES15 SP2)
173 4.15.0-48 (Ubuntu 18.04)
174 * Other clients known to build on these kernels at some point (others may also work):
175 3.10.0-862.14.4.el7 (RHEL7.5)
176 3.10.0-957.27.2.el7 (RHEL7.6)
177 4.14.0-49.13.1.el7a (RHEL7.5)
178 4.14.0-115.2.2.el7a (RHEL7.6)
179 4.18.0-80.11.2.el8 (RHEL8.0)
180 4.18.0-147.8.1.el8 (RHEL8.1)
181 4.18.0-193.28.1.el8 (RHEL8.2)
182 4.18.0-240.1.1.el8 (RHEL8.3)
183 4.4.120-92.70 (SLES12 SP2)
184 4.4.180-94.100 (SLES12 SP3)
185 4.4.0-131 (Ubuntu 16.04)
186 5.4.0-37 (Ubuntu 20.04)
187 5.4.0 (vanilla kernel.org)
188 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
189 * Recommended ZFS version: 2.0.0
190 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
191 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
192 stack overflow. For more information, please refer to bugzilla 17630.
193 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
194 the re-exporting nodes to be mounted with "32bitapi" mount option
196 --------------------------------------------------------------------------------
200 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
201 for currently supported client and server kernel versions.
202 * Server primary kernels built and tested during release cycle:
203 3.10.0-1062.1.1.el7 (RHEL7.7)
204 * Other server kernels known to build and work at some point (others may also work):
205 3.10.0-862.14.4.el7 (RHEL7.5)
206 3.10.0-957.27.2.el7 (RHEL7.6)
207 4.18.0-80.7.1.el8 (RHEL8.0)
208 4.4.120-92.70 (SLES12 SP2)
209 4.4.180-94.100 (SLES12 SP3)
210 4.4.0-131 (Ubuntu 16.04)
211 4.15.0-32 (Ubuntu 18.04)
212 vanilla linux 4.6.7 (ZFS only)
213 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
214 * Client primary unpatched kernels built and tested during release cycle:
215 3.10.0-1062.1.1.el7 (RHEL7.7)
216 4.12.14-95.29 (SLES12 SP4)
217 4.15.0-48 (Ubuntu 18.04)
218 * Other clients known to build on unpatched kernels at some point (others may also work):
219 3.10.0-862.14.4.el7 (RHEL7.5)
220 3.10.0-957.27.2.el7 (RHEL7.6)
221 4.14.0-49.13.1.el7a (RHEL7.5)
222 4.14.0-115.2.2.el7a (RHEL7.6)
223 4.18.0-80.7.1.el8 (RHEL8.0)
224 4.4.120-92.70 (SLES12 SP2)
225 4.4.180-94.100 (SLES12 SP3)
226 4.4.0-131 (Ubuntu 16.04)
227 vanilla linux 4.15.0 (ZFS only)
228 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
229 * Recommended ZFS / SPL version: 0.7.13
230 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
231 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
232 stack overflow. For more information, please refer to bugzilla 17630.
233 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
234 the re-exporting nodes to be mounted with "32bitapi" mount option
235 * Note for Ubuntu 14.04.5:
236 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
237 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
239 --------------------------------------------------------------------------------
243 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
244 for currently supported client and server kernel versions.
245 * Server known to build on the following kernels (others may also work):
246 2.6.32-696.18.7.el6 (RHEL6.9)
247 3.10.0-862.14.4.el7 (RHEL7.5)
248 3.10.0-957.el7 (RHEL7.6)
249 4.4.120-92.70 (SLES12 SP2)
250 4.4.162-94.69 (SLES12 SP3)
251 4.4.0-131 (Ubuntu 16.04)
252 4.15.0-32 (Ubuntu 18.04)
253 vanilla linux 4.6.7 (ZFS only)
254 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
255 * Client known to build on unpatched kernels (others may also work):
256 2.6.32-696.18.7.el6 (RHEL6.9)
257 2.6.32-754.6.3.el6 (RHEL6.10)
258 3.10.0-862.14.4.el7 (RHEL7.5)
259 3.10.0-957.el7 (RHEL7.6)
260 4.14.0-49.13.1.el7a (RHEL7.5)
261 4.14.0-115.2.2.el7a (RHEL7.6)
262 4.4.120-92.70 (SLES12 SP2)
263 4.4.162-94.69 (SLES12 SP3)
264 4.4.0-131 (Ubuntu 16.04)
265 4.15.0-32 (Ubuntu 18.04)
266 vanilla linux 4.15.0 (ZFS only)
267 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
268 * Recommended ZFS / SPL version: 0.7.9
269 * Tested with ZFS / SPL version: 0.7.9
270 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
271 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
272 stack overflow. For more information, please refer to bugzilla 17630.
273 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
274 the re-exporting nodes to be mounted with "32bitapi" mount option
275 * Note for Ubuntu 14.04.5:
276 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
277 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
279 --------------------------------------------------------------------------------
281 2018-04-03 Intel Corporation
283 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
284 for currently supported client and server kernel versions.
285 * Server known to build on patched kernels:
286 2.6.32-431.29.2.el6 (RHEL6.5)
287 2.6.32-504.30.3.el6 (RHEL6.6)
288 2.6.32-573.26.1.el6 (RHEL6.7)
289 2.6.32-642.15.1.el6 (RHEL6.8)
290 2.6.32-696.18.7.el6 (RHEL6.9)
291 3.10.0-693.21.1.el7 (RHEL7.4)
292 3.0.101-0.47.71 (SLES11 SP3)
293 3.0.101-107 (SLES11 SP4)
294 3.12.74-60.64.40 (SLES12 SP1)
295 4.4.114-92.64 (SLES12 SP2)
296 4.4.114-94.11 (SLES12 SP3)
297 3.13.0-101 (Ubuntu 14.04, ZFS only)
298 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
299 4.4.0-116 (Ubuntu 16.04)
300 vanilla linux 4.6.7 (ZFS only)
301 * Client known to build on unpatched kernels:
302 2.6.32-431.29.2.el6 (RHEL6.5)
303 2.6.32-504.30.3.el6 (RHEL6.6)
304 2.6.32-573.26.1.el6 (RHEL6.7)
305 2.6.32-642.15.1.el6 (RHEL6.8)
306 2.6.32-696.18.7.el6 (RHEL6.9)
307 3.10.0-693.21.1.el7 (RHEL7.4)
308 3.0.101-0.47.71 (SLES11 SP3)
309 3.0.101-107 (SLES11 SP4)
310 3.12.74-60.64.40 (SLES12 SP1)
311 4.4.114-92.64 (SLES12 SP2)
312 4.4.114-94.11 (SLES12 SP3)
313 3.13.0-101 (Ubuntu 14.04)
314 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
315 4.4.0-116 (Ubuntu 16.04)
317 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
318 * Recommended ZFS / SPL version: 0.7.6
319 * Tested with ZFS / SPL version: 0.7.6
320 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
321 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
322 stack overflow. For more information, please refer to bugzilla 17630.
323 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
324 the re-exporting nodes to be mounted with "32bitapi" mount option
325 * Note for Ubuntu 14.04.5:
326 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
327 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
329 --------------------------------------------------------------------------------
331 2017-07-05 Intel Corporation
333 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
334 for currently supported client and server kernel versions.
335 * Server known to build on patched kernels:
336 2.6.32-431.29.2.el6 (RHEL6.5)
337 2.6.32-504.30.3.el6 (RHEL6.6)
338 2.6.32-573.26.1.el6 (RHEL6.7)
339 2.6.32-642.15.1.el6 (RHEL6.8)
340 2.6.32-696.3.1.el6 (RHEL6.9)
341 3.10.0-514.21.1.el7 (RHEL7.3)
342 3.0.101-0.47.71 (SLES11 SP3)
343 3.0.101-100 (SLES11 SP4)
344 3.12.74-60.64.40 (SLES12 SP1)
345 4.4.59-92.17 (SLES12 SP2)
346 vanilla linux 4.6.7 (ZFS only)
347 * Client known to build on unpatched kernels:
348 2.6.32-431.29.2.el6 (RHEL6.5)
349 2.6.32-504.30.3.el6 (RHEL6.6)
350 2.6.32-573.26.1.el6 (RHEL6.7)
351 2.6.32-642.15.1.el6 (RHEL6.8)
352 2.6.32-696.3.1.el6 (RHEL6.9)
353 3.10.0-514.21.1.el7 (RHEL7.3)
354 3.0.101-0.47.71 (SLES11 SP3)
355 3.0.101-100 (SLES11 SP4)
356 3.12.74-60.64.40 (SLES12 SP1)
357 4.4.59-92.17 (SLES12 SP2)
359 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
360 * Recommended ZFS / SPL version: 0.6.5.9
361 * Tested with ZFS / SPL version: 0.6.5.9
362 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
363 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
364 stack overflow. For more information, please refer to bugzilla 17630.
365 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
366 the re-exporting nodes to be mounted with "32bitapi" mount option
368 --------------------------------------------------------------------------------
370 12-07-2016 Intel Corporation
372 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
373 for currently supported client and server kernel versions.
374 * Server known to build on patched kernels:
375 2.6.32-431.29.2.el6 (RHEL6.5)
376 2.6.32-504.30.3.el6 (RHEL6.6)
377 2.6.32-573.26.1.el6 (RHEL6.7)
378 2.6.32-642.11.1.el6 (RHEL6.8)
379 3.10.0-514.el7 (RHEL7.3)
380 3.0.101-0.47.71 (SLES11 SP3)
381 3.0.101-80 (SLES11 SP4)
382 3.12.62-60.64.8 (SLES12 SP1)
383 vanilla linux 4.5.7 (ZFS only)
384 * Client known to build on unpatched kernels:
385 2.6.32-431.29.2.el6 (RHEL6.5)
386 2.6.32-504.30.3.el6 (RHEL6.6)
387 2.6.32-573.26.1.el6 (RHEL6.7)
388 2.6.32-642.11.1.el6 (RHEL6.8)
389 3.10.0-514.el7 (RHEL7.3)
390 3.0.101-0.47.71 (SLES11 SP3)
391 3.0.101-80 (SLES11 SP4)
392 3.12.62-60.64.8 (SLES12 SP1)
394 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
395 * Recommended ZFS / SPL version: 0.6.5.8
396 * Tested with ZFS / SPL version: 0.6.5.8
397 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
398 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
399 stack overflow. For more information, please refer to bugzilla 17630.
400 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
401 the re-exporting nodes to be mounted with "32bitapi" mount option
403 --------------------------------------------------------------------------------
405 02-29-2016 Intel Corporation
407 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
408 for currently supported client and server kernel versions.
409 * Server known to build on patched kernels:
410 2.6.32-431.29.2.el6 (RHEL6.5)
411 2.6.32-504.30.3.el6 (RHEL6.6)
412 2.6.32-573.12.1.el6 (RHEL6.7)
413 3.10.0-327.3.1.el7 (RHEL7.2)
414 3.0.101-0.47.71 (SLES11 SP3)
415 3.0.101-68 (SLES11 SP4)
416 vanilla linux 4.2.1 (ZFS only)
417 * Client known to build on unpatched kernels:
418 2.6.32-431.29.2.el6 (RHEL6.5)
419 2.6.32-504.30.3.el6 (RHEL6.6)
420 2.6.32-573.12.1.el6 (RHEL6.7)
421 3.10.0-327.3.1.el7 (RHEL7.2)
422 3.0.101-0.47.71 (SLES11 SP3)
423 3.0.101-68 (SLES11 SP4)
426 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
427 * Recommended ZFS / SPL version: 0.6.4.2
428 * Tested with ZFS / SPL version: 0.6.4.2
429 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
430 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
431 stack overflow. For more information, please refer to bugzilla 17630.
432 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
433 the re-exporting nodes to be mounted with "32bitapi" mount option
435 --------------------------------------------------------------------------------
437 03-10-2015 Intel Corporation
439 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
440 for currently supported client and server kernel versions.
441 * Server known to build on patched kernels:
442 2.6.32-431.29.2.el6 (RHEL6.5)
443 2.6.32-504.8.1.el6 (RHEL6.6)
444 3.0.101-0.46 (SLES11 SP3)
445 * Client known to build on unpatched kernels:
446 2.6.32-431.29.2.el6 (RHEL6.5)
447 2.6.32-504.8.1.el6 (RHEL6.6)
448 3.10.0-123.20.1.el7 (RHEL7)
449 3.0.101-0.46 (SLES11 SP3)
450 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
451 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
452 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
453 stack overflow. For more information, please refer to bugzilla 17630.
454 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
455 the re-exporting nodes to be mounted with "32bitapi" mount option
457 Severity : enhancement
459 Description: control OST-index in IDIF via ROCOMPAT flag.
460 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
461 last_rcvd file. For new formatted OST device, it will be auto set;
462 for the case of upgrading from old OST device, you can enable it
463 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
464 enabled, for new created OST-object, its IDIF-in-LMA will contain
465 the OST-index; for the existing OST-object, the OSD will convert
466 old format IDIF as new format IDIF with OST-index stored in the
467 LMA EA when accessing such OST-object or via OI scrub. Once such
468 flag is enabled, it cannot be reverted back, so the system cannot
469 be downgraded to the orignal incompatible version.
471 --------------------------------------------------------------------------------
473 07-30-2014 Intel Corporation
475 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
476 for currently supported client and server kernel versions.
477 * Server known to build on patched kernels:
478 2.6.32-431.20.3.el6 (RHEL6.5)
479 3.0.101-0.31 (SLES11 SP3)
480 * Client known to build on unpatched kernels:
481 2.6.32-431.20.3.el6 (RHEL6.5)
482 3.0.101-0.31 (SLES11 SP3)
485 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
486 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
487 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
488 stack overflow. For more information, please refer to bugzilla 17630.
489 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
490 the re-exporting nodes to be mounted with "32bitapi" mount option
492 --------------------------------------------------------------------------------
494 10-31-2013 Intel Corporation
496 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
497 * Server builds for kernels:
498 2.6.32-358.18.1.el6 (RHEL6.4)
499 * Client builds for unpatched kernels:
500 2.6.32-358.18.1.el6 (RHEL6.4)
501 3.0.92-0.8 (SLES11 SP3)
503 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
504 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
505 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
506 stack overflow. For more information, please refer to bugzilla 17630.
507 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
508 the re-exporting nodes to be mounted with "32bitapi" mount option
510 --------------------------------------------------------------------------------
512 05-30-2013 Intel Corporation
514 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
515 * Server builds for kernels:
516 2.6.32-358.6.2.el6 (RHEL6.4)
517 * Client builds for unpatched kernels:
518 2.6.32-358.6.2.el6 (RHEL6.4)
519 3.0.74-0.6.6 (SLES11 SP2)
521 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
522 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
523 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
524 stack overflow. For more information, please refer to bugzilla 17630.
525 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
526 the re-exporting nodes to be mounted with "32bitapi" mount option
528 --------------------------------------------------------------------------------
530 9-30-2012 Whamcloud, Inc.
532 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
533 * Server support for kernels:
534 2.6.32-279.5.1.el6 (RHEL6)
535 * Client support for unpatched kernels:
536 2.6.32-279.5.1.el6 (RHEL6)
538 2.6.18-238.19.1.el5 (RHEL5)
539 2.6.16.60-0.69.1 (SLES10)
540 * Recommended e2fsprogs version: 1.41.90.wc4
541 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
542 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
543 stack overflow. For more information, please refer to bugzilla 17630.
544 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
545 the re-exporting nodes to be mounted with "32bitapi" mount option
547 Severity : enhancement
549 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
551 Severity : enhancement
553 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
555 Severity : enhancement
557 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
559 Severity : enhancement
561 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
563 Severity : enhancement
565 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
567 --------------------------------------------------------------------------------
569 04-30-2012 Whamcloud, Inc.
571 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
572 * Server support for kernels:
573 2.6.32-220.4.2.el6 (RHEL6)
574 * Client support for unpatched kernels:
575 2.6.32-220.4.2.el6 (RHEL6)
577 2.6.18-238.19.1.el5 (RHEL5)
578 2.6.16.60-0.69.1 (SLES10)
579 * Recommended e2fsprogs version: 1.41.90.wc4
580 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
581 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
582 stack overflow. For more information, please refer to bugzilla 17630.
583 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
584 the re-exporting nodes to be mounted with "32bitapi" mount option
586 Severity : enhancement
588 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
590 Severity : enhancement
592 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
594 Severity : enhancement
596 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
598 Severity : enhancement
600 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
602 --------------------------------------------------------------------------------
604 09-30-2011 Whamcloud, Inc.
606 * Server support for kernels:
607 2.6.18-238.19.1.el5 (RHEL5)
608 2.6.32-131.6.1.el6 (RHEL6)
609 * Client support for unpatched kernels:
610 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
611 2.6.18-238.19.1.el5 (RHEL5)
612 2.6.32-131.6.1.el6 (RHEL6)
613 2.6.29.4-167.fc11 (FC11)
614 2.6.16.60-0.69.1 (SLES10)
616 * Recommended e2fsprogs version: 1.41.90.wc3
617 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
618 Lustre filesystem with 4K stack may cause a stack overflow. For more
619 information, please refer to bugzilla 17630.
620 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
621 mounted with "32bitapi" mount option
623 Severity : enhancement
624 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
626 Severity : enhancement
627 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
629 Severity : enhancement
630 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
633 Frequency : only with SLES10
635 Description: Use OFED "KMP" provided by Novell
636 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
637 Lustre is now built against this package. That means you need to
638 install the ofed-kmp package from Novell for the patchless client.
639 Note that the ofed-kmp that Novell ships may not exactly match the
640 kernel version but should still be compatible.
642 Severity : enhancement
644 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
645 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
647 --------------------------------------------------------------------------------
649 2010-07-15 Oracle, Inc.
651 * Server support for kernels:
652 2.6.18-164.11.1.el5 (RHEL 5)
653 2.6.18-164.11.1.0.1.el5 (OEL 5)
654 * Client support for unpatched kernels:
655 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
656 2.6.18-164.11.1.el5 (RHEL 5),
657 2.6.18-164.11.1.0.1.el5 (OEL 5)
658 2.6.16.60-0.42.8 (SLES 10),
660 2.6.29.4-167.fc11 (FC11)
661 * Recommended e2fsprogs version: 1.41.10-sun2
663 Severity : enhancement
665 Description: Distribute lfs_migrate script with lustre
666 Details : The lfs_migrate mentioned in the manual is now included in the
667 lustre tree in lustre/scripts and will distributed in lustre RPMs.
669 Severity : enhancement
671 Description: Failover nid list added to import proc file.
673 Severity : enhancement
675 Description: "som_preview" mount option required to use SOM
676 Details : The clients will need to use som_preview mount option
680 Description: Lustre does not do 1MB IOs to HW RAID
681 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
683 Severity : enhancement
685 Description: lfs quota output is not very convenient for awk/sed-parsing
686 Details : Some positions in lfs quota output table could be empty or
687 non-empty which made it hard to parse it with scripts, now
688 a dash is put instead of space where there is not supposed
693 Description: Update to OFED 1.5.1
694 Details : OFED 1.5.1 will be used in environments that don't use
695 vendor supplied OFED.
699 Description: Remove lnet. prefix from lctl params display
700 Details : Also renamed lnet.memused to lnet_memused to
705 Description: Use conf_param -d to remove permanant settings.
709 Description: Add list_param -R
710 Details : lctl list_param now supports recursive mode.
713 Bugzilla : 21829, 22312
714 Description: Fix llobdstat.
715 Details : 1. Now updating the write_bytes counter
716 2. Port script fixes from 1.8.
720 Description: decrease the usage of memory on clients.
721 Details : 1. On clients, recycle dentries and inodes unused.
722 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
727 Description: llobdstat fix and enhancement
728 Details : add a counter to set a limit to how many samples will be returned
729 fix a wildcard in the path to limit to obdfilter stats only
733 Description: lfs check only allowed for root
734 Details : Code cleanups and added a regresion test of non-root use of
739 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
740 Details : Fix hr->hr_index overload.
744 Description: Fix FIEMAP functionality.
748 Description: FC11 patchless client support added
752 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
753 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
757 Description: New acc-sm script and various test-framework enhancments.
761 Description: Previously landed patch was reverted.
765 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
769 Description: Add ldiskfs maxdirsize mount option.
773 Description: Fix mkfs.lustre support for 16TB LUNs.
777 Description: Print mmp_check_interval
778 Details : print mmp_check_interval and make it possible to abort
779 mount operation in case it takes too long.
783 Description: Add cascading_rw to lustre mpi tests.
787 Description: Store SOM data in EA.
788 Details : Also several SOM related cleanups.
792 Description: Add refcount to md_open_data
796 Description: Use refcount on lov_request_set to prevent memory corruption.
800 Description: Statahead: use parent dir's i_mutex for synchronization.
804 Description: loadgen improvements.
808 Description: Enable peer health detection in ksocklnd and ko2iblnd.
812 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
816 Description: Fix race on obd_devcie:md_stats
820 Description: Fix SLES 11 client builds.
821 Details : Resolved uninitalized pointer gcc warnings.
825 Description: Allow non-root access for "lfs check".
826 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
830 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
832 Severity : enhancement
834 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
835 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
839 Description: Sanity test_27p fails.
840 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
841 Sanity tests 27 were cleaned up and improved.
845 Description: Unify req format on client/servers
846 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
847 buffers and swabbers approach..
852 Description: drop join file stuff.
854 Severity : enhancement
856 Description: Add 2.6.27 and SLES11 (patchless client) support
860 Description: Fix in ptlrpc_expire_one_request() to print the signed time
863 Severity : enhancement
865 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
868 Severity : enhancement
870 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
872 Severity : enhancement
874 Description: correctly handle big reply message.
875 Details : send LNet event if reply is bigger than buffer and adjust this buffer
880 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
882 Severity : enhancement
884 Description: Add support for OEL5.
888 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
890 Severity : enhancement
892 Description: "weak-modules" support
893 Details : Implement "weak-modules" support which enables kernel modules
894 to be used with any kernel that implements the same kABI. In
895 order to achieve this modules are now installed in
896 /lib/modules/$(uname -r)/updates/kernel on all distributions.
900 Description: client could not reconnect to OST because of an active request.
901 Details : The client now retries to reconnect to the same server, if a
902 connect request failed with EBUSY or -EAGAIN. Test case 35b
903 conf-sanity.sh has been created.
907 Description: Don't allow make backward step on assignin osc next id.
908 Details : race between allocation next id and ll_sync thread can be cause
909 of set wrong osc next id and can be kill valid ost objects.
913 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
917 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
921 Description: Error handling in osc_statfs_interpret() has been improved.
922 Details : Check in osc_statfs_interpret() for EBADR.
926 Description: Do not send statfs() requests to OSTs disabled by administrator.
927 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
929 Severity : enhancement
931 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
933 Severity : enhancement
935 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
937 Severity : enhancement
939 Description: Enhancement for lfs(1) command to use numeric uid/gid.
943 Description: Improved cleanup procedure for conf-sanity test_46a.
947 Description: Do not start bulk transfer, if client tried to reconnect.
948 Details : Do not start bulk transfer, if client tried to reconnect,
949 but failed due to active rpcs. Check for exp_abort_active_req
950 has been implemented.
954 Description: Prevent inconsistences between linux and lustre mount structures.
955 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
956 Make the sleep interruptible.
958 Severity : enhancement
960 Description: provide server to client comms path
961 Details : server to client communications path via new LDLM_SET_INFO rpc
965 Description: truncate starts GFP_FS allocation under transaction causing deadlock
966 Details : ldiskfs_truncate calls grab_cache_page which may start page
967 allocation under an open transaction. This may lead to
968 calling prune_icache with consequent lustre reentrance.
972 Description: send correctly lsm on open replay
973 Details : MDS is trust to LSM size on replay open, but client can set wrong size
978 Description: enable client interface failover
979 Details : When a child reconnects from another NID, properly update export
980 nid hash position and ldlm reverse import.
982 Severity : enhancement
984 Description: Communicate OST degraded/readonly state via statfs to MDS
985 Details : Flags in the statfs returned from OSTs indicate whether the
986 OST is in a degraded RAID state, or if the filesystem has
987 turned read-only after a filesystem error is detected.
992 Description: don't panic if EPROTO was hit when reading symlink
993 Details : correctly handling request reference in error cases.
995 Severity : enhancement
997 Description: Add LustreNetLink, a kernel-userspace communcation path.
999 Severity : enhancement
1001 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1006 Description: access to llog context before init.
1007 Details : move handling CATALOGS file at osc layer and forbid access to llog
1008 context before init.
1011 Frequency : rare, only when client is writing to OST on local node
1013 Description: Avoid deadlock for local client writes
1014 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1015 memory freeing context. This allows OST threads to set the
1016 PF_MEMALLOC flag on task structures in order to allocate memory
1017 from reserved pools and complete IO.
1018 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1019 so that the OST threads generate memory pressure and allow
1020 inactive pages to be reclaimed.
1022 Severity : enhancement
1024 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1029 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1030 Details : move ldlm namespace creation in setup phase to avoid grab
1031 _lprocfs_lock with cli_sem held.
1035 Description: Temporarily disable grant shrink.
1036 Details : Disable the feature for debugging.
1039 Frequency : only when formatting test filesystems
1041 Description: Unable to run several mkfs.lustre on loop devices at the same time
1042 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1043 formatting. The solution is to proper handle the error.
1045 Severity : enhancement
1047 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1049 Severity : enhancement
1051 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1056 Description: resolve race between obd_disconnect and class_disconnect_exports
1057 Details : if obd_disconnect will be called to already disconnected export he
1058 forget release one reference and osc module can't unloaded.
1060 Severity : enhancement
1062 Description: Allow tuning service thread via /proc
1063 Details : For each service a new
1064 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1065 created that can be used to set min/max thread counts, and get the
1066 current number of running threads.
1071 Descriptoin: don't return error if have partially created objects for file.
1072 Details : lov_update_create_set uses set->set_success as index for created
1073 objects, so if some requests will be failed, they will have hole
1074 at end of array and qos_shrink_lsm for allocate correct lsm.
1076 Severity : enhancement
1078 Description: Update OFED support to 1.4.1
1083 Description: Reduce small size read RPC
1084 Details : Set read-ahead limit for every file and only do read-ahead when
1085 available read-ahead pages are bigger than 1M to avoid small size
1088 Severity : enhancement
1090 Description: move AT tunable parameters for more consistent usage
1091 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1093 Severity : enhancement
1095 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1096 Details : allow skip disconnected ost for send statfs request and hide error
1102 Description: fix racy locking of mballoc block bitmaps causing BUG
1103 Details : The locking of the mballoc buddy bitmap and the in-memory
1104 block bitmap was using two different spin locks in some
1105 cases. This made it possible to incorrectly access the
1106 mballoc bitmap while another process was modifying it,
1107 causing a sanity assertion to fail. While no on-disk corruption
1108 was reported, there was some risk of this happening.
1111 Frequency : rare, on llog test 6
1113 Descriptoin: don't allow connect to already connected import
1114 Details : allowing connect to already connected import is hide connecting problem.
1119 Descriptoin: allow kill process which wait statahead result
1120 Details : for some reasons 'ls' can stick in waiting result from statahead,
1121 in this case need way for kill this process.
1123 Severity : enhancement
1125 Description: Add state history info file, enhance import info file
1126 Details : Track import connection state changes in a new osc/mdc proc file;
1127 add overview-type data to the osc/mdc import proc file.
1129 Severity : enhancement
1131 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1133 Severity : enhancement
1135 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1140 Descriptoin: OST grant too much space to client even there are not enough space.
1141 Details : Client will shrink its grant cache to OST if there are no write
1142 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1143 this grant cache if there are already not enough avaible space
1144 (left_space < total_clients * 32M).
1147 Frequency : start MDS on uncleanly shutdowned MDS device
1149 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1150 Details : stay in waiting mds<>ost recovery finished produce random bugs
1151 due race between two ll_sync thread for one lov target. send
1152 ACTIVATE event only if connect realy finished and import have
1156 Frequency : rare, connect and disconnect target at same time
1158 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1159 Details : don't call obd_disconnect under lov_lock. this long time
1160 operation and can block ptlrpcd which answer to connect request.
1165 Descriptoin: don't lose wakeup for imp_recovery_waitq
1166 Details : recover_import_no_retry or invalidate_import and import_close can
1167 both sleep on imp_recovery_waitq, but we was send only one wakeup
1171 Frequency : always with long access acl
1173 Descriptoin: mds can't pack reply with long acl.
1174 Details : mds don't control size of acl but they limited by reint/getattr
1177 Severity : enhancement
1179 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1181 Severity : enhancement
1183 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1186 Frequency : start MDS on uncleanly shutdowned MDS device
1188 Descriptoin: aborting recovery hang on MDS
1189 Details : don't throttle destroy RPCs for the MDT.
1192 Frequency : on remount
1194 Description: external journal device not working after the remount
1195 Details : clear dev_rdonly flag for external journal devices in
1201 Description: shutdown vs evict race
1202 Details : client_disconnect_export vs connect request race.
1203 if client will evicted at this time - we start invalidate
1204 thread without referece to import and import can be freed
1208 Frequency : rare, need acl's on inode.
1210 Description: client can't handle ost additional correctly
1211 Details : if ost was added after client connected to mds client can have
1212 hit lnet_try_match_md ... to big messages to wide striped files.
1213 in this case need teach client to handle config events about add
1214 lov target and update client max ea size at that event.
1216 Severity : enhancement
1218 Description: Changelogs
1219 Details : Changelogs are a lightweight mechanism to track filesystem
1220 metadata and namespace changes. The changelog is recorded
1221 permanently on the MDTs, and is periodically "consumed" / purged
1222 when records are no longer needed.
1224 Severity : enhancement
1226 Description: compact fld format with extents
1227 Details : Store range of seq rather than every seq in FLD. Seq
1228 controller update FLD rather than clients. In Case of CMD, mdt0
1229 has FLD, all other metadata server act as non persistent proxy
1230 for FLD queries and cache fld entries in fld cache.
1235 Description: don't skip ost target if they assigned to file
1236 Details : Drop slow OSCs if we can, but not for requested start idx.
1237 This means "if OSC is slow and it is not the requested
1238 start OST, then it can be skipped, otherwise skip it only
1239 if it is inactive/recovering/out-of-space.
1243 Description: more cleanup in mds_lov
1244 Details : not send LOV EA under replay, we can't know about they size at this
1245 time. Don't allow client connect to mds before any ost connected,
1246 for avoid problems with LOV EA size and returning EIO to client.
1248 Severity : enhancement
1250 Description: Interoperability at server side (Disk interoperability)
1252 Severity : enhancement
1254 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1256 Severity : enhancement
1258 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1260 Severity : enhancement
1262 Description: New client IO stack (CLIO).
1264 Severity : enhancement
1266 Description: Commit on sharing. Eliminate inter-client dependencies between
1267 uncommitted transactions by doing transaction commits.
1268 Thereby clients may recovery independently.
1271 Frequency : Create a symlink file with a very long name
1273 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1274 Details : If there is no extra space in the request for early cancels,
1275 ldlm_req_handles_avail() returns 0 instead of a negative value.
1277 Severity : enhancement
1279 Description: Add /proc entry for import status
1280 Details : The mdc, osc, and mgc import directories now have
1281 an import directory that contains useful import data for debugging
1282 connection problems.
1284 Severity : enhancement
1286 Description: Re-disable certain /proc logging
1287 Details : Enable and disable client's offset_stats, extents_stats and
1288 extents_stats_per_process stats logging on the fly.
1291 Frequency : Only on FC kernels 2.6.22+
1293 Description: oops in statahead
1294 Details : Do not drop reference count for the dentry from VFS when lookup,
1295 VFS will do that by itself.
1297 Severity : enhancement
1299 Description: Generic /proc file permissions
1300 Details : Set /Proc file permissions in a more generic way to enable non-
1301 root users operate on some /proc files.
1305 Description: Hitting mdc_commit_close() ASSERTION
1306 Details : Properly handle request reference release in
1307 ll_release_openhandle().
1311 Description: quota recovery deadlock during mds failover
1312 Details : This patch includes att18982, att18236, att18237 in bz14840.
1314 1. fix osts hang when mds does failover with quotaon
1315 2. prevent watchdog storm when osts threads wait for the
1320 Frequency : only patchless client
1321 Description: add workaround for race between add/remove dentry from hash
1323 Severity : enhancement
1325 Description: Allow OST glimpses to return PW locks
1329 Description: LBUG when llog conf file is full
1330 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1334 Description: Prevent import from entering FULL state when server in recovery
1338 Description: service mount cannot take device name with ":"
1339 Details : Only when device name contains ":/" will mount treat it as
1345 Description: replace ptlrpcd with the statahead thread to interpret the async
1346 statahead RPC callback
1350 Frequency : on recovery
1351 Description: I/O failures after umount during fail back
1352 Details : if client reconnected to restarted server we need join to recovery
1353 instead of find server handler is changed and process self
1354 eviction with cancel all locks.
1356 Severity : enhancement
1358 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1360 Severity : enhancement
1362 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1364 Severity : enhancement
1366 Description: Upcall on Lustre log has been dumped
1367 Details : Allow for a user mode script to be called once a Lustre log has
1368 been dumped. It passes the filename of the dumped log to the
1369 script, the location of the script can be specified via
1370 /proc/sys/lnet/debug_log_upcall.
1375 Description: avoid idr_remove called for id which is not allocated.
1376 Details : Move assigment s_dev for clustered nfs to end of initialization,
1377 to avoid problem with error handling.
1382 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1383 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1384 into CONN_USED_HASH and this prodice warning when put connection
1385 again in unused hash.
1388 Severity : enhancement
1390 Description: File striping can now be set to use an arbitrary pool of OSTs.
1392 Severity : enhancement
1394 Description: Export bytes_read/bytes_write count on OSC/OST.
1398 Description: Early reply size mismatch, MGC loses connection
1399 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1400 the connect flags are properly negotiated.
1405 Description: quotas are not honored with O_DIRECT
1406 Details : all writes with the flag O_DIRECT will use grants which leads to
1407 this problem. Now using OBD_BRW_SYNC to guard this.
1411 Description: add quota statistics
1412 Details : 1. sort out quota proc entries and proc code.
1413 2. add quota statistics
1415 Severity : enhancement
1417 Description: enable quota support for HEAD.
1421 Description: Properly propagate oinfo flags from lov to osc for statfs
1422 Details : restore missing copy oi_flags to lov requests.
1424 Severity : enhancement
1426 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1427 : l_getgroups(8), lst(8), routerstat(8)
1428 Details : included man pages for llobdstat(8), llstat(8),
1429 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1431 Severity : enhancement
1433 Description: configure's --enable-quota should check the
1434 : kernel .config for CONFIG_QUOTA
1435 Details : configure is terminated if --enable-quota is passed but
1436 : no quota support is in kernel
1440 Description: Remove portals compatibility
1441 Details : Remove portals compatibility, not interoperable with releases
1446 Description: Resolve device initialization race
1447 Details : Prevent proc handler from accessing devices added to the
1448 obd_devs array but yet be intialized.
1450 Severity : enhancement
1452 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1454 Severity : enhancement
1456 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1461 Description: Using wrong pointer in osc_brw_prep_request
1462 Details : Access to array[-1] can produce panic if kernel compiled with
1463 CONFIG_PAGE_ALLOC enabled
1467 Description: Client runs out of low memory
1468 Details : Consider only lowmem when counting initial number of llap pages
1472 Description: *optional* service tags registration
1473 Details : if the "service tags" package is installed on a Lustre node
1474 When the filesystem is mounted, a local-node service tag will
1475 be created. See http://inventory.sun.com/ for more information
1476 about the Service Tags asset management system.
1480 Description: Kernel BUG tries to release flock
1481 Details : Lustre does not destroy flock lock before last reference goes
1482 away. So always drop flock locks when client is evicted and
1483 perform unlock regardless of successfulness of speaking to MDS.
1487 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1491 Description: sanity test 65a fails if stripecount of -1 is set
1492 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1497 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1498 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1501 Severity : enhancement
1503 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1507 Description: do not process already freed flock
1508 Details : flock can possibly be freed by another thread before it reaches
1509 to ldlm_flock_completion_ast.
1513 Description: LBUG during stress test
1514 Details : Need properly lock accesses the flock deadlock detection list.
1518 Description: oops in page fault handler
1519 Details : kernel page fault handler can return two special 'pages' in error
1520 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1524 Description: timeout with invalidate import.
1525 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1526 should be handled by ptlrpcd. This produce long age waiting and
1527 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1529 Severity : enhancement
1531 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1535 Description: enable MGS and MDT services start separately
1536 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1537 a co-located MGS without starting the MGS, which is a complement
1538 to 'nosvc' mount option.
1542 Frequency : after recovery
1543 Description: precreate to many object's after del orphan.
1544 Details : del orphan st in oscc last_id == next_id and this triger growing
1545 count of precreated objects. Set flag LOW to skip increase count
1546 of precreated objects.
1550 Frequency : rare, on clear nid stats
1551 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1552 Details : when clean nid stats sometimes try destroy live entry,
1553 and this produce panic in free.
1557 Description: Stack overflow during MDS log replay
1558 ease stack pressure by using a thread dealing llog_process.
1562 Description: wait until IO finished before start new when do lock cancel.
1563 Details : VM protocol want old IO finished before start new, in this case
1564 need wait until PG_writeback is cleared until check dirty flag and
1565 call writepages in lock cancel callback.
1567 Severity : enhancement
1569 Description: using special macro for print time and cleanup in includes.
1573 Description: mds_mfd_close() ASSERTION(rc == 0)
1574 Details : In mds_mfd_close(), we need protect inode's writecount change
1575 within its orphan write semaphore to prevent possible races.
1579 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1583 Frequency : rare, on shutdown ost
1584 Description: don't hit live lock with umount ost.
1585 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1586 use shrink_dcache_sb instead.
1590 Description: don't panic with use echo client
1591 Details : echo client pass NULL as client nid pointer and this produce null
1592 pointer dereference.
1596 Description: fix build on ppc32
1597 Details : compile code with -m64 flag produce wrong object file for ppc32.
1601 Description: add message levels for liblustreapi
1605 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1606 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1607 will result in an -EAGAIN and in the exit call path we call
1608 alloc_rr() which will with increasing aggressiveness attempt to
1609 aquire precreated objects on the minimum number of required OSCs.
1613 Description: Use old size assignment to avoid deadlock
1614 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1615 the scheduling while holding a spinlock. We do not need locking
1616 for size in ll_update_inode() because size is only updated from
1617 the MDS for directories or files without objects, so there is no
1618 other place to do the update, and concurrent access to such inodes
1619 are protected by the inode lock.
1623 Description: resolve "_IOWR redefined" build error on SLES10
1627 Description: dump the memory debugging after all modules are unloaded to
1628 suppress false negative in conf_sanity test 39
1630 Severity : enhancement
1632 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1637 Description: fix SLES kernel versioning
1638 Details : the kernel version for our SLES 10 kernel did not include a "-"
1639 before the "smp" at the end. while this was not a problem in
1640 general, it did mean that software trying to use the kernel
1641 version to try to detect a vendor specific kernel would fail.
1642 this was most evident by the OFED build scripts.
1646 Description: Don't update lov_desc members until making sure they are valid
1647 Details : When updating lov_desc members via proc fs, need fix their
1648 validities before doing the real update.
1652 Description: don't put request into delay list while invalidate in flight.
1653 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1654 import in flight. this produce timeout for invalidate and sometimes
1655 can cause stale data.
1659 Frequency : on ppc only
1660 Description: not convert ost objects for directory because it's not exist.
1661 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1665 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1667 Severity : enhancement
1669 Description: Update to support 2.6.22.14 vanilla kernel.
1673 Frequency : rare, on recovery
1674 Description: read procfs can produce deadlock in some situation
1675 Details : Holding lprocfs lock which send rpc can produce block for destroy
1676 obd objects and this also block reconnect with -EALREADY. This
1677 isn't fix all lprocfs bugs - but make it rare.
1679 Severity : enhancement
1681 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1684 Frequency : frequent on X2 node
1686 Description: mdc_set_open_replay_data LBUG
1687 Details : Set replay data for requests that are eligible for replay.
1691 Description: lustre_mgs: operation 101 on unconnected MGS
1692 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1693 MGC, and late on MGC cannot successfully connect to MGS and a lot
1694 of the error messages complaining that MGS is not connected.
1698 Frequency : on network error
1699 Description: panic with double free request if network error
1700 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1701 but it's true only for synchronus enqueue, for async enqueue
1702 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1705 Severity : enhancement
1707 Description: client-side metadata stat-ahead during readdir(directory readahead)
1708 Details : perform client-side metadata stat-ahead when the client detects
1709 readdir and sequential stat of dir entries therein
1712 Frequency : on start mds
1714 Description: Implement get_info(last_id) in obdfilter.
1717 Frequency : occasional
1719 Description: Correctly check stale fid, not start epoch if ost not support SOM
1720 Details : open with flag O_CREATE need set old fid in op_fid3 because
1721 op_fid2 was overwritten with new generated fid, but mds can answer
1722 with one of these two fids and both is not stale. Setattr
1723 incorrectly started an epoch and assume will be called
1724 done_writing, but without SOM done_writing ever being called.
1727 Frequency : rare, depends on device drivers and load
1729 Description: MDS or OSS nodes crash due to stack overflow
1730 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1731 In some cases, in conjunction with device drivers that use a lot
1732 of stack the MDS (or possibly OSS) service threads could overflow
1733 the stack. One change which was identified to consume additional
1734 stack has been reworked to avoid the extra stack usage.
1737 Frequency : occasional
1739 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1740 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1741 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1742 oscc_has_objects and see OSC has no objests and call
1743 oscc_internal_create to resend create request.
1745 Severity : enhancement
1747 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1749 Severity : enhancement
1751 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1754 Frequency : very rare
1756 Description: Fix replay if there is an un-replied request and open
1757 Details : In some cases, older replay request will revert the
1758 mcd->mcd_last_xid on MDS which is used to record the client's
1759 latest sent request.
1761 Severity : enhancement
1763 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1765 Severity : enhancement
1767 Description: Add rhel5 support to HEAD.
1769 Serverity : enhancement
1771 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1776 Description: Don't allow skipping OSTs if index has been specified.
1777 Details : Don't allow skipping OSTs if index has been specified, make
1778 locking in internal create lots better.
1782 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1783 Details : access to bitfield in structure is always rounded to long
1784 and this produce problem with not atomic change any bit.
1788 Description: Lustre make rpms failed.
1789 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1790 builds Lustre rpms from tarball.
1793 Frequency : rare on shutdown ost
1795 Description: If llog cancel was not send before clean_exports phase, this can
1796 produce deadlock in llog code.
1797 Details : If llog thread has last reference to obd and call class_import_put
1798 this produce deadlock because llog_cleanup_commit_master wait when
1799 last llog_commit_thread exited, but this never success because was
1800 called from llog_commit_thread.
1804 Description: allow userland application know is lost one of stripes.
1805 Details : fill lvb_blocks with error code on ost and return it to
1806 application if error flag found.
1810 Description: NULL lov_tgts causing MDS oops
1811 Details : more safe checks for NULL lov_tgts for avoid oops.
1813 Severity : enhancement
1815 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1819 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1820 Details : If an OST is down the MDS will hang indefinitely in
1821 obd_statfs_rqset() waiting for the statfs data. While for
1822 MDS QOS usage of statfs, it should not stuck in waiting.
1824 Severity : enhancement
1826 Description: remote_acl support
1827 Details : Support ACL-based permission check for remote user.
1828 Support setfacl/getfacl for remote user with the utils
1829 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1830 the system "{s,g}etfacl" utils.
1832 Severity : enhancement
1834 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1836 Severity : enhancement
1838 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1842 Description: make mgs_setparam() handle fsname containing dash
1843 Details : fsname containing a dash does not work with lctl conf_param
1845 Severity : enhancement
1847 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1849 Severity : enhancement
1851 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1855 Frequency : rare, at shutdown
1856 Description: access already free / zero obd_namespace.
1857 Details : if client_disconnect_export was called without force flag set,
1858 and exist connect request in flight, this can produce access to
1859 NULL pointer (or already free pointer) when connect_interpret
1860 store ocd flags in obd_namespace.
1864 Frequency : only at startup
1865 Description: not alloc memory with spinlock held.
1866 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1867 if any spinlock held.
1869 Severity : enhancement
1871 Description: make lustre randomly fail allocating memory
1872 Details : Make lustre randomly failed allocating memory for testing purpose.
1874 Severity : enhancement
1876 Description: lost problems with lov objid file
1877 Details : Fixes some scability and access to not inited memory problems
1878 in work with lov objdid file.
1883 Description: lfs find does not continue on file error
1884 Details : Continue other files processing when a file/dir is absent.
1888 Description: Inconsistent usage of lustre_pack_reply()
1889 Details : Standardize the usage of lustre_pack_reply() such that it
1890 always generate a CERROR on failure.
1893 Frequency : occasional
1895 Description: MDS hang or stay in waiting lock
1896 Details : If client receive lock with CBPENDING flag ldlm need send lock
1897 cancel as separate rpc, to avoid situation when cancel request
1898 can't processed due all i/o threads stay in wait lock.
1902 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1906 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1910 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1914 Description: Update to RHEL4 latest kernel.
1916 Severity : enhancement
1918 Description: Build SLES10 patchless client fails
1919 Details : The configure was broken by run ./configure with
1920 --with-linux-obj=.... argument for patchless client. When the
1921 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1922 can't search header adequately. Use absolute path such as
1923 -I($LINUX)/include instead.
1927 Description: interrupt oig_wait produce painc on resend.
1928 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1929 private set, and this produce situation when rq_ptlrpcd_data not
1930 copyed to new allocated request and triggered LBUG on assert
1931 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1932 ptlrpcd set if request is changed and can be safety changed to use
1937 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1938 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1939 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1940 these macros and replace them with their difinitions instead.
1944 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1945 Details : Update which_patch & target file for SLES10 latest kernel.
1948 Description: Remove client patches from SLES 10 kernel.
1949 Details : This causes SLES 10 clients to behave as patchless clients
1950 even on a Lustre-patched (server) kernel.
1952 Severity : enhancement
1954 Description: self-adjustable client's lru lists
1955 Details : use adaptive algorithm for managing client cached locks lru
1956 lists according to current server load, other client's work
1957 pattern, memory activities, etc. Both, server and client
1958 side namespaces provide number of proc tunables for controlling
1961 Severity : enhancement
1963 Description: light-weight GSS support
1964 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1965 and reduce performance overhead.
1967 Severity : enhancement
1969 Description: Linux keyring support
1970 Details : Support using service of Linux keyring for Lustre GSS internal
1971 context refresh/cache mechanism.
1975 Description: Fix errors in lfs documentation
1976 Details : Fixes man pages
1980 Description: don't use GFP_* in generic Lustre code.
1981 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1985 Description: obdclass is limited by single OBD_ALLOC(idarray)
1986 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1990 Description: warnings with build patchless client with vanila 2.6.19 and up
1991 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1992 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1996 Frequency : Only for SLES
1997 Description: sanity_quota fail test_1
1998 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1999 file, which makes TSTID[2] inunique.
2003 Frequency : Only for SLES9
2004 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2005 Details : Update target/ChangeLog/which_patch .
2009 Frequency : Only for test_10 in sanity-quota.sh
2010 Description: a bug in quota test code
2011 Details : checking a test flag in a wrong place
2015 Frequency : Only for test_13 in sanity-quota.sh
2016 Description: a bug in quota test script
2017 Details : shouldn't check overfull allocated quota
2021 Description: jbd statistics
2022 Details : Port older jbd statistics patch for sles10
2024 Severity : enhancement
2026 Description: Add printing inode info into message about error in writepage.
2031 Description: reply_lock_interpret crash due to race with it and lock cancel.
2032 Details : Do not replay locks that are being cancelled. Do not reference
2033 locks by their address during replay, just by their handle.
2037 Frequency : When flocks are used.
2038 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2039 Details : Flock locks might destroy just granted lock if it could be merged
2040 with another existing flock, this is done in completion handler,
2041 so teach ldlm_cli_enquque_fini that this is a valid case for
2047 Description: Oops in read and write path when failing to allocate lock.
2048 Details : Check if lock allocation failed and return error back.
2052 Description: lstripe command fails for valid OST index
2053 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2054 instead of lov->desc.ld_active_tgt_count.
2058 Description: sanity.sh failed test 103
2059 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2064 Description: df doesn't work properly if diskfs blocksize != 4K
2065 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2069 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2070 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2072 Severity : enhancement
2074 Description: Adaptive timeouts
2075 Details : RPC timeouts adapt to changing server load and network
2076 conditions to reduce resend attempts and improve recovery time.
2080 Description: llapi_file_create() does not allow some changes
2081 Details : add llapi_file_open() that allows specifying the mode and
2082 open flags, and also returns an open file handle.
2086 Description: merge and cleanup kernel patches.
2087 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2091 Description: Add journal checksum support.(Kernel part)
2092 Details : The journal checksum feature adds two new flags i.e
2093 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2094 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2095 indicates that the commit block contains the checksum for
2096 the blocks described by the descriptor blocks. Now commit
2097 record can be sent to disk without waiting for descriptor
2098 blocks to be written to disk. This behavior is controlled
2099 using JBD2_FEATURE_ASYNC_COMMIT flag.
2103 Description: OSS needs mutliple precreate threads
2104 Details : Add ability to start more than one create thread per OSS.
2108 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2109 Details : execute lfs setstripe on client
2113 Description: mds_obd_create error creating tmp object
2114 Details : When the user sets quota on root, llog will be affected and can't
2115 create files and write files.
2118 Frequency : Always on ia64 patchless client, and possibly others.
2120 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2121 Details : This allows the patchless client to be loaded on architectures
2122 without this export.
2126 Description: RedHat Update kernel for RHEL5
2127 Details : Add the kernel config file for RHEL5.
2131 Description: RedHat Update kernel for RHEL5
2132 Details : Modify the kernel config file more closer RHEL5.
2136 Description: Build failure against Centos5 (RHEL5)
2137 Details : Define PAGE_SIZE when it isn't present.
2141 Description: test_7 on sanity-quota.sh can't be used on separate machines
2142 Details : add facet to handle it
2146 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2147 Details : with NFS, the anon dentry's parent was set to itself in
2148 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2149 corresponding dentry other than use rec->ur_name.
2151 Severity : enhancement
2153 Description: lfs setstripe enhancement
2154 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2158 Description: enable data checksumming by default
2159 Details : enable checksum by default, allow --disable-checksum
2160 configure option and "-o nochecksum" mount option. Checksums
2161 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2165 Description: lustre support for RHEL5
2166 Details : Add support for RHEL5.
2170 Description: Client eviction due to failover config
2171 Details : after a connection loss, the lustre client should attempt to
2172 reconnect to the last active server first before trying the
2173 other potential connections.
2177 Description: when mds and osts use different quota unit(32bit and 64bit),
2178 quota will be released repeatly.
2179 Details : void sending multiple quota reqs to mds, which will keep the
2180 status between the reqs.
2184 Description: osts not allocated evenly to files
2185 Details : change the condition to increase offset_idx
2188 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2190 Description: Data corruption for OSTs that are formatted larger than 2TB
2192 Details : When generating the bio request for lustre file writes the
2193 sector number would overflow a temporary variable before being
2194 used for the IO. The data reads correctly from Lustre (which
2195 will overflow in a similar manner) but other file data or
2196 filesystem metadata may be corrupted in some cases.
2200 Description: Tune the kernel for good SCSI performance.
2201 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2202 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2207 Description: rewrite ext2-derived code in obdclass/uuid.c
2208 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2209 from scratch preserving functionality.
2213 Description: "lfs find -obd UUID" prints directories
2214 Details : "lfs find -obd UUID" will return all directory names instead
2215 of just file names. It is incorrect because the directories
2216 do not reside on the OSTs.
2220 Description: use i_size_read and i_size_write in 2.6 port
2221 Details : replace inode->i_size access with i_size_read/write()
2224 Frequency : when removing large files
2226 Description: scheduling issue during removal of large Lustre files
2227 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2228 It causes scheduling issues when removing large files (17TB in the
2232 Frequency : only with liblustre clients on XT3
2234 Description: evictions taking too long
2235 Details : allow llrd to evict clients directly on OSTs
2238 Frequency : only on ppc
2240 Description: /proc/fs/lustre/devices broken on ppc
2241 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2242 We should make sure the type of 'index' is loff_t and then cast
2243 explicitly as needed below. Do not assign an explicitly cast
2247 Frequency : only for rhel5
2249 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2250 Details : Modify the target file & which_kernel.
2252 Severity : enhancement
2254 Description: omit set fsid for export NFS
2255 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2256 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2260 Frequency : Always, for kernels after 2.6.16
2261 Description: Fix warning idr_remove called for id=.. which is not allocated.
2262 Details : Last kernels save old s_dev before kill super and not allow
2263 to restore from callback - restore it before call kill_anon_super.
2267 Description: buffer overruns could theoretically occur
2268 Details : llapi_semantic_traverse() modifies the "path" argument by
2269 appending values to the end of the origin string, and a
2270 overrun may occur. Adding buffer overrun check in liblustreapi.
2274 Description: Fix error on 'ls .' at the top of the Lustre mount.
2275 Details : Don't revalidate dentry if it is a root dentry.
2279 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2280 Details : Modify vm-tunables-rhel4.patch.
2284 Description: Kernel config for 2.6.18-vanilla.
2285 Details : Modify targets/2.6-vanilla.target.in.
2286 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2287 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2288 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2289 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2293 Description: improve handling recoverable errors
2294 Details : if request processig with error which can be recoverable on server
2295 request should be resend, otherwise page released from cache and
2300 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2301 Details : Modify target file & which_patch.
2302 A flaw was found in the IA32 system call emulation provided
2303 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2304 value could be stored in the %RAX register, which could trigger an
2305 out-of-bounds system call table access. An untrusted local user
2306 could exploit this flaw to run code in the kernel
2307 (ie a root privilege escalation). (CVE-2007-4573).
2311 Description: O_DIRECT bypasses client statistics.
2312 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2313 read_ahead_stats were updated. copied the stats section from
2314 osc_send_oap_rpc() into async_internal().
2318 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2322 Description: change order of libsysio includes
2323 Details : '#include sysio.h' should always come before '#include xtio.h'
2327 Description: handle "serious error: objid * already exists" more gracefully
2328 Details : If LAST_ID value on disk is smaller than the objects existing in
2329 the O/0/d* directories, it indicates disk corruption and causes an
2330 LBUG(). If the object is 0-length, then we should use the existing
2331 object. This will help to avoid a full fsck in most cases.
2333 Severity : enhancement
2335 Description: adapt the lustre_config script to support the upgrade case
2336 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2337 server targets to 1.6 in parallel.
2341 Description: To avoid grant space > avaible space when the disk is almost
2342 full. Without this patch you might see the error "grant XXXX >
2343 available" or some LBUG about grant, when the disk is almost
2345 Details : In filter_check_grant, for non_grant cache write, we should
2346 check the left space by if (*left > ungranted + bytes), instead
2347 of (*left > ungranted), because only we are sure the left space
2348 is enough for another "bytes", then the ungrant space should be
2349 increase. In client, we should update cl_avail_grant only there
2350 is OBD_MD_FLGRANT in the reply.
2354 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2355 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2356 described in CVE-2007-4573. This problem could allow untrusted
2357 local users to gain root access.
2360 Frequency : when using O_DIRECT and quotas
2362 Description: Incorrect file ownership on O_DIRECT output files
2363 Details : block usage reported by 'lfs quota' does not take into account
2364 files that have been written with O_DIRECT.
2369 Description: touch file failed when fs is not full
2370 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2371 otherwise we can get ENOSP while fs is not full.
2375 Description: parallel lock callbacks
2376 Details : Instead of sending blocking and completion callbacks as separated
2377 requests, adding them to a set and sending in parallel.
2380 Frequency : only for Cray XT3
2381 Bugzilla : 12829/13455
2382 Description: Changing primary group doesn't change the group lustre assigns to
2384 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2385 overridden with the primary group provided by the group upcall,
2386 whereas the supplied fsgid can be trusted if it is in the list of
2387 supplementary groups returned by the group upcall.
2389 Severity : enhancement
2391 Description: Allow masking D_WARNING, D_ERROR messages from console
2392 Details : Console messages can now be disabled via lnet.printk.
2396 Description: User code with malformed file open parameter crashes client node
2397 Details : Before packing join_file req, all the related reference should be
2398 checked carefully in case some malformed flags cause fake
2399 join_file req on client.
2403 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2404 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2405 mode is equal to the granted mode. However, it should also check
2406 that LDLM_FL_CP_REQD is not set.
2410 Description: Heavy nfs access might result in deadlocks
2411 Details : After ELC code landed, it is now improper to enqueue any mds
2412 locks under och_sem, because enqueue might want to decide to
2413 cancel open locks for same inode we are holding och_sem for.
2417 Description: Client eviction while running blogbench
2418 Details : A lot of unlink operations with concurrent I/O can lead to a
2419 deadlock causing evictions. To address the problem, the number of
2420 oustanding OST_DESTROY requests is now throttled to
2421 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2422 callbacks are processed in priority.
2426 Description: enable ACLs on MDS by default
2427 Details : ACLs must be enabled on MDS by default.
2430 Frequency : PPC/PPC64 only
2432 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2433 Details : fix duplicated definitions between asm-ppc64/types.h and
2434 lustre_types.h on PPC.
2437 Frequency : PPC/PPC64 only
2439 Description: asm-ppc/segment.h does not exist
2440 Details : fix compile issue on PPC.
2444 Description: better handle error messages in extents code
2447 Frequency : RHEL4 only
2449 Description: mkfs is very slow on IA64/RHEL4
2450 Details : A performance regression has been discovered in the MPT Fusion
2451 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2452 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2453 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2455 Severity : enhancement
2457 Description: SNMP support enhancement
2458 Details : Adding total number of sampled request for an MDS node in snmp
2461 Severity : enhancement
2463 Description: Optimize ldlm waiting list processing for PR extent locks
2464 Details : When processing waiting list for read extent lock and meeting read
2465 lock that is same or wider to it that is not contended, skip
2466 processing rest of the list and immediatelly return current
2467 status of conflictness, since we are guaranteed there are no
2468 conflicting locks in the rest of the list.
2472 Description: Time out and refuse to reconnect
2473 Details : When the failover node is the primary node, it is possible
2474 to have two identical connections in imp_conn_list. We must
2475 compare not conn's pointers but NIDs, otherwise we can defeat
2476 connection throttling.
2480 Description: port llog fixes from b1_6 into HEAD
2481 Details : Port llog reference couting and some llog cleanups from b1_6
2482 (bug 10800) into HEAD, for protect from panic and access to already
2483 free llog structures.
2487 Description: Detect stride IO mode in read-ahead
2488 Details : When a client does stride read, read-ahead should detect that and
2489 read-ahead pages according to the detected stride pattern.
2493 Description: data checksumming impacts single node performance
2494 Details : add support for several checksum algorithm. Currently, only CRC32
2495 and Adler-32 are supported. The checksum type can be changed on
2496 the fly via /proc/fs/lustre/osc/*/checksum_type.
2500 Description: use adler32 for page checksums
2501 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2506 Description: build for x2 fails
2507 Details : fix compile issue on Cray systems.
2511 Description: Properly match for duplicate locks
2512 Details : Due to different lock order from skiplists code, we need to
2513 traverse entire list for now
2516 Frequency : only on PPC/SLES10
2518 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2519 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2520 be defined. Add a hack in mkfs_lustre.c to work around this compile
2525 Description: LASSERT on MDS when client holding flock lock dies
2526 Details : ldlm pool logic depends on number of granted locks equal to
2527 number of released locks which is not true for flock locks, so
2528 just exclude such locks from consideration.
2532 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2533 Details : Use fsfilt_sync() for both the whole filesystem sync and
2534 individual file sync to eliminate dangerous inode locking
2535 with I_LOCK that can lead to a deadlock.
2539 Description: performance in 1.6.3
2540 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2541 sure that 1MB requests are not fragmented by the block layer.
2543 Severity : enhancement
2545 Description: organize the server-side client stats on per-nid basis
2546 Details : Change the structure of stats under obdfilter and mds to
2554 The "uuid"s file would list the uuids of _active_ exports.
2555 And the clear entry is to clear all stats and stale nids.
2557 Severity : enhancement
2559 Description: eliminate client locks in face of contention
2560 Details : file contention detection and lockless i/o implementation
2561 for contended files.
2565 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2568 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2570 Description: MDS may lose file striping (and hence file data) in some cases
2571 Details : If there are additional extended attributes stored on the MDS,
2572 in particular ACLs, SELinux, or user attributes (if user_xattr
2573 is specified for the client mount options) then there is a risk
2574 of attribute loss. Additionally, the Lustre file striping
2575 needs to be larger than default (e.g. striped over all OSTs),
2576 and an additional attribute must be stored initially in the
2577 inode and then increase in size enough to be moved to the
2578 external attribute block (e.g. ACL growing in size) for file
2583 Description: skiplist implementation simplification
2584 Details : skiplists are used to group compatible locks on granted list
2585 that was implemented as tracking first and last lock of each
2586 lock group the patch changes that to using doubly linked lists
2590 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2591 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2592 because LASSERT on both the data supplied by a client, and the
2593 data on disk is dangerous and incorrect.
2595 Severity : enhancement
2597 Description: Slow truncate/writes to huge files at high offsets.
2598 Details : Directly associate cached pages to lock that protect those pages,
2599 this allows us to quickly find what pages to write and remove
2600 once lock callback is received.
2604 Description: more ldlm soft lockups
2605 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2606 starve other threads from the resource lock for a long time in
2607 case of long waiting queue, so change the debug level from
2608 D_OTHER to the less frequently used D_INFO.
2610 Severity : enhancement
2612 Description: add -gid, -group, -uid, -user options to lfs find
2616 Description: Hung threads in invalidate_inode_pages2_range
2617 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2618 once one is completed. As a result, some RPCs are stuck in the
2619 queue and are never sent.
2623 Description: filter threads hungs on waiting journal commit
2624 Details : Cleanup filter group llog code, then only filter group llog will
2625 be only created in the MDS/OST syncing process.
2629 Description: Procfs and llog threads access destoryed import sometimes.
2630 Details : Sync the import destoryed process with procfs and llog threads by
2631 the import refcount and semaphore.
2633 Severity : enhancement
2635 Description: openlock cache of b1_6 port to HEAD
2640 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2641 Details : If insertion of an extent fails, then discard the inode
2642 preallocation and free data blocks else it can lead to duplicate
2647 Description: don't always update ctime in ext3_xattr_set_handle()
2648 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2649 In some cases the ctime should not be updated, for example for
2650 2.0->1.8 compatibility it is necessary to delete an xattr and it
2651 should not update the ctime.
2655 Bugzilla : 15713/16362
2656 Description: Assertion in iopen_connect_dentry in 1.6.3
2657 Details : looking up an inode via iopen with the wrong generation number can
2658 populate the dcache with a disconneced dentry while the inode
2659 number is in the process of being reallocated. This causes an
2660 assertion failure in iopen since the inode's dentry list contains
2661 both a connected and disconnected dentry.
2665 Description: assertion failure in ldlm_handle2lock()
2666 Details : fix a race between class_handle_unhash() and class_handle2object()
2667 introduced in lustre 1.6.5 by bug 13622.
2672 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2673 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2678 Description: oops in sd_iostats_seq_show()
2679 Details : unloading/reloading the scsi low level driver triggers a kernel
2680 bug when trying to access the sd iostat file.
2685 Description: Kernel panics during QLogic driver reload
2686 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2687 patch, causing memory corruption.
2692 Description: journal_dev option does not work in b1_6
2693 Details : pass mount option during pre-mount.
2695 Severity : enhancement
2697 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2698 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2699 extent information of a file. It can be used to map logical blocks
2700 in a file to physical blocks in the block device.
2704 Description: LDLM soft lockups - improvement
2705 Details : It is be possible to send the lock handle along with each read
2706 or write request because the client is already doing a lock match
2707 itself so there isn't any reason the OST should have to re-do that
2711 Frequency : only on Cray X2
2713 Description: X2 build failures
2714 Details : fix build failures on Cray X2.
2718 Description: xid & resent requests
2719 Details : Initialize RPC XID from clock at startup (randomly if clock is
2722 Severity : enhancement
2724 Description: Add lustre_start utility to start or stop multiple Lustre servers
2729 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2730 Details : In case of memory pressure, list_del() can be called twice on
2731 req->rq_history_list, causing a kernel oops.
2735 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2736 Details : fix stack overflow in the distributed lock manager by defering
2737 export eviction after a failed AST to the elt thread instead of
2738 handling it in the dlm interpret routine.
2742 Description: Convert some comments to new format.
2743 Details : Update documenting comments to match doxygen conventions.
2747 Description: Grammar fixes.
2748 Details : A couple of trivial spelling fixes.
2752 Description: OSD_COUNTERS-mandatory
2753 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2754 to simplify the code.
2758 Description: simplify cmm_device freeing logic.
2759 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2763 Description: Add lockdep support to dt_object_operations locking interface.
2764 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2765 parameter indicating lock ordering. Update mdd code to use new
2770 Description: Introduce failloc constants for lockless IO tests.
2771 Details : Add two new failloc constants to test lockless IO. Only one of
2772 them in implemented---another is checked in yet to be landed
2777 Description: Add lockdep support for inode mutex.
2778 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2779 in the situations where Lustre has to lock more than one inode
2784 Description: Add optional invariants checking support.
2785 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2786 configure switch. This macro is to be used for consistency and
2787 sanity checks that are too expensive to be left in `production'
2792 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2793 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2798 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2799 Details : Use spin_lock_nested() in (the only) situation where more than
2800 one ldlm_lock is locked simultaneously. Also, fix possible
2801 dead-lock in ldlm_lock_change_resource() by enforcing particular
2806 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2807 Details : Instead of specifying each ldlm_lock call-back through separate
2808 parameter, wrap them into struct ldlm_callback_suite.
2812 Description: Kill join_lru obd method and its callers.
2813 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2814 in memory---a requirement implemented through ->o_join_lru() obd
2815 method in HEAD. Remove this method and its users.
2819 Description: Add asynchronous ldlm ENQUEUE completion handler.
2820 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2821 case that stock ldlm_completion_ast() cannot handle as it waits
2822 until lock is granted. Introduce new ldlm_completion_ast_async()
2823 for this. Also comment ldlm_completion_ast().
2827 Description: ldlm_error <-> errno conversion.
2828 Details : Add functions to map (rather arbitrary) between LDLM error codes
2829 and standard errno values. CLIO needs this to prevent LDLM specific
2830 constants from escaping ldlm and osc.
2834 Description: Kill unused ldlm_handle2lock_ns() function.
2835 Details : Kill unused ldlm_handle2lock_ns() function.
2839 Description: Add lu_ref support to ldlm_lock
2840 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2841 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2842 and ldlm_resource. LDLM interface has to be changed, because code that
2843 releases a reference on a lock, has to "know" what reference this is.
2844 In the most frequent case
2846 lock = ldlm_handle2lock(handle);
2848 LDLM_LOCK_PUT(lock);
2850 no changes are required. When any other reference (received _not_ from
2851 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2852 instead of LDLM_LOCK_PUT().
2854 Arguably, changes are pervasive, and interface requires some discipline
2855 for proper use. On the other hand, it was very instrumental in finding
2856 a few leaked lock references.
2860 Description: Add ldlm_lock_addref_try().
2861 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2862 attempts to addref a lock that might be being canceled
2867 Description: Add ldlm_weigh_callback().
2868 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2869 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2870 hard-coded `number of pages' logic.
2874 Description: Add lockdep annotations to llog code.
2875 Details : Use appropriately tagged _nested() locking calls in the places
2876 where llog takes more than one ->lgh_lock lock.
2880 Description: Add loi_kms_set().
2881 Details : Wrap kms updates into a helper function.
2885 Description: Constify instances of struct lsm_operations.
2886 Details : Constify instances of struct lsm_operations.
2890 Description: lu_conf support.
2891 Details : On a server, a file system object is uniquely identified
2892 by a fid, which is sufficient to locate and load all object
2893 state (inode). On a client, on the other hand, more data are
2894 necessary instantiate an object. Change lu_object_find() and
2895 friends to take additional `lu_conf' argument describing object.
2896 Typically this includes layout information.
2900 Description: lu_context fixes.
2901 Details : Introduce new lu_context functions that are needed on the client
2902 side, where some system threads (ptlrpcd) are shared by multiple
2903 modules, and so cannot be stopped during module shutdown.
2907 Description: Add start and stop methods to lu_device_type_operations.
2908 Details : Introduce two new methods in lu_device_type_operations, that are
2909 invoked when first instance of a given type is created and last one
2910 is destroyed respectively. This is need by CLIO.
2914 Description: Add lu_ref support to struct lu_device.
2915 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2916 track leaked references.
2920 Description: Introduce lu_kmem_descr.
2921 Details : lu_kmem_descr and its companion interface allow to create
2922 and destroy a number of kmem caches at once.
2926 Description: Fix lu_object finalization race.
2927 Details : Fix a race between lu_object_find() finding an object and its
2928 concurrent finalization. This race is (most likely) not possible
2929 on the server, but might happen on the client.
2933 Description: Introduce lu_ref interface.
2934 Details : lu_ref is a debugging module allowing to track references to
2935 a given object. It is quite cpu expensive, and has to be
2936 explicitly enabled with --enable-lu_ref. See usage description
2941 Description: Factor lu_site procfs stats into a separate function.
2942 Details : Separate lu_site stats printing code into a separate function
2943 that can be reused on a client.
2947 Description: Constify instances of struct {lu,dt,md}_device_operations.
2948 Details : Constify instances of struct {lu,dt,md}_device_operations.
2952 Description: Introduce struct md_site and move meta-data specific parts of
2953 struct lu_site here.
2954 Details : Move md-specific fields out of struct lu_site into special struct
2955 md_site, so that lu_site can be used on a client.
2959 Description: Kill mdd_lov_destroy().
2960 Details : Remove unused mdd code.
2964 Description: Add st_block checking to multistat.c.
2965 Details : Add st_block checking to multistat.c.
2969 Description: Add lu_ref support to struct obd_device.
2970 Details : Add lu_ref tracking to obd_device.
2974 Description: Kill obd_set_fail_loc().
2975 Details : Remove unused code.
2979 Description: Add special type for ptlrpc_request interpret functions.
2980 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2981 there. Actual usage will be in CLIO.
2985 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2986 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2987 doesn't work with lockdep.
2991 Description: Add rwv.c test program.
2992 Details : New testing program exercising readv(2) and writev(2) (Qian).
2996 Description: Add sendfile.c test program.
2997 Details : New testing program exercising sendfile(2) (Jay).
3001 Description: Ratelimit a message that can be very frequent.
3002 Details : Ratelimit a memory allocation failure message that can
3007 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3008 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3012 Description: Liblustre build fixes.
3013 Details : Liblustre build fixes.
3017 Description: libcfs: add cfs_{need,cond}_resched() interface.
3018 Details : libcfs: add cfs_{need,cond}_resched() definition and
3019 implementations for Linux, NT, and liblustre.
3021 Severity : enhancement
3023 Description: More exported tunables for mballoc
3024 Details : Add support for tunable preallocation window and new tunables for
3025 large/small requests
3029 Description: Detect corruption of block bitmap and checking for preallocations
3030 Details : Checks validity of on-disk block bitmap. Also it does better
3031 checking of number of applied preallocations. When corruption is
3032 found, it turns filesystem readonly to prevent further corruptions.
3036 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3037 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3038 window to 0. If it does hit cache miss, and read-pattern is still
3039 stride-io mode, does not reset the stride window, but also does not
3040 increase the stride window length in this case.
3044 Frequency : only for big-endian servers
3045 Description: Check if system is big-endian while mounting fs with extents feature
3046 Details : Mounting a filesystem with extents feature will fail on big-endian
3047 systems since ext3-based ldiskfs is not supported on big-endian
3048 systems. This can be over-riden with "bigendian_extents" mount option.
3050 Severity : enhancement
3052 Description: The root squash functionality
3053 Details : A security feature, which is to prevent users from being able
3054 to mount lustre on their desktop, run as root, and delete
3055 all of the files in the filesystem. The goal is accomplished by
3056 remapping user id (UID) and group id (GID) of the root user to
3057 a UID and GID specified by the system administartor via Lustre
3058 configuration management server (MGS). The functionality also
3059 allows to specify sets of clients for which the remapping does
3064 Description: Excessive recovery window
3065 Details : With AT enabled, the recovery window can be excessively long (6000+
3066 seconds). To address this problem, we no longer use
3067 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3068 timeout no longer depends on the service time, it is set to
3069 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3070 time via pb_service_time.
3074 Description: Watchdog triggered on MDS failover
3075 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3076 the OSTs know that the MDS "UUID" can be reused for the same export
3077 from a different NID, so we do not need to wait for the export to be
3081 Frequency : rare, only if using MMP with Linux RAID
3083 Description: MMP doesn't work with Linux RAID
3084 Details : While using HA for Lustre servers with Linux RAID, it is possible
3085 that MMP will not detect multiple mounts. To make this work we
3086 need to unplug the device queue in RAID when the MMP block is being
3087 written. Also while reading the MMP block, we should read it from
3088 disk and not the cached one.
3090 Severity : enhancement
3092 Description: open file using fid
3093 Details : A file can be opened using just its fid, like
3094 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3097 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3100 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3101 Details : add proper locking for creation of jbd_4k slab cache
3105 Description: MMP check in ext3_remount() fails without displaying any error
3106 Details : When multiple mount protection fails during remount, proper error
3109 Severity : enhancement
3111 Description: Allow stripe size to be up to 4G-64k
3112 Details : Fix math logic to allow large stripe sizes.
3116 Description: add check for >8TB ldiskfs filesystems
3117 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3118 Don't allow >8TB ldiskfs filesystems to be mounted without
3119 force_over_8tb mount option
3124 Descriptoin: OSS complains "too many blocks" and crashes
3125 Details : In filter_commitrw_write():
3126 1. lquota_chkquota();
3127 2. obdo_from_inode();
3128 3. lquota_pending_commit();
3129 "2." has changed uid and gid in this case, then caused
3130 lquota_chkquota() and lquota_pending_commit() to use
3131 different uid and gid. That is the root of the bug.
3135 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3136 Details : Detect and print out the 'stripe count = -1' case properly.
3140 Description: Out or order replies might be lost on replay
3141 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3142 request with tid smaller than one currently being added, add it
3143 to the start, not end of the list.
3145 --------------------------------------------------------------------------------
3147 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3149 * Support for kernels:
3150 2.6.5-7.283 (SLES 9),
3151 2.6.9-55.EL (RHEL 4),
3152 2.6.16.46-0.14 (SLES 10),
3153 2.6.18.8 vanilla (kernel.org)
3154 * Client support for unpatched kernels:
3155 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3156 2.6.16 - 2.6.22 vanilla (kernel.org)
3157 * Due to recently discovered recovery problems, we do not recommend
3158 using patchless RHEL 4 clients with this or any earlier release.
3159 * Recommended e2fsprogs version: 1.39.cfs8
3160 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3161 * Starting with this release, the ldiskfs backing filesystem required
3162 by Lustre is now in its own package, lustre-ldiskfs. This package
3163 should be installed. It is versioned separately from Lustre and
3164 may be released separately in future.
3168 Description: block reactivating mgc import until all deactivates complete
3169 Details : Fix race when failing back MDT/MGS to itself (testing)
3172 Frequency : at statup only
3174 Description: mds_lov_synchronize race leads to various problems
3175 Details : simultaneous MDT->OST connections at startup can cause the
3176 sync to abort, leaving the OSC in a bad state.
3178 Severity : enhancement
3180 Description: add optional extra BUILD_VERSION info
3181 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3182 to override the lustre version.
3185 Frequency : 2.6.18 servers only
3187 Description: ll_kern_mount() doesn't release the module reference
3188 Details : The ldiskfs module reference count never drops down to 0
3189 because ll_kern_mount() doesn't release the module reference.
3194 Description: server LBUG when using old ost_num_threads parameter
3195 Details : Accept the old ost_num_threads parameter but warn that it
3196 is deprecated, and fix an off-by-one error that caused an LBUG.
3201 Description: Transient SCSI error results in persistent IO issue
3202 Details : iobuf->dr_error is not reinitialized to 0 between two
3206 Frequency : sometimes when underlying device returns I/O errors
3208 Description: OSTs not going read-only during write failures
3209 Details : OSTs are not remounted read-only when the journal commit threads
3210 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3211 instead of the ext3 wrappers.
3215 Description: poor connect scaling with increasing client count
3216 Details : Don't run filter_grant_sanity_check for more than 100 exports
3217 to improve scaling for large numbers of clients.
3220 Frequency : SLES10 only
3222 Description: sanity-quota.sh quotacheck failed: rc = -22
3223 Details : Quotas cannot be enabled on SLES10.
3226 Frequency : liblustre clients only
3228 Description: getdirentries does not give error when run on compute nodes
3229 Details : getdirentries does not fail when the size specified as an argument
3230 is too small to contain at least one entry
3232 Severity : enhancement
3234 Description: Add LNET router traceability for debug purposes
3235 Details : If a checksum failure occurs with a router as part of the
3236 IO path, the NID of the last router that forwarded the bulk data
3237 is printed so it can be identified.
3242 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3243 Details : Due to a race condition, liblustre clients were occasionally
3244 evicted incorrectly.
3246 Severity : enhancement
3248 Description: lfs setstripe use optional parameters instead of postional
3251 Severity : enhancement
3253 Description: Nanosecond timestamp support for ldiskfs
3254 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3255 resolution timestamps. There is not yet support for this at
3256 the Lustre filesystem level.
3259 Frequency : during server recovery
3261 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3262 Details : request with rq_no_resend flag not awake l_wait_event if they get
3266 Frequency : nfs export on patchless client
3268 Description: connectathon hang when test nfs export over patchless client
3269 Details : Disconnected dentry cannot be found with lookup, so we do not need
3270 to unhash it or make it invalid
3273 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3274 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3275 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3280 Description: open req refcounting wrong on reconnect
3281 Details : If reconnect happened between getting open reply from server and
3282 call to mdc_set_replay_data in ll_file_open, we will schedule
3283 replay for unreferenced request that we are about to free.
3284 Subsequent close will crash in variety of ways.
3285 Check that request is still eligible for replay in
3286 mdc_set_replay_data().
3291 Description: disable writes to filesystem when reading health_check file
3292 Details : the default for reading the health_check proc file has changed
3293 to NOT do a journal transaction and write to disk, because this
3294 can cause reads of the /proc file to hang and block HA state
3295 checking on a healthy but otherwise heavily loaded system. It
3296 is possible to return to the previous behaviour during configure
3297 with --enable-health-write.
3299 Severity : enhancement
3301 Description: 64-bit inode version
3302 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3303 to the inode. This will be required for version-based recovery.
3308 Description: MDS fails to start if a duplicate client export is detected
3309 Details : in some rare cases it was possible for a client to connect to
3310 an MDS multiple times. Upon recovery the MDS would detect this
3311 and fail during startup. Handle this more gracefully.
3313 Severity : enhancement
3315 Description: Add -o localflock option to simulate old noflock
3317 Details : This will achieve local-only flock/fcntl locks
3323 Description: log_commit_thread vs filter_destroy race leads to crash
3324 Details : Take import reference before releasing llog record semaphore
3329 Description: Wrong request locking in request set processing
3330 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3331 imp_delayed_list, in this place should be used imp_lock.
3334 Frequency : when reconnection
3336 Description: Grant Leak when osc reconnect to OST
3337 Details : When osc reconnect ost, OST(filter) should check whether it
3338 should grant more space to client by comparing fed_grant and
3339 cl_avail_grant, and return the granted space to client instead
3340 of "new granted" space, because client will call osc_init_grant
3341 to update the client grant space info.
3344 Frequency : when client reconnect to OST
3346 Description: Grant Leak when osc do resend and replay bulk write
3347 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3348 bulk write request, because the grant info while be sync between
3349 OSC and OST when reconnect, and we should ignore the grant info
3350 these of resend/replay write req.
3355 Description: Grant space more than available left space sometimes.
3356 Details : When then OST is about to be full, if two bulk writing from
3357 different clients came to OST. Accord the avaliable space of the
3358 OST, the first req should be permitted, and the second one
3359 should be denied by ENOSPC. But if the seconde arrived before
3360 the first one is committed. The OST might wrongly permit second
3361 writing, which will cause grant space > available space.
3364 Frequency : when client is evicted
3366 Description: Grant might be wrongly erased when osc is evicted by OST
3367 Details : when the import is evicted by server, it will fork another
3368 thread ptlrpc_invalidate_import_thread to invalidate the
3369 import, where the grant will be set to 0. While the original
3370 thread will update the grant it got when connecting. So if
3371 the former happened latter, the grant will be wrongly errased
3372 because of this race.
3377 Description: Checking Stale with correct fid
3378 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3379 to check whether it is stale, because sometimes, we want the
3380 enqueue happened anyway, and op_data.fid2 will not be initialized.
3382 Severity : enhancement
3384 Description: update patchless client
3385 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3388 Frequency : only with 2.4 kernel
3390 Description: random memory corruption
3391 Details : size of struct ll_inode_info is to big for union inode.u and this
3392 can be cause of random memory corruption.
3397 Description: Memory leak in recovery
3398 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3399 It should also check obd_fail before fsfilt_start, otherwise if
3400 fsfilt_start return -EROFS,(failover mds during mds recovery).
3401 then the req will return with repmsg->transno = 0 and rc = EROFS.
3402 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3403 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3404 be freed no matter whether fsfilt_commit success or not.
3407 Frequency : only with huge count clients
3409 Description: Prevents from taking the superblock lock in llap_from_page for
3411 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3412 for llap_from_page call in ll_removepage prevents from taking the
3413 superblock lock for a soon died page.
3418 Description: Not check open intent error before release open handle
3419 Details : in some rare cases, the open intent error is not checked before
3420 release open handle, which may cause
3421 ASSERTION(open_req->rq_transno != 0), because it tries to release
3422 the failed open handle.
3427 Description: Set cat log bitmap only after create log success.
3428 Details : in some rare cases, the cat log bitmap is set too early. and it
3429 should be set only after create log success.
3433 Description: Accessing a block bevice can re-enable I/O when Lustre is
3434 tearing down a device.
3435 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3440 Description: service threads may hog cpus when there are a lot of requests
3441 Details : Insert cond_resched to give other threads a chance to use some CPU
3446 Description: the cat log was not initialized in recovery
3447 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3448 unlink log on mds will not be initialized until mds post
3449 recovery. And also in mds post recovery, the unlink log will
3450 initialization will be done asynchronausly, so there will be race
3451 between add unlink log and unlink log initialization.
3455 Description: brw_stats were being printed incorrectly
3456 Details : brw_stats were being printed as log2 but all of them were not
3457 recorded as log2. Also remove some code duplication arising from
3458 filter_tally_{read,write}.
3462 Frequency : rare, only in recovery.
3463 Description: ASSERTION(req->rq_type != LI_POISON) failed
3464 Details : imp_lock should be held while iterating over imp_sending_list for
3465 prevent destroy request after get timeout in ptlrpc_queue_wait.
3469 Description: replay-single.sh test 52 fails
3470 Details : A lock's skiplist need to be cleanup when it being unlinked
3471 from its resource list.
3475 Description: Short directio read returns full requested size rather than
3477 Details : Direct I/O operations should return actual amount of bytes
3478 transferred rather than requested size.
3480 Severity : enhancement
3482 Description: metadata RPC reduction (e.g. for rm performance)
3483 Details : decrease the amount of synchronous RPC between clients and servers
3484 by canceling conflicing lock before the operation on the client
3485 and packing thier handles into the main operation RPC to server.
3487 Severity : enhancement
3489 Description: Async OSC create to avoid the blocking unnecessarily.
3490 Details : If an OST has no remaining object, system will block on the
3491 creation when it needs to create a new object on this OST. Now,
3492 ways use pre-created objects when available, instead of blocking on
3493 an empty osc while others are not empty. If we must block, we
3494 block for the shortest possible period of time.
3498 Description: improve handling recoverable errors
3499 Details : if request processig with error which can be recoverable on server
3500 request should be resend, otherwise page released from cache and
3503 Severity : enhancement
3505 Description: refine locking for avoid write wrong info into lov_objid file
3506 Details : fix possible races with add new target and write/update data in
3509 --------------------------------------------------------------------------------
3511 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3516 Frequency : on some architectures
3518 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3519 Details : Uninitialized flags sometimes cause configuration commands to
3523 Frequency : patchless clients only
3525 Description: missing __iget() symbol export
3526 Details : The __iget() symbol export is missing. To avoid the need for
3527 this on patchless clients the deathrow inode reaper is turned
3528 off, and we depend on the VM to clean up old inodes. This
3529 dependency was during via the fix for bug 12181.
3531 --------------------------------------------------------------------------------
3533 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3535 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3536 INTEROPERATE with older versions automatically. In many cases a
3537 special upgrade step is needed. Please read the
3538 user documentation before upgrading any part of a live system.
3539 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3540 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3541 * WARNING: Lustre configuration and startup changes are required with
3542 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3546 Severity : enhancement
3548 Description: MountConf
3549 Details : Lustre configuration is now managed via mkfs and mount
3550 commands instead of lmc and lconf. New obd types (MGS, MGC)
3551 are added for dynamic configuration management. See
3552 https://mail.clusterfs.com/wikis/lustre/MountConf for
3555 Severity : enhancement
3557 Description: dynamic OST addition
3558 Details : OSTs can now be added to a live filesystem
3560 Severity : enhancement
3562 Description: startup order invariance
3563 Details : MDTs and OSTs can be started in any order. Clients only
3564 require the MDT to complete startup.
3566 Severity : enhancement
3568 Description: parallel, asynchronous orphan cleanup
3569 Details : orphan cleanup is now performed in separate threads for each
3570 OST, allowing parallel non-blocking operation.
3572 Severity : enhancement
3574 Description: optimized stripe assignment
3575 Details : stripe assignments are now made based on ost space available,
3576 ost previous usage, and OSS previous usage, in order to try
3577 to optimize storage space and networking resources.
3579 Severity : enhancement
3581 Description: Permanently set tunables
3582 Details : All writable /proc/fs/lustre tunables can now be permanently
3583 set on a per-server basis, at mkfs time or on a live system.
3585 Severity : enhancement
3587 Description: Lustre message v2
3588 Details : Add lustre message format v2.
3590 Severity : enhancement
3592 Description: client OST exclusion list
3593 Details : Clients can be started with a list of OSTs that should be
3594 declared "inactive" for known non-responsive OSTs.
3598 Description: SPEC SFS validation failure on NFS v2 over lustre.
3599 Details : Changes the blocksize for regular files to be 2x RPC size,
3600 and not depend on stripe size.
3602 Severity : enhancement
3604 Description: Multiple MD RPCs in flight.
3605 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3606 To avoid overly-overloading MDS, introduce a limit on number of
3607 MDS RPCs in flight for a single client and add /proc controls
3608 to adjust this limit.
3610 Severity : enhancement
3612 Description: client read/write statistics
3613 Details : Add client read/write call usage stats for performance
3614 analysis of user processes.
3615 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3616 file access. extents_stats shows chunk size distribution.
3617 extents_stats_per_process show chunk size distribution per
3620 Severity : enhancement
3622 Description: mds statistics
3623 Details : Add detailed mds operations statistics in
3624 /proc/fs/lustre/mds/*/stats.
3628 Description: Failure of copying files with lustre special EAs.
3629 Details : Client side always return success for setxattr call for lustre
3630 special xattr (currently only "trusted.lov").
3635 Description: Refcount LNET uuids
3636 Details : The global LNET uuid list grew linearly with every startup;
3637 refcount repeated list entries instead of always adding to
3640 Severity : enhancement
3642 Description: Dynamic service threads
3643 Details : Within a small range, start extra service threads
3644 automatically when the request queue builds up.
3647 Frequency : mixed-endian client/server environments
3649 Description: mixed-endian crashes
3650 Details : The new msg_v2 system had some failures in mixed-endian
3653 Severity : enhancement
3655 Description: Easy OST removal
3656 Details : OSTs can be permanently deactivated with e.g. 'lctl
3657 conf_param lustre-OST0001.osc.active=0'
3659 Severity : enhancement
3661 Description: MGS proc entries
3662 Details : Added basic proc entries for the MGS showing what filesystems
3665 Severity : enhancement
3667 Description: provide MGS failover
3668 Details : Added config lock reacquisition after MGS server failover.
3670 Severity : enhancement
3672 Description: add Linux 2.4 support
3673 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3677 Description: a large application tries to do I/O to the same resource and dies
3678 in the middle of it.
3679 Details : Check the req->rq_arrival time after the call to
3680 ost_brw_lock_get(), but before we do anything about
3681 processing it & sending the BULK transfer request. This
3682 should help move old stale pending locks off the queue as
3683 quickly as obd_timeout.
3686 Frequency : when an incorrect nid is specified during startup
3688 Description: ptlrpc connect to non-existant node causes kernel crash
3689 Details : LNET can't be re-entered from an event callback, which
3690 happened when we expire a message after the export has been
3691 cleaned up. Instead, hand the zombie cleanup off to another
3694 Severity : enhancement
3696 Description: plain/inodebits lock performance improvement
3697 Details : Grouping plain/inodebits in granted list by their request modes
3698 and bits policy, thus improving the performance of search through
3702 Frequency : only if OST filesystem is corrupted
3704 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3705 Details : for a short time RPCs with bulk IO are in the replay list,
3706 but replay of bulk IOs is unimplemented. If the OST filesystem
3707 is corrupted due to disk cache incoherency and then replay is
3708 started it is possible to trip an assertion. Avoid putting
3709 committed RPCs into the replay list at all to avoid this issue.
3712 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3714 Description: System hang on startup
3715 Details : This bug allowed the liblustre (e.g. catamount) client to
3716 return to the app before handling all startup RPCs. This
3717 could leave the node unresponsive to lustre network traffic
3718 and manifested as a server ptllnd timeout.
3720 Severity : enhancement
3722 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3723 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3724 boolean to control whether to print peer debug info when a
3725 client's RPC times out.
3728 Frequency : only for kernels with patches from Lustre below 1.4.3
3730 Description: Remove old rdonly API
3731 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3734 Frequency : only for devices with external journals
3736 Description: Set external device read-only also
3737 Details : During a commanded failover stop, we set the disk device
3738 read-only while the server shuts down. We now also set any
3739 external journal device read-only at the same time.
3742 Frequency : when upgrading from 1.4 while trying to change parameters
3744 Description: The wrong (new) MDC name was used when setting parameters for
3745 upgraded MDT's. Also allows changing of OSC (and MDC)
3746 parameters if --writeconf is specified at tunefs upgrade time.
3749 Frequency : when setting specific ost indicies
3751 Description: QOS code breaks on skipped indicies
3752 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3753 created with --index need not be sequential.
3755 Severity : enhancement
3757 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3758 Details : The uninit_groups feature works in conjunction with the kernel
3759 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3760 up the pass1 processing of e2fsck. This is a read-only feature
3761 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3762 mount filesystems that have had this feature enabled.
3764 Severity : enhancement
3766 Description: Improve multi-block allocation algorithm to avoid fragmentation
3767 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3768 allocation locality and avoid filesystem fragmentation.
3770 ------------------------------------------------------------------------------
3772 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3774 * Support for kernels:
3775 2.4.21-47.0.1.EL (RHEL 3)
3776 2.6.5-7.283 (SLES 9)
3777 2.6.9-42.0.10.EL (RHEL 4)
3778 2.6.12.6 vanilla (kernel.org)
3779 2.6.16.27-0.9 (SLES 10)
3780 * Recommended e2fsprogs version: 1.39.cfs5
3782 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3786 Frequency : occasional, depends on client load and configuration
3787 Bugzilla : 12181, 12203
3788 Description: data loss for recently-modified files
3790 Details : In some cases it is possible that recently written or created
3791 files may not be written to disk in a timely manner (this should
3792 normally be within 30s unless client IO load is very high).
3793 The problem appears as zero-length files or files that are a
3794 multiple of 1MB in size after a client crash or client eviction
3795 that are missing data at the end of the file.
3797 This problem is more likely to be hit on clients where files are
3798 repeatedly created and unlinked in the same directory, clients
3799 have a large amount of RAM, have many CPUs, the filesystem has
3800 many OSTs, the clients are rebooted frequently, and/or the files
3801 are not accessed by other nodes after being written.
3803 The presence of the problem can be detected by looking at
3804 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3805 smaller than the second (nr_unused) then dirty files will not
3806 be flushed automatically to disk. "sync; sleep 10" should be
3807 run several times on the node before unmounting it to update
3808 Lustre (this is also safe to run on nodes without this problem).
3810 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3811 that can cause this same problem, so customers using that kernel
3812 also need to update the kernel in addition to Lustre. In order
3813 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3815 It is normal that files written just before a client crash (less
3816 than 30s) may not yet have been flushed to disk, even for local
3820 Frequency : frequent on thin XT3 nodes
3822 Description: UUID collision on thin XT3 Linux nodes
3823 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3824 randomly, since we relied on an insufficiently-seeded PRNG.
3829 Description: OSS hangs after "All ost request buffers busy"
3830 Details : A deadlock between quota and journal operations caused OSS
3831 hangs after printing "All ost request buffers busy."
3834 Frequency : always on liblustre builds
3836 Description: Cleanup compiler warnings on liblustre
3839 Frequency : always on liblustre builds on XT3
3841 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3845 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3846 > 32000 subdirectories
3847 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3848 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3849 subdirectory count crosses 32000. This will aid e2fsck to
3850 correctly handle more than 32000 subdirectories.
3853 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3855 Description: System hang on startup
3856 Details : This bug allowed the liblustre (e.g. catamount) client to
3857 return to the app before handling all startup RPCs. This
3858 could leave the node unresponsive to lustre network traffic
3859 and manifested as a server ptllnd timeout.
3861 Severity : enhancement
3863 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3864 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3865 boolean to control whether to print peer debug info when a
3866 client's RPC times out.
3871 Description: make O_SYNC working on 2.6 kernels
3872 Details : 2.6 kernels use different method for mark pages for write,
3873 so need add a code to lustre for O_SYNC work.
3878 Description: Failure to close file and release space on NFS
3879 Details : Put inode details into lock acquired in ll_intent_file_open.
3880 Use mdc_intent_lock in ll_intent_open to properly
3881 detect all kind of errors unhandled by mdc_enqueue.
3886 Description: proc file read during shutdown sometimes raced obd removal,
3888 Details : Add lock to prevent obd access after proc file removal.
3891 Frequency : Only for files larger than 4GB on 32-bit clients.
3893 Description: improperly doing page alignment of locks
3894 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3895 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3900 Description: RPCs being resent when they shouldn't be
3901 Details : Some RPCs that should not be resent are being resent. This
3902 can cause inconsistencies in the RPC state machine. Do not
3903 resend such requests.
3906 Frequency : rare, only with NFS export
3908 Description: Crash on NFS re-export node
3909 Details : under very unusual load conditions an assertion is hit in
3910 ll_intent_file_open()
3913 Frequency : only if OST filesystem is corrupted
3915 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3916 Details : for a short time RPCs with bulk IO are in the replay list,
3917 but replay of bulk IOs is unimplemented. If the OST filesystem
3918 is corrupted due to disk cache incoherency and then replay is
3919 started it is possible to trip an assertion. Avoid putting
3920 committed RPCs into the replay list at all to avoid this issue.
3925 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3926 Details : Large single O_DIRECT read and write calls can fail to allocate
3927 a sufficiently large buffer to process the request. In case of
3928 allocation failure the allocation is retried with a smaller
3929 buffer and broken into smaller requests.
3931 Severity : enhancement
3933 Description: Add -o localflock option to simulate old noflock behaviour.
3934 Details : This will achieve local-only flock/fcntl locks coherentness.
3939 Description: versioning check is incomplete
3940 Details : Checking the version difference of client vs. server, report
3941 error if the gap is too big.
3946 Description: add support PG_writeback bit
3947 Details : add support for PG_writeback bit for Lustre, for more carefull
3948 work with page cache in 2.6 kernel. This also fix some deadlocks
3949 and remove hack for work O_SYNC with 2.6 kernel.
3951 Severity : enhancement
3953 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3954 Details : The uninit_groups feature works in conjunction with the kernel
3955 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3956 up the pass1 processing of e2fsck. This is a read-only feature
3957 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3958 mount filesystems that have had this feature enabled.
3960 Severity : enhancement
3962 Description: Improve multi-block allocation algorithm to avoid fragmentation
3963 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3964 allocation locality and avoid filesystem fragmentation.
3966 ------------------------------------------------------------------------------
3968 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3970 * Support for kernels:
3971 2.6.9-42.0.3.EL (RHEL 4)
3972 2.6.5-7.276 (SLES 9)
3973 2.4.21-47.0.1.EL (RHEL 3)
3974 2.6.12.6 vanilla (kernel.org)
3975 2.6.16.21-0.8 (SLES10)
3976 * Recommended e2fsprogs version: 1.39.cfs2-0
3978 * The backwards-compatible /proc/sys/portals symlink has been removed
3979 in this release. Before upgrading, please ensure that you change
3980 any configuration scripts or /etc/sysctl.conf files that access
3981 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3982 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3983 in advance of the upgrade on any system running Lustre 1.4.6 or
3984 newer, since /proc/sys/lnet was added in that version.
3985 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3989 Frequency : only when quota is used
3991 Description: avoid scanning export list for quota master
3992 Details : Change the algorithms to avoid scanning export list in order
3993 to improve the efficiency.
3996 Frequency : MDS failover only, very rarely
3998 Description: "went back in time" messages on mds failover
3999 Details : The greatest transno may be lost when the current operation
4000 finishes with an error (transno==0) and the client's last_rcvd
4001 record is over-written. Save the greatest transno in the
4002 mds_last_transno for this case.
4005 Frequency : always for specific kernels and striping counts
4007 Description: client may get "Matching packet too big" without ACL support
4008 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4009 when trying to access files in certain configurations. The
4010 clients should in fact be denied when mounting because they do
4011 not understand ACLs.
4014 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4016 Description: many clients connecting with IO in progress causes connect timeouts
4017 Details : Avoid synchronous journal commits to avoid delays caused by many
4018 clients connecting/disconnecting when bulk IO is in progress.
4019 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4020 OST_IO_PORTAL to avoid delays behind potentially many pending
4024 Frequency : occasionally with multiple writers to a single file
4026 Description: shared writes to file may result in wrong size reported by stat()
4027 Details : Allow growing of kms when extent lock is cancelled
4030 Frequency : always with random mmap IO to multi-striped file
4032 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4033 Details : Only if the hole is at the end of OST object so that kms is too
4034 small. Fix is to increase kms accordingly in ll_nopage.
4037 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4039 Description: writes to a missing object would leak memory on the OST
4040 Details : If there is an inconsistency between the MDS and OST filesystems,
4041 such that the MDS references an object that doesn't exist, writes
4042 to that object will leak memory due to incorrect cleanup in the
4043 error handling path, eventually running out of memory on the OST.
4048 Description: Creating too long symlink causes lustre errors
4049 Details : Check symlink and name lengths before sending requests to MDS.
4052 Frequency : only if flock is enabled (not on by default)
4054 Description: posix locks not released on fd closure on 2.6.9+
4055 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4056 kernels, this caused such locks not to be released on fd close and
4057 then assertions on fs unmount about still used locks.
4060 Frequency : MDS failover only, very rarely
4062 Description: clients may get ASSERTION(granted_lock != NULL)
4063 Details : When request was taking a long time, and a client was resending
4064 a getattr by name lock request. The were multiple lock requests
4065 with the same client lock handle and
4066 mds_getattr_name->fixup_handle_for_resent_request found one of the
4067 lock handles but later failed with ASSERTION(granted_lock != NULL).
4072 Description: handle->h_buffer_credits > 0, assertion failure
4073 Details : h_buffer_credits is zero after truncate, causing assertion
4074 failure. This patch extends the transaction or creates a new
4078 Frequency : NFS re-export or patchless client
4079 Bugzilla : 11179, 10796
4080 Description: Crash on NFS re-export node (__d_move)
4081 Details : We do not want to hash the dentry if we don't have a lock.
4082 But if this dentry is later used in d_move, we'd hit uninitialised
4083 list head d_hash, so we just do this to init d_hash field but
4084 leave dentry unhashed.
4087 Frequency : NFS re-export or patchless client
4089 Description: NFS exports has problem with symbolic link
4090 Details : lustre client didn't properly install dentry when re-exported
4091 to NFS or running patchless client.
4094 Frequency : NFS re-export or patchless client
4096 Description: Various nfs/patchless fixes.
4097 Details : fixes reuse disconected alias for lookup process - this fixes
4098 warning "find_exported_dentry: npd != pd",
4099 fix permission error with open files at nfs.
4100 fix apply umask when do revalidate.
4103 Frequency : occasional
4105 Description: Crash on NFS re-export node
4106 Details : calling clear_page() on the wrong pointer triggered oops in
4107 generic_mapping_read().
4110 Frequency : rarely, using O_DIRECT IO
4112 Description: unaligned directio crashes client with LASSERT
4113 Details : check for unaligned buffers before trying any requests.
4116 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4118 Description: stale data returned from RAID cache
4119 Details : If only a small amount of IO is done to the RAID device before
4120 reading it again it is possible to get stale data from the RAID
4121 cache instead of reading it from disk.
4124 Frequency : always for sles10 kernel
4126 Description: sles10 support
4127 Details : ll_follow_link: compile fixes and using of nd_set_link
4128 under newer kernels.
4131 Frequency : depends on arch, kernel and compiler version, always on sles10
4134 Description: recursive or deep enough symlinks cause stack overflow
4135 Details : getting rid of large stack-allocated variable in
4139 Frequency : depends on hardware
4141 Description: lustre write performance loss in the SLES10 kernel
4142 Details : the performance loss is caused by using of write barriers in the
4143 ext3 code. The SLES10 kernel turns barrier support on by
4144 default. The fix is to undo that change for ldiskfs.
4146 ------------------------------------------------------------------------------
4148 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4150 * Support for kernels:
4151 2.6.9-42.0.3EL (RHEL 4)
4152 2.6.5-7.276 (SLES 9)
4153 2.4.21-47.0.1.EL (RHEL 3)
4154 2.6.12.6 vanilla (kernel.org)
4158 Frequency : quota enabled and large files being deleted
4160 Description: releasing more than 4GB of quota at once hangs OST
4161 Details : If a user deletes more than 4GB of files on a single OST it
4162 will cause the OST to spin in an infinite loop. Release
4163 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4168 Description: statfs data retrieved from /proc may be stale or zero
4169 Details : When reading per-device statfs data from /proc, in the
4170 {kbytes,files}_{total,free,avail} files, it may appear
4171 as zero or be out of date.
4174 Frequency : systems with MD RAID1 external journal devices
4176 Description: lconf's call to blkid is confused by RAID1 journal devices
4177 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4178 of returning all block devices that match the journal UUID.
4181 Frequency : always, for aggregate stripe size over 4GB
4183 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4184 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4185 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4188 Frequency : always if "lfs find" used on a local file/directory
4190 Description: "lfs find" segfaults if used on a local file/directory
4191 Details : The case where a directory component was not specified wasn't
4192 handled correctly. Handle this properly.
4195 Frequency : always on ppc64
4197 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4198 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4199 perform endianness conversion.
4202 Frequency : rarely (truncate to non-zero file size after write under load)
4203 Bugzilla : 10730, 10687
4204 Description: Files padded with zeros to next 4K multiple
4205 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4206 it is possible that files that are truncated to a non-zero size
4207 immediately after being written are filled with zero bytes beyond
4208 the truncated size. No file data is lost.
4210 Severity : enhancement
4212 Description: Allow recovery/failover for liblustre clients.
4213 Details : liblustre clients were unaware of failover configurations until
4216 Severity : enhancement
4218 Description: user file locks should fail when not mounting with flock option
4219 Details : Set up an error-returning stub in ll_file_operations.lock field
4220 to prevent incorrect behaviour when client is mounted without
4221 flock option. Also, set up properly f_op->flock field for
4225 Frequency : always on ia64
4227 Description: "lfs df" loops on printing out MDS statfs information
4228 Details : The obd_ioctl_data was not initialized and in some systems
4229 this caused a failure during the ioctl that did not return
4230 an error. Initialize the struct and return an error on failure.
4233 Frequency : SLES 9 only
4235 Description: Error of copying files with lustre special EAs as root
4236 Details : Client side always return success for setxattr call for lustre
4237 special xattr (currently only "trusted.lov").
4240 Frequency : rarely on clusters with both ia64+i386 clients
4242 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4243 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4244 clients are doing concurrent writes to the same file it is
4245 possible that the ia64 clients may LASSERT because the OST
4246 extent locks are not PAGE_SIZE aligned. Ensure that grown
4247 locks are always aligned on the request boundary.
4250 Frequency : specific use, occasional
4252 Description: Overwriting in use executable truncates on-disk binary image
4253 Details : If one node attempts to overwrite an executable in use by
4254 another node, we now correctly return ETXTBSY instead of
4255 truncating the file.
4257 Severity : enhancement
4259 Description: Async OSC create to avoid the blocking unnecessarily.
4260 Details : If an OST has no remaining object, system will block on the creating
4261 when need to create a new object on this OST. Now, ways use
4262 pre-created objects when available, instead of blocking on an
4263 empty osc while others are not empty. If we must block, we block
4264 for the shortest possible period of time.
4269 Description: chmod on Lustre root is propagated to other clients
4270 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4271 invalid by the follow_mount time.
4274 Frequency : liblustre clients only
4276 Description: Race in 'instant cancel' lock handling could lead to such locks
4277 never to be granted in case of SMP MDS
4278 Details : Do not destroy not yet granted but cbpending locks in
4282 Frequency : replay/resend of open
4284 Description: non null lock assetion failure in mds_intent_policy
4285 Details : Trying to replay/resend lockless open requests resulted in
4286 mds_open() returning 0 with no lock. Now it sets a flag if
4287 a lock is going to be returned.
4289 Severity : enhancement
4291 Description: Checksum enhancements
4292 Details : New checksum enhancements allow for resending RPCs that failed
4295 Severity : enhancement
4297 Description: Tunables on number of dirty pages in cacche
4298 Details : Allow to set limit on number of dirty pages cached.
4303 Description: client crash on unmount - lock still has references
4304 Details : In some error handling cases it was possible to leak a lock
4305 reference on a client while accessing a file. This was not
4306 harmful to the client during operation, but would cause the
4307 client to crash when the filesystem is unmounted.
4310 Frequency : specific case, rare
4312 Description: ETXTBSY on mds though file not in use by client
4313 Details : ETXTBSY is no longer incorrectly returned when attempting to
4314 chmod or chown a directory that the user previously tried to
4315 execute or a currently-executing binary.
4318 Frequency : extremely rare except on liblustre-based clients
4320 Description: Lustre space not freed when files are deleted
4321 Details : Clean up open-unlinked files after client eviction. Previously
4322 the unlink was skipped and the files remained as orphans.
4327 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4328 Details : In some cases it was possible to send a blocking callback to
4329 a client doing a glimpse, even though that client didn't get
4330 a lock granted. When the glimpse lock is cancelled on the OST
4331 the freed lock is left on the waiting list and corrupted the list.
4334 Frequency : all core dumps
4336 Description: Broke core dumps to lustre
4337 Details : Negative dentry may be unhashed if parent does not have UPDATE
4338 lock, but some callers, e.g. do_coredump, expect dentry to be
4339 hashed after successful create, hash it in ll_create_it.
4341 ------------------------------------------------------------------------------
4343 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4345 * Support for kernels:
4346 2.6.9-42.0.2.EL (RHEL 4)
4347 2.6.5-7.276 (SLES 9)
4348 2.4.21-40.EL (RHEL 3)
4349 2.6.12.6 vanilla (kernel.org)
4353 Frequency : always on RHEL 3
4355 Description: Number of open files grows over time
4356 Details : The number of open files grows over time, whether or not
4357 Lustre is started. This was due to a filp leak introduced
4358 by one of our kernel patches.
4360 ------------------------------------------------------------------------------
4362 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4364 * Support for kernels:
4365 2.6.9-42.EL (RHEL 4)
4366 2.6.5-7.276 (SLES 9)
4367 2.4.21-40.EL (RHEL 3)
4368 2.6.12.6 vanilla (kernel.org)
4373 Bugzilla : 5719, 9635, 9792, 9684
4374 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4375 Details : If a server is under heavy load and cannot reply to new
4376 connection requests before the client resends the (re)connect,
4377 the connection handling code can behave badly if two service
4378 threads are concurrently handing separate (re)connections from
4379 the same client. Add better locking to the connection handling
4380 code, and ensure that only a single connection will be processed
4381 for a given client UUID, even if the lock is dropped.
4383 Severity : enhancement
4385 Description: add TCP zero-copy support to kernel
4386 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4387 sends if the hardware supports scatter-gather and checksumming.
4388 This allows socklnd to do client-write and server-read more
4389 efficiently and reduce CPU utilization from skbuf copying.
4392 Frequency : only if NFS exporting from client
4394 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4395 Details : If setting attributes on a file created under NFS that had
4396 never been opened it would be possible to oops the client
4397 if the file had no objects.
4400 Frequency : always for liblustre
4402 Description: liblustre client does MDS+OSTs setattr RPC for each write
4403 Details : When doing a write from a liblustre client, the client
4404 incorrectly issued an RPC to the MDS and each OST the file was
4405 striped over in order to update the timestamps. When writing
4406 with small chunks and many clients this could overwhelm the MDS
4407 with RPCs. In all cases it would slow down the write because
4408 these RPCs are unnecessary.
4410 Severity : enhancement
4412 Description: allow number of MDS service threads to be changed at module load
4413 Details : It is now possible to change the number of MDS service threads
4414 running. Adding "options mds mds_num_threads={N}" to the MDS's
4415 /etc/modprobe.conf will set the number of threads for the next
4416 time Lustre is restarted (assuming the "mds" module is also
4417 reloaded at that time). The default number of threads will
4418 stay the same, 32 for most systems.
4423 Description: OST crash if filesystem is unformatted or corrupt
4424 Details : If an OST is started on a device that has never been formatted
4425 or if the filesystem is corrupt and cannot even mount then the
4426 error handling cleanup routines would dereference a NULL pointer.
4431 Description: NULL pointer deref in llap_from_page.
4432 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4433 incorrect) mapping if the page was truncated/reclaimed while it was
4434 searched for. Check for this condition and skip such pages when
4435 doing readahead. Introduce extra check to llap_from_page() to
4436 verify page->mapping->host is non-NULL (so page is not anonymous).
4439 Frequency : Sometimes when using sys_sendfile
4441 Description: "page not covered by a lock" warnings from ll_readpage
4442 Details : sendfile called ll_readpage without right page locks present.
4443 Now we introduced ll_file_sendfile that does necessary locking
4444 around call to generic_file_sendfile() much like we do in
4448 Frequency : with certain MDS communication failures at client mount time
4450 Description: NULL pointer deref after failed client mount
4451 Details : a client connection request may delayed by the network layer
4452 and not be sent until after the PTLRPC layer has timed out the
4453 request. If the client fails the mount immediately it will try
4454 to clean up before the network times out the request. Add a
4455 reference from the request import to the obd device and delay
4456 the cleanup until the network drops the request.
4459 Frequency : occasionally during client (re)connect
4461 Description: assertion failure during client (re)connect
4462 Details : processing a client connection request may be delayed by the
4463 client or server longer than the client connect timeout. This
4464 causes the client to resend the connection request. If the
4465 original connection request is replied in this interval, the
4466 client may trip an assertion failure in ptlrpc_connect_interpret()
4467 which thought it would be the only running connect process.
4470 Frequency : only with obd_echo servers and clients that are rebooted
4472 Description: kernel BUG accessing uninitialized data structure
4473 Details : When running an obd_echo server it did not start the ping_evictor
4474 thread, and when a client was evicted an uninitialized data
4475 structure was accessed. Start the ping_evictor in the RPC
4476 service startup instead of the OBD startup.
4478 Severity : enhancement
4479 Bugzilla : 10193 (patchless)
4480 Description: Remove dependency on various unexported kernel interfaces.
4481 Details : No longer need reparent_to_init, exit_mm, exit_files,
4482 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4485 Frequency : rare (only users of deprecated and unsupported LDAP config)
4487 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4488 Details : LDAP apparently contains 'lustreName' attributes instead of
4489 'name'. A simple remapping of the name is sufficient.
4492 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4494 Description: waiting_locks_callback trips kernel BUG if client is evicted
4495 Details : Running with the dump_on_timeout debug flag turned on makes
4496 it possible that the waiting_locks_callback() can try to dump
4497 the Lustre kernel debug logs from an interrupt handler. Defer
4498 this log dumping to the expired_lock_main() thread.
4500 Severity : enhancement
4502 Description: Support NFS exporting on 2.6 kernels.
4503 Details : Implement non-rawops metadata methods for NFS server to use without
4504 changing NFS server code.
4507 Frequency : very rare (synthetic metadata workload only)
4509 Description: two racing renames might cause an MDS thread to deadlock
4510 Details : Running the "racer" program may cause one MDS thread to rename
4511 a file from being the source of a rename to being the target of
4512 a rename at exactly the same time that another thread is doing
4513 so, and the second thread has already enqueued these locks after
4514 doing a lookup of the target and is trying to relock them in
4515 order. Ensure that we don't try to re-lock the same resource.
4518 Frequency : only very large systems with liblustre clients
4520 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4521 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4522 all OSTs in parallel. This allows the eviction of stale liblustre
4523 clients to proceed much faster than if they were done in series,
4524 and also offers similar improvements for other set_info RPCs.
4529 Description: excessive CPU usage during initial read phase on client
4530 Details : During the initial read phase on a client, it would agressively
4531 retry readahead on the file, consuming too much CPU and impacting
4532 performance (since 1.4.5.8). Improve the readahead algorithm
4533 to avoid this, and also improve some other common cases (read
4534 of small files in particular, where "small" is files smaller than
4535 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4540 Description: MDS crash when receiving packet with unknown intent.
4541 Details : Do not LBUG in unknown intent case, just return -EFAULT
4543 Severity : enhancement
4544 Bugzilla : 9293, 9385
4545 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4546 Details : Do not serialize getattr (non-intent version) and statfs.
4549 Frequency : occasional, when OST network is overloaded/intermittent
4551 Description: client evicted by OST after bulk IO timeout
4552 Details : If a client sends a bulk IO request (read or write) the OST
4553 may evict the client if it is unresposive to its data GET/PUT
4554 request. This is incorrect if the network is overloaded (takes
4555 too long to transfer the RPC data) or dropped the OST GET/PUT
4556 request. There is no need to evict the client at all, since
4557 the pinger and/or lock callbacks will handle this, and the
4558 client can restart the bulk request.
4561 Frequency : Always when mmapping file with no objects
4563 Description: client crashes when mmapping file with no objects
4564 Details : Check that we actually have objects in a file before doing any
4565 operations on objects in ll_vm_open, ll_vm_close and
4571 Description: Request leak when working with deleted CWD
4572 Details : Introduce advanced request refcount tracking for requests
4573 referenced from lustre intent.
4575 Severity : Enhancement
4577 Description: Cache open file handles on client.
4578 Details : MDS now will return special lock along with openhandle, if
4579 requested and client is allowed to hold openhandle, even if unused,
4580 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4581 closing files for every read/write openration.
4583 Severity : Enhancement
4585 Description: Cache open negative dentries on client when possible.
4586 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4587 negative dentries on lock revocation.
4592 Description: Remounting a client read-only wasn't possible with a zconf mount
4593 Details : It wasn't possible to remount a client read-only with llmount.
4595 Severity : enhancement
4596 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4597 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4598 MPICH is included to add Lustre-specific ADIO interfaces.
4599 This is based closely on the UFS ADIO layer and only differs
4600 in file creation, in order to allow the OST striping to be set.
4601 This is user-contributed code and not supported by CFS.
4606 Description: extended inode attributes (immutable, append-only) work improperly
4607 when 2.4 and 2.6 kernels are used on client/server or vice versa
4608 Details : Introduce kernel-independent values for these flags.
4610 Severity : enhancement
4613 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4614 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4615 granularity. Now, Fractional MB granularity is be supported,
4616 this is very useful for low memory system.
4618 Severity : enhancement
4620 Description: Getattr by fid
4621 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4622 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4623 have locks and avoids getattr rpc requests for it on every stat.
4626 Frequency : Always, for filesystems larger than 2TB
4628 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4629 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4630 filesystems larger than 2TB to be mounted with mballoc enabled.
4633 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4635 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4636 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4637 filesystems larger than 2TB is mounted then the kernel will
4638 silently corrupt the start of the filesystem. CONFIG_LBD is
4639 enabled for all CFS-supported kernels, but the possibility of
4640 this happening with a modified kernel config exists.
4642 Severity : enhancement
4644 Description: add client O_DIRECT support for 2.6 kernels
4645 Details : It is now possible to do O_DIRECT reads and writes to files
4646 in the Lustre client mountpoint on 2.6 kernel clients.
4648 Severity : enhancement
4650 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4651 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4652 parallel, not waiting for response from every OST before sending
4653 a rpc to the next OST.
4658 Description: setattr vs write race when updating file timestamps
4659 Details : Client processes that update a file timestamp into the past
4660 right after writing to the file (e.g. tar) it is possible that
4661 the updated file modification time can be reset to the current
4662 time due to a race between processing the setattr and write RPC.
4664 Severity : enhancement
4666 Description: Bring 'lfs find' closer in line with regular Linux find.
4667 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4668 -print0 options and obtains all the needed info through the lustre
4671 Severity : enhancement
4673 Description: support up to 1024 configured devices on one node
4674 Details : change obd_dev array from statically allocated to dynamically
4675 allocated structs as they are first used to reduce memory usage
4680 Description: Flush dirty partially truncated pages during truncate
4681 Details : Immediatelly flush partially truncated pages in filter_setattr,
4682 this way we completely avoid having any pages in page cache on OST
4683 and can retire ugly workarounds during writes to flush such pages.
4688 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4689 Details : protect inode from truncation within vfs_unlink() context
4690 just take a reference before calling vfs_unlink() and release it
4691 when parent's i_sem is free.
4696 Description: last_id value checked outside lock on OST caused LASSERT failure
4697 Details : If there were multiple MDS->OST object precreate requests in
4698 flight, it was possible that the OST's last object id was checked
4699 outside a lock and incorrectly tripped an assertion. Move checks
4700 inside locks, and discard old precreate requests.
4703 Frequency : always, if extents are used on OSTs
4705 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4706 Details : OSTs using the extents format would not zero the high 16 bits of
4707 the index physical block number. This is not a problem for any
4708 OST filesystems smaller than 16TB, and no kernels support ext3
4709 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4710 new/modified files) and can be fixed for existing filesystems
4711 with e2fsprogs-1.39-cfs1.
4716 Description: import connection selection may be incorrect if timer wraps
4717 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4718 import connections to be ignored if the 32-bit jiffies counter
4719 wraps. Use a 64-bit jiffies counter.
4722 Frequency : very large clusters immediately after boot
4724 Description: LNET request buffers exhausted under heavy short-term load
4725 Details : If a large number of client requests are generated on a service
4726 that has previously never seen so many requests it is possible
4727 that the request buffer growth cannot keep up with the spike in
4728 demand. Instead of dropping incoming requests, they are held in
4729 the LND until the RPC service can accept more requests.
4732 Frequency : Sometimes during replay
4734 Description: Assertion failure in ll_local_open after replay.
4735 Details : If replay happened on an open request reply before we were able
4736 to set replay handler, reply will become not swabbed tripping the
4737 assertion in ll_local_open. Now we set the handler right after
4738 recognising of open request
4741 Frequency : very rare
4743 Description: kernel reports "badness in vsnprintf"
4744 Details : Reading from the "recovery_status" /proc file in small chunks
4745 may cause a negative length in lprocfs_obd_rd_recovery_status()
4746 call to vsnprintf() (which is otherwise harmless). Exit early
4747 if there is no more space in the output buffer.
4749 Severity : enhancement
4751 Description: clear OBD RPC statistics by writing to them
4752 Details : It is now possible to clear the OBD RPC statistics by writing
4753 to the "stats" file.
4758 Description: Client mtime is not the same on different clients after utimes
4759 Details : In some cases, the client was using the utimes() syscall on
4760 a file cached on another node. The clients now validate the
4761 ctime from the MDS + OSTs to determine which one is right.
4766 Description: Inability to activate failout mode
4767 Details : lconf script incorrectly assumed that in pythong string's numeric
4768 value is used in comparisons.
4771 Frequency : always with multiple stripes per file
4773 Description: Inefficient object allocation for mutli-stripe files
4774 Details : When selecting which OSTs to stripe files over, for files with
4775 a stripe count that divides evenly into the number of OSTs,
4776 the MDS is always picking the same starting OST for each file.
4777 Return the OST selection heuristic to the original design.
4782 Description: mount failures may take full timeout to return an error
4783 Details : Under some heavy load conditions it is possible that a
4784 failed mount can wait for the full obd_timeout interval,
4785 possibly several minutes, before reporting an error.
4786 Instead return an error as soon as the status is known.
4788 Frequency : quota enabled and large files being deleted
4790 Description: releasing more than 4GB of quota at once hangs OST
4791 Details : If a user deletes more than 4GB of files on a single OST it
4792 will cause the OST to spin in an infinite loop. Release
4793 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4798 Description: statfs data retrieved from /proc may be stale or zero
4799 Details : When reading per-device statfs data from /proc, in the
4800 {kbytes,files}_{total,free,avail} files, it may appear
4801 as zero or be out of date.
4804 Frequency : systems with MD RAID1 external journal devices
4806 Description: lconf's call to blkid is confused by RAID1 journal devices
4807 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4808 of returning all block devices that match the journal UUID.
4811 Frequency : always, for aggregate stripe size over 4GB
4813 Description: assertion fails when trying to use 4GB stripe size
4814 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4815 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4818 Frequency : always on ppc64
4820 Description: the first write on an ext3 filesystem with mballoc got stuck
4821 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4822 perform endianness conversion.
4824 ------------------------------------------------------------------------------
4826 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4828 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4829 INTEROPERATE with older versions automatically. Please read the
4830 user documentation before upgrading any part of a live system.
4831 * WARNING: Lustre networking configuration changes are required with
4832 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4835 * Support for kernels:
4836 2.6.9-22.0.2.EL (RHEL 4)
4837 2.6.5-7.244 (SLES 9)
4838 2.6.12.6 vanilla (kernel.org)
4841 Severity : enhancement
4842 Bugzilla : 7981/8208
4843 Description: Introduced Lustre Networking (LNET)
4844 Details : LNET is new networking infrastructure for Lustre, it includes
4845 a reorganized network configuration mode (see the user
4846 documentation for full details) as well as support for routing
4847 between different network fabrics. Lustre Networking Devices
4848 (LNDs) for the supported network fabrics have also been
4849 created for this new infrastructure.
4851 Severity : enhancement
4852 Description: Introduced Access control lists
4853 Details : clients can set ACLs on files and directories in order to have
4854 more fine-grained permissions than the standard Unix UGO+RWX.
4855 The MDS must be started with the "-o acl" mount option.
4857 Severity : enhancement
4858 Description: Introduced filesystem quotas
4859 Details : Administrators may now establish per-user quotas on the
4862 Severity : enhancement
4864 Description: Configuration change for the XT3
4865 The PTLLND is now used to run Lustre over Portals on the XT3
4866 The configure option(s) --with-cray-portals are no longer used.
4867 Rather --with-portals=<path-to-portals-includes> is used to
4868 enable building on the XT3. In addition to enable XT3 specific
4869 features the option --enable-cray-xt3 must be used.
4874 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4875 Details : A race between allocating a new llcd and re-getting the llcd_lock
4876 allowed another thread to grab newly-allocated llcd.
4878 Severity : enhancement
4880 Description: 2.6 OST async journal commit and locking fix to improve performance
4881 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4882 2.6 kernels are now async as they already were in 2.4 kernels so
4883 that they can commit concurrently with the network bulk transfer.
4884 For block-allocated files the filter allocation semaphore is held
4885 to avoid filesystem fragmentation during allocation. BKL lock
4886 removed for 2.6 xattr operations where it is no longer needed.
4891 Description: lconf incorrectly determined whether two IP networks could talk
4892 Details : In some more complicated routing and multiple-network
4893 configurations, lconf will avoid trying to make a network
4894 connection to a disjoint part of the IP space. It was doing the
4895 math incorrectly for one set of cases.
4900 Description: Fix for potential infinite loop processing records in an llog.
4901 Details : If an llog record is corrupted/zeroed, it is possible to loop
4902 forever in llog_process(). Validate the llog record length
4903 and skip the remainder of the block on error.
4906 Frequency : occasional (liblustre only)
4908 Description: liblustre could not open files whose last component is a symlink
4909 Details : sysio_path_walk() would incorrectly pass the open intent to
4910 intermediate path components.
4913 Frequency : rare (liblustre only with non-standard tuning)
4914 Bugzilla : 7201 (7350)
4915 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4916 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4917 after it has been released, only do this for referenced locks
4919 Severity : enhancement
4921 Description: specifying an (invalid) directory default stripe_size of -1
4922 would reset the directory default striping
4923 Details : stripe_size -1 was used internally to signal directory stripe
4924 removal, now use "all default" to signal dir stripe removal
4925 as a directory striping of "all default" is not useful
4928 Frequency : common for large clusters running liblustre clients
4930 Description: doing an ls when liblustre clients are running is slow
4931 Details : sending a glimpse AST to a liblustre client waits for every AST
4932 to time out, as liblustre clients will not respond. Since they
4933 cannot cache data we refresh the OST lock LVB from disk instead.
4935 Severity : enhancement
4937 Description: doing an ls at the same time as file IO can be slow
4938 Details : enqueue and other "small" requests can be blocked behind many
4939 large IO requests. Create a new OST IO portal for non-IO
4940 requests so they can be processed faster.
4943 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4945 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4946 created file with mode 000
4947 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4948 allow an ftruncate() as a non-root user to a file with mode 000.
4949 Lustre now respects this flag to disable mode checks when
4950 truncating a file owned by the user
4953 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4956 Description: Revoking locks from clients that went dead or catatonic might take
4958 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4959 cancellation of such locks instant on servers without waiting for
4960 any reply from clients. Clients drops these locks when cancel
4961 notification from server is received without replying.
4964 Frequency : liblustre-only, when liblustre client dies or becomes busy
4966 Description: Doing ls on Linux clients can take a long time with active
4968 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4969 granting such locks to it in the first place if possible. Locks
4970 are taken by proxy on the OST during the read or write and
4971 dropped immediately afterward. Add connect flags handling, do
4972 not grant locks to liblustre clients for glimpse ASTs.
4974 Severity : enhancement
4976 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4977 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4978 sequential reads and behaves badly with applications doing random
4979 reads. Improve it by reading ahead at least read region, and
4980 avoiding excessive large RPC for small reads.
4982 Severity : enhancement
4984 Description: Creating more than 1000 files for a single job may cause a load
4985 imbalance on the OSTs if there are also a large number of OSTs.
4986 Details : qos_prep_create() uses an OST index reseed value that is an
4987 even multiple of the number of available OSTs so that if the
4988 reseed happens in the middle of the object allocation it will
4989 still utilize the OSTs as uniformly as possible.
4994 Description: OST or MDS may oops in ping_evictor_main()
4995 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4996 but doesn't restart at beginning of obd_exports_timed list
4999 Severity : enhancement
5001 Description: improve by-nid export eviction on the MDS and OST
5002 Details : allow multiple exports with the same NID to be evicted at one
5003 time without re-searching the exports list.
5006 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5008 Description: MDS may oops in groups_free()
5009 Details : in rare race conditions a newly allocated group_info struct is
5010 freed again, and this can be NULL. The 2.4 compatibility code
5011 for groups_free() checked for a NULL pointer, but 2.6 did not.
5014 Frequency : common for liblustre clients doing little filesystem IO
5015 Bugzilla : 9352, 7313
5016 Description: server may evict liblustre clients accessing contended locks
5017 Details : if a client is granted a lock or receives a completion AST
5018 with a blocking AST already set it would not reply to the AST
5019 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5022 Frequency : lfs setstripe, only systems with more than 160 OSTs
5024 Description: unable to set striping with a starting offset beyond OST 160
5025 Details : llapi_create_file() incorrectly limited the starting stripe
5026 index to the maximum single-file stripe count.
5029 Frequency : LDAP users only
5031 Description: lconf did not handle in-kernel recovery with LDAP properly
5032 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5034 Severity : enhancement
5036 Description: bind OST threads to NUMA nodes to improve performance
5037 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5038 node and do their allocations there to localize memory access
5040 Severity : enhancement
5042 Description: llmount can determine client NID directly from Myrinet (GM)
5043 Details : the client NID code from gmnalnid was moved directly into
5044 llmount, removing the need to use this or specifying the
5045 client NID explicitly when mounting GM clients with zeroconf
5048 Frequency : if client is started with down MDS
5050 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5051 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5052 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5053 ptlrpc_import_delayed_req() because we didn't check if the
5054 request was interrupted, and we also didn't break out of the
5055 event loop if there was no timeout
5060 Description: data loss during non-page-aligned writes to a single file from
5061 both multiple nodes and multiple threads on one node at same time
5062 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5063 inconsistency led to false short-reads, that were cached and later
5064 used by ->prepare_write() to fill in partially written page,
5065 leading to data loss.
5068 Frequency : always, if lconf --abort_recovery used
5070 Description: lconf --abort_recovery fails with 'Operation not supported'
5071 Details : lconf was attempting to abort recovery on the MDT device and not
5074 Severity : enhancement
5076 Description: remove cleanup logs
5077 Details : replace lconf-generated cleanup logs with lustre internal
5078 cleanup routines. Eliminates the need for client-cleanup and
5081 Severity : enhancement
5083 Description: add support for EAs (user and system) on lustre filesystems
5084 Details : it is now possible to store extended attributes in the Lustre
5085 client filesystem, and with the user_xattr mount option it
5086 is possible to allow users to store EAs on their files also
5088 Severity : enhancement
5090 Description: Add possibility (config option) to show minimal available OST free
5092 Details : When compiled with --enable-mindf configure option, statfs(2)
5093 (and so, df) will return least minimal free space available from
5094 all OSTs as amount of free space on FS, instead of summary of
5095 free spaces of all OSTs.
5097 Severity : enhancement
5099 Description: do not expand extent locks acquired on OST-side
5100 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5101 by server: they are not cached anyway.
5104 Frequency : when mmap is used/binaries executed from Lustre
5106 Description: Unmmap pages before throwing them away from read cache.
5107 Details : llap_shrink cache now attempts to unmap pages before discarding
5108 them (if unmapping failed - do not discard). SLES9 kernel has
5109 extra checks that trigger if this unmapping is not done first.
5114 Description: lconf didn't resolve symlinks before checking to see whether a
5115 given mountpoint was already in use
5118 Frequency : when migrating failover services
5119 Bugzilla : 6395, 9514
5120 Description: When migrating a subset of services from a node (e.g. failback
5121 from a failover service node) the remaining services would
5122 time out and evict clients.
5123 Details : lconf --force (implied by --failover) sets the global obd_timeout
5124 to 5 seconds in order to quickly disconnect, but this caused
5125 other RPCs to time out too quickly. Do not change the global
5126 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5128 Severity : enhancement
5129 Frequency : if MDS is started with down OST
5130 Bugzilla : 9439,5706
5131 Description: Allow startup/shutdown of an MDS without depending on the
5132 availability of the OSTs.
5133 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5134 Add appropriate locking and lov-osc refcounts for safe
5135 cleaning. Add osc abort_inflight calls in case the
5136 synchronize never started.
5139 Frequency : occasional (Cray XT3 only)
5141 Description: root not authorized to access files in CRAY_PORTALS environment
5142 Details : The client process capabilities were not honoured on the MDS in
5143 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5144 been accessed by an authorized user then root was able to access
5145 the file on the local client also. The root user capabilities
5146 are now allowed on the MDS, as this environment has secure UID.
5149 Frequency : occasional
5151 Description: ldiskfs "too long searching" message happens too often
5152 Details : A debugging message (otherwise harmless) prints too often on
5153 the OST console. This has been reduced to only happen when
5154 there are fragmentation problems on the filesystem.
5159 Description: Division by zero in statfs when all OSCs are inactive
5160 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5161 lov_statfs divides by value returned by lov_get_stripecnt().
5165 Bugzilla : 9489, 3273
5166 Description: First write from each client to each OST was only 4kB in size,
5167 to initialize client writeback cache, which caused sub-optimal
5168 RPCs and poor layout on disk for the first written file.
5169 Details : Clients now request an initial cache grant at (re)connect time
5170 and so that they can start streaming writes to the cache right
5171 away and always do full-sized RPCs if there is enough data.
5172 If the OST is rebooted the client also re-establishes its grant
5173 so that client cached writes will be honoured under the grant.
5178 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5179 Details : Now I/O RPCs go to different portal number and (presumably) fast
5180 lock requests (and glimses) and other RPCs get their own service
5181 threads pool that should be able to service those RPCs
5184 Severity : enhancement
5186 Description: Ability to exchange lustre version between client and servers and
5187 issue warnings at client side if client is too old. Also for
5188 liblustre clients there is ability to refuse connection of too old
5190 Details : New 'version' field is added to connect data structure that is
5191 filled with version info. That info is later checked by server and
5195 Frequency : rare, liblustre only.
5196 Bugzilla : 9296, 9581
5197 Description: Two simultaneous writes from liblustre at offset within same page
5198 might proceed at the same time overwriting eachother with stale
5200 Details : I/O lock withing llu_file_prwv was released too early, before data
5201 actually was hitting the wire. Extended lock-holding time until
5202 server acknowledges receiving data.
5205 Frequency : extremely rare. Never observed in practice.
5207 Description: avoid generating lustre_handle cookie of 0.
5208 Details : class_handle_hash() generates handle cookies by incrementing
5209 global counter, and can hit 0 occasionaly (this is unlikely, but
5210 not impossible, because initial value of cookie counter is
5211 selected randonly). Value of 0 is used as a sentinel meaning
5212 "unassigned handle" --- avoid it. Also coalesce two critical
5213 sections in this function into one.
5215 Severity : enhancement
5217 Description: allow liblustre clients to delegate truncate locking to OST
5218 Details : To avoid overhead of locking, liblustre client instructs OST to
5219 take extent lock in ost_punch() on client's behalf. New connection
5220 flag is added to handle backward compatibility.
5222 Severity : enhancement
5223 Bugzilla : 4928, 7341, 9758
5224 Description: allow number of OST service threads to be specified
5225 Details : a module parameter allows the number of OST service threads
5226 to be specified via "options ost ost_num_threads={N}" in the
5227 OSS's /etc/modules.conf or /etc/modprobe.conf.
5231 Bugzilla : 6146, 9635, 9895
5232 Description: servers crash with bad pointer in target_handle_connect()
5233 Details : In rare cases when a client is reconnecting it was possible that
5234 the connection request was the last reference for that export.
5235 We would temporarily drop the export reference and get a new
5236 one, but this may have been the last reference and the export
5237 was just destroyed. Get new reference before dropping old one.
5239 Severity : enhancement
5240 Frequency : if client is started with failover MDS
5242 Description: Allow multiple MDS hostnames in the mount command
5243 Details : Try to read the configuration from all specified MDS
5244 hostnames during a client mount in case the "primary"
5247 Severity : enhancement
5249 Description: Stop sending data to evicted clients as soon as possible.
5250 Details : Check if the client we are about to send or are sending data to
5251 was evicted already. (Check is done every second of waiting,
5252 for which l_wait_event interface was extended to allow checking
5253 of exit condition at specified intervals).
5256 Frequency : rare, normally only when NFS exporting is done from client
5258 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5260 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5263 Frequency : always, if srand() is called before liblustre initialization
5265 Description: Liblustre uses system PRNG disturbing its usage by user application
5266 Details : Introduce internal to lustre fast and high-quality PRNG for
5267 lustre usage and make liblustre and some other places in generic
5268 lustre code to use it.
5270 Severity : enhancement
5271 Bugzilla : 9477, 9557, 9870
5272 Description: Verify that the MDS configuration logs are updated when xml is
5273 Details : Check if the .xml configuration logs are newer than the config
5274 logs stored on the MDS and report an error if this is the case.
5275 Request --write-conf, or allow starting with --old_conf.
5277 Severity : enhancement
5279 Description: Handle symlinks in the path when checking if Lustre is mounted.
5280 Details : Resolve intermediate symlinks when checking if a client has
5281 mounted a filesystem to avoid duplicate client mounts.
5286 Description: lconf can hit an error exception but still return success.
5287 Details : The lconf command catches the Command error exception at the top
5288 level script context and will exit with the associated exit
5289 status, but doesn't ensure that this exit status is non-zero.
5294 Description: failure of ptlrpc thread startup can cause oops
5295 Details : Starting a ptlrpc service thread can fail if there are a large
5296 number of threads or the server memory is very fragmented.
5297 Handle this without oopsing.
5300 Frequency : always, only if liblustre and non-default acceptor port was used
5302 Description: liblustre cannot connect to servers with non-default acceptor port
5303 Details : tcpnal_set_default_params() was not called and was therefore
5304 ignoring the environment varaible TCPNAL_PORT, as well as other
5305 TCPNAL_ environment variables
5310 Description: two objects could be created on the same OST for a single file
5311 Details : If an OST is down, in some cases it was possible to create two
5312 objects on a single OST for a single file. No problems other
5313 than potential performance impact and spurious error messages.
5317 Bugzilla : 5681, 9562
5318 Description: Client may oops in ll_unhash_aliases
5319 Details : Client dcache may become inconsistent in race condition.
5320 In some cases "getcwd" can fail if the current directory is
5326 Description: Inode refcounting problems in NFS export code
5327 Details : link_raw functions used to call d_instantiate without obtaining
5328 extra inode reference first.
5332 Bugzilla : 9942, 9903
5333 Description: Referencing freed requests leading to crash, memleaks with NFS.
5334 Details : We used to require that call to ll_revalidate_it was always
5335 followed by ll_lookup_it. Also with revalidate_special() it is
5336 possible to call ll_revalidate_it() twice for the same dentry
5337 even if first occurence returned success. This fix changes semantic
5338 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5339 reference on a request referred from the intent.
5340 ll_intent_release() then releases such a request.
5343 Frequency : rare, normally benchmark loads only
5345 Description: unlinked inodes were kept in memory on the client
5346 Details : If a client is repeatedly creating and unlinking files it
5347 can accumulate a lot of stale inodes in the inode slab cache.
5348 If there is no other client load running this can cause the
5349 client node to run out of memory. Instead flush old inodes
5350 from client cache that have the same inode number as a new inode.
5353 Frequency : SLES9 2.6.5 kernel and long filenames only
5354 Bugzilla : 9969, 10379
5355 Description: utime reports stale NFS file handle
5356 Details : SLES9 uses out-of-dentry names in some cases, which confused
5357 the lustre dentry revalidation. Change it to always use the
5361 Frequency : rare, unless heavy write-truncate concurrency is continuous
5362 Bugzilla : 4180, 6984, 7171, 9963, 9331
5363 Description: OST becomes very slow and/or deadlocked during object unlink
5364 Details : filter_destroy() was holding onto the parent directory lock
5365 while truncating+unlinking objects. For very large objects this
5366 may block other threads for a long time and slow overall OST
5367 responsiveness. It may also be possible to get a lock ordering
5368 deadlock in this case, or run out of journal credits because of
5369 the combined truncate+unlink. Solution is to do object truncate
5370 first in one transaction without parent lock, and then do the
5371 final unlink in a new transaction with the parent lock. This
5372 reduces the lock hold time dramatically.
5375 Frequency : rare, 2.4 kernels only
5377 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5378 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5379 before unmounting their filesystem in order to avoid deadlock.
5380 The kernel_locked() function in 2.4 kernels only checks whether
5381 the kernel lock is held, not whether it is this process that is
5382 holding it as 2.6 kernels do.
5387 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5388 Details : The client ptlrpc code may be trying to reconnect to a down
5389 server before a previous connection attempt has timed out.
5390 Increase the reconnect interval to be longer than the connection
5391 timeout interval to avoid sending duplicate connections to
5395 Frequency : echo_client brw_test command
5397 Description: fix echo_client to work with OST preallocated code
5398 Details : OST preallocation code (5137) didn't take echo_client IO path
5399 into account: echo_client calls filter methods outside of any
5400 OST thread and, hence, there is no per-thread preallocated
5401 pages and buffers to use. Solution: hijack pga pages for IO. As
5402 a byproduct, this avoids unnecessary data copying.
5406 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5407 Description: Client can oops in mdc_commit_close() after open replay
5408 Details : It was possible for the MDS to return an open request with no
5409 transaction number in mds_finish_transno() if the client was
5410 evicted, but without actually returning an error. Clients
5411 would later try to replay that open and may trip an assertion
5412 Simplify the client close codepath, and always return an error
5413 from the MDS in case the open is not successful.
5416 Frequency : rare, 2.6 OSTs only
5418 Description: OST may deadlock under high load on fragmented files
5419 Details : If there was a heavy load and highly-fragmented OST filesystems
5420 it was possible to have all the OST threads deadlock waiting on
5421 allocation of biovecs, because the biovecs were not released
5422 until the entire RPC IO was completed. Instead, release biovecs
5423 as soon as they are complete to ensure forward IO progress.
5425 Severity : enhancement
5427 Description: Support for specifying external journal device at mount
5428 Details : If an OST or MDS device is formatted with an external journal
5429 device, this device major/minor is stored in the ext3 superblock
5430 and may not be valid for failover. Allow detecting and
5431 specifying the external journal at mount time.
5436 Description: Mounting an MDS with pending unlinked files may cause oops
5437 Details : target_finish_recovery() calls mds_postrecov() which returned
5438 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5439 considers this an error and immediately begins cleaning up the
5440 lov, just after starting the mds_lov process
5442 Severity : enhancement
5444 Description: Implement 'lfs df' to report actual free space on per-OST basis
5445 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5446 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5447 usage of objects; '-h' to report in human readable format.
5449 ------------------------------------------------------------------------------
5451 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5458 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5459 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5460 at startup during orphan unlinking. Instead, skip buddy update
5461 before setup, as it will be regenerated after recovery is complete.
5466 Description: If an OST is inactive, its locks might reference stale inodes.
5467 Details : lov_change_cbdata() must iterate over all namespaces, even if
5468 they are inactive to clear inode references from the lock.
5470 Severity : enhancement
5471 Frequency : occasional, if non-standard max_dirty_mb used
5473 Description: Client will block write RPCs if not enough grant
5474 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5475 then the client will block writes while waiting for another RPC
5476 to complete instead of consuming its dirty limit. With change
5477 we get improved performance when max_dirty_mb is small.
5479 Severity : enhancement
5480 Bugzilla : 3389, 6253
5481 Description: Add support for supplementary groups on the MDS.
5482 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5483 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5484 lookups for user supplementary groups into a cache.
5488 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5489 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5490 for liblustre O_EXCL is enforced by MDS.
5494 Description: readdir never returns NULL in liblustre.
5495 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5496 dentry in struct dirent.
5500 Frequency : liblustre only, depends on application IO pattern
5501 Description: liblustre clients evicted if not contacting servers
5502 Details : Don't put liblustre clients into the ping_evictor list, so
5503 they will not be evicted by the pinger ever.
5505 Severity : enhancement
5507 Description: Add ability to evict clients by NID from MDS.
5508 Details : By echoing "nid:$NID" string into
5509 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5510 $NID would be instantly evicted from this MDS and from all active
5511 OSTs connected to it.
5515 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5516 Details : lookup_it_finish() used to query file size from OSTs that was not
5521 Description: service threads change working directory to that of init
5522 Details : Starting lustre service threads may pin the working directory
5523 of the parent thread, making that filesystem busy. Threads
5524 now change to the working directory of init to avoid this.
5528 Frequency : during shutdown only
5529 Description: shutdown with a failed MDS or OST can cause unmount to hang
5530 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5535 Frequency : 2.6 only
5536 Description: chmod/chown may include an extra supplementary group
5537 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5538 supplementary group and if none is specified this is used.
5541 Bugzilla : 5479 (6816)
5543 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5544 Details : If lookup is in progress on a file that is unlinked we might try
5545 to revalidate the inode and fail in revalidate after lookup is
5546 complete and ll_file_open() enqueues the open again but
5547 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5550 Frequency : always, if lconf --abort_recovery used
5552 Description: lconf --abort_recovery fails with 'Operation not supported'
5553 Details : lconf was attempting to abort recovery on the MDT device and not
5556 ------------------------------------------------------------------------------
5558 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5563 Frequency : rare (only unsupported configurations with a node running as an
5565 Bugzilla : 6514, 5137
5566 Description: Mounting a Lustre file system on a node running as an OST could
5568 Details : OSTs now preallocates memory needed to write out data at
5569 startup, instead of when needed, to avoid having to
5570 allocate memory in possibly low memory situations.
5571 Specifically, if the file system is mounted on on OST,
5572 memory pressure could force it to try to write out data,
5573 which it needed to allocate memory to do. Due to the low
5574 memory, it would be unable to do so and the node would
5575 become unresponsive.
5577 Severity : enhancement
5579 Description: Addition of lconf --service command line option
5580 Details : lconf now accepts a '--service <arg>' option, which is
5581 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5583 Severity : enhancement
5585 Description: Failover mode is now the default for OSTs.
5586 Details : By default, OSTs will now run in failover mode. To return to
5587 the old behaviour, add '--failout' to the lmc line for OSTs.
5589 Severity : enhancement
5591 Description: Health checks are now provided for MDS and OSTs
5592 Details : Additional detailed health check information on MSD and OSTs
5593 is now provided through the procfs health_check value.
5596 Frequency : occasional, depends on IO load
5598 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5599 after numerous create/delete cycles
5600 Details : The ext3 inode allocation policy would not allocate new inodes
5601 very well on the OSTs because there are no new directories
5602 being created. Instead we look for groups with free space if
5603 the parent directories are nearly full.
5608 Description: Network or server problems during mount may cause partially
5609 mounted clients instead of returning an error.
5610 Details : The config llog parsing code may overwrite the error return
5611 code during mount error handling, returning success instead
5617 Description: MDS can fail to allocate large reply buffers
5618 Details : After long uptimes the MDS can fail to allocate large reply
5619 buffers (e.g. zconf client mount config records) due to memory
5620 fragmentation or consumption by the buffer cache. Preallocate
5621 some large reply buffers so that these replies can be sent even
5622 under memory pressure.
5626 Frequency : rare (liblustre)
5627 Description: fsx running with liblustre complained that using truncate() to
5628 extend the file doesn't work. This patch corrects that issue.
5629 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5630 ATTR_SIZE and lsm use in llu_setattr_raw.
5634 Frequency : rare, only 2.6 kernels
5635 Description: Unusual file access patterns on the MDS may result in inode
5636 data being lost in very rare circumstances.
5637 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5638 code in the 2.6 kernel caused the inode and/or EA data not to
5639 be read from disk, causing single-file corruption.
5643 Frequency : rare, only 2.6 filesystems using extents
5644 Description: Heavy concurrent write and delete load may cause data corruption.
5645 Details : It was possible under high-load situations to have an extent
5646 metadata block in the block device cache from a just-unlinked
5647 file overwrite a newly-allocated data block. We now unmap any
5648 metadata buffers that alias just-allocated data blocks.
5652 Frequency : filesystems with default stripe_count larger than 77
5653 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5654 Details : lconf specifies an inode size of 4096 bytes when the default
5655 stripe_count is larger than 77. This conflicts with the default
5656 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5657 this case to avoid pinning too much memory for large EAs.
5659 ------------------------------------------------------------------------------
5661 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5666 Frequency : rare (extremely heavy IO load with hundreds of clients)
5668 Description: Client is evicted, gets IO error writing to file
5669 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5670 caused clients to be evicted for AST timeouts. The fixes in
5671 bug 5192 mean we no longer need to have such short AST timeouts
5672 so ldlm_timeout has been increased.
5675 Frequency : occasional during --force or --failover shutdown under load
5676 Bugzilla : 5949, 4834
5677 Description: Server oops/LBUG if stopped with --force or --failover under load
5678 Details : a collection of import/export refcount and cleanup ordering
5679 issues fixed for safer force cleanup
5682 Frequency : only filesystems larger than 120 OSTs
5683 Bugzilla : 5990, 6223
5684 Description: lfs getstripe would oops on a very large filesystem
5685 Details : lov_getconfig used kfree on vmalloc'd memory
5688 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5689 Bugzilla : 6242, 6243
5690 Description: reading from files that had been truncated to a non-zero size
5691 but never opened returned no data
5692 Details : ll_file_read() reads zeros from no-object files to EOF
5697 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5698 Details : There's little harm in aborting MDS/OSS recovery and letting it
5699 try again, so I removed the LASSERT and return an error instead.
5701 Severity : enhancement
5703 Description: New debugging infrastructure for tracking down data corruption
5704 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5705 (b) cover more of the client-side code path, and (c) try to narrow
5706 down where problems occurred
5710 Bugzilla : 3819, 4364, 4397, 6313
5711 Description: Racing close and eviction MDS could cause assertion in mds_close
5712 Details : It was possible to get multiple mfd references during close and
5713 client eviction, leading to one thread referencing a freed mfd.
5715 Severity: : enhancement
5716 Bugzilla : 3262, 6359
5717 Description: Attempts to reconnect to servers are now more aggressive.
5718 Details : This builds on the enhanced upcall-less recovery that was added
5719 in 1.4.2. When trying to reconnect to servers, clients will
5720 now try each server in the failover group every 10 seconds. By
5721 default, clients would previously try one server every 25 seconds.
5726 Description: After recovery, certain operations trigger a failed
5727 assertion on a client.
5728 Details : Failing over an mds, using lconf -d --failover, while a
5729 client was doing a readdir() call would cause the client to
5730 LBUG after recovery completed and the readdir() was resent.
5732 Severity : enhancement
5734 Description: Default groups are now added by lconf
5735 Details : You can now run lconf --group <servicename> without having to
5736 manually add groups with lmc.
5739 Frequency : occasional
5741 Description: Nodes with an elan id of 0 trigger a failed assertion
5744 Frequency : always when accessing e.g. tty/console device nodes
5746 Description: tty and some other devices nodes cannot be used on lustre
5747 Details : file's private_data field is used by device data and lustre
5748 values in there got lost. New field was added to struct file to
5749 store fs-specific private data.
5752 Frequency : when exporting Lustre via NFS
5754 Description: NFSD failed occasionally when looking up a path component
5755 Details : NFSD is looking up ".." which was broken in ext3 directories
5756 that had grown large enough to become hashed.
5759 Frequency : Clusters with multiple interfaces not on the same subnet
5761 Description: Nodes will repeatedly try to reconnect to an interface which it
5762 cannot reach and report an error to the log.
5763 Details : Extra peer list entries will be created by lconf with some peers
5764 unreachable. lconf now validates the peer before adding it.
5767 Frequency : Only if a default stripe is set on the filesystem root.
5769 Description: Setting a default stripe on the filesystem root prevented the
5770 filesystem from being remounted.
5771 Details : The client was sending extra request flags in the root getattr
5772 request and did not allocate a reply buffer for the dir EA.
5775 Frequency : occasional, higher if lots of files are accessed by one client
5776 Bugzilla : 6159, 6097
5777 Description: Client trips assertion regarding lsm mismatch/magic
5778 Details : While revalidating inodes the VFS looks up inodes with ifind()
5779 and in rare cases can find an inode that is being freed.
5780 The ll_test_inode() code will free the lsm during ifind()
5781 when it finds an existing inode and then the VFS later attaches
5782 this free lsm to a new inode.
5786 Bugzilla : 6422, 7030
5787 Description: MDS deadlock between mkdir and client eviction
5788 Details : Creating a new file via mkdir or mknod (starting a transaction
5789 and getting the ns lock) can deadlock with client eviction
5790 (gets ns lock and trying to finish a synchronous transaction).
5793 Frequency : occasional
5794 Description: While starting a server, the fsfilt_ext3 module could not be
5796 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5797 kernels. Previously, lconf would still use the ext3 name
5798 when trying to load modules. Now, it will correctly use
5799 ext3 on 2.4 and ldiskfs on 2.6.
5801 Severity : enhancement
5802 Description: The default stripe count has been changed to 1
5803 Details : The interpretation of the default stripe count (0, to lfs
5804 or lmc) has been changed to mean striping across a single
5805 OST, rather than all available. For general usage we have
5806 found a stripe count of 1 or 2 works best.
5808 Severity : enhancement
5809 Description: Add support for compiling against Cray portals.
5810 Details : Conditional compiling for some areas that are different
5814 Frequency : occasional
5815 Bugzilla : 6409, 6834
5816 Description: Creating files with an explicit stripe count may lead to
5817 a failed assertion on the MDS
5818 Details : If some OSTs are full or unavailable, creating files may
5819 trigger a failed assertion on the MDS. Now, Lustre will
5820 try to use other servers or return an error to the
5824 Frequency : occasional
5826 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5827 caused lock timeout during readahead (since 1.4.2).
5828 Details : Processes doing readahead might match a lock that hasn't been
5829 granted yet if there are overlapping and conflicting lock
5830 requests. The readahead process waits on ungranted lock
5831 (original lock is CBPENDING), while OST waits for that process
5832 to cancel CBPENDING read lock and eventually evicts client.
5834 Severity : enhancement
5836 Description: Initial enabling of flock support for clients
5837 Details : Implements fcntl advisory locking and file status functions.
5838 This feature is provided as an optional mount flag (default
5839 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5840 locking are implemented yet, and those that are are not guaranteed
5841 to be completely correct in production environments.
5842 mount -t lustre -o [flock|noflock] ...
5845 Frequency : occasional
5847 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5848 assertion in the ext3 JBD (journaling) layer.
5849 Details : The b_committed_data struct is protected by the big kernel lock
5850 in 2.4 kernels, serializing journal_commit_transaction() and
5851 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5852 access to this struct. In 2.6 kernels there is finer grained
5853 locking to improve SMP performance of the JBD layer.
5857 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5862 Description: merge and cleanup kernel patches.
5865 -----------------------------------------------------------------------------
5867 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5869 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5870 versions of Lustre. Please update all servers and clients to
5871 version 1.4.2 or later at the same time. You must also run
5872 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5873 to update the configuration logs.
5875 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5876 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5877 - swab llog records of type '0' so we get proper header size/idx (5861)
5878 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5879 - fix rename of one directory over another leaking an inode (5953)
5880 - avoid SetPageDirty on 2.6 (5981)
5881 - don't re-add just-being-destroyed locks to the waiting list (5653)
5882 - when creating new directories, inherit the parent's custom
5883 striping settings if present parent (3048)
5884 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5885 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5886 - don't hold client locks on temporary worklist from l_lru (5666)
5887 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5888 - automatically evict dead clients (5921)
5889 - Update file size properly in create+truncate+fstat case (6196)
5890 - Do not unhash mountpoint dentries, do not allow removal of
5892 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5893 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5894 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5895 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5896 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5897 - Correctly handle reads of files with no objects (6243)
5898 - lctl recover will also mark a device active if deactivate used (5933)
5900 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5901 - allow --write-conf on an MDS with different nettype than client (5619)
5902 - don't write config llogs to MDS for mounts not from that MDS (5617)
5903 - lconf should create multiple TCP connections from a client (5201)
5904 - init scripts are now turned off by default; run chkconfig --on
5905 lustre and chkconfig --on lustrefs to use them
5906 - upcalls are no longer needed for clients to recover to failover
5908 - add --abort-recovery option to lconf to abort recovery on device
5910 - add support for an arbitrary number of OSTs (3026)
5911 - Quota support protocol changes.
5912 - forward compatibility changes to wire structs (6007)
5913 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5914 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5915 - improved "lustre status" script
5916 - initialize blocksize for non-regular files (6062)
5917 - added --disable-server and --disable-client configure options (5782)
5918 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5919 - Vanilla 2.4.29 support
5920 - increase maximum number of obd devices to 520 (6242)
5921 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5922 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5923 SLES 9 2.6.5 kernels
5924 - limit stripes per file to 160 (the maximum EA size) (6093)
5926 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5929 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5930 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5931 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5932 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5933 - small ext3 extents cleanups and fixes (5733)
5934 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5935 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5936 - avoid needless client->OST connect, fix handle mismatch (5317)
5937 - fix DLM error path that led to out-of-sync client, long delays (5779)
5938 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5939 - fix several locking issues related to i_size (5492,5624,5654,5672)
5940 - don't move pending lock onto export if it is already evicted (5683)
5941 - fix kernel oops when creating .foo in unlinked directory (5548)
5942 - fix deadlock in obdfilter statistics vs. object create (5811)
5943 - use time_{before,after} to avoid timer jiffies wrap (5882)
5944 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5945 - Do not leak request if server was not able to process it (5154)
5946 - If mds_open unable to find parent dir, make that negative lookup(5154)
5947 - don't create new directories with extent-mapping (5909, 5936)
5949 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5950 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5951 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5952 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5953 - support for RHEL4 kernel on i686 (5773)
5954 - provide error messages when incompatible logs are encountered (5898)
5956 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5957 * version 1.4.0.10 (1.4.1 release candidate 1)
5959 - don't keep a lock reference when lock is not granted (4238)
5960 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5961 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5962 - drop import inflight refcount on signal_completed_replay error (5255)
5963 - unlock page after async write error during send (3677)
5964 - handle missing objects in filter_preprw_read properly (5265)
5965 - no transno return for symlink open, don't save no-trasno open (3440)
5966 - don't try to complete elan receive that already failed (4012)
5967 - free RPC server reply state on error (5406)
5968 - clean up thread from ptlrpc_start_thread() on error (5160)
5969 - readahead could read extra page into cache that wasn't ejected (5388)
5970 - prevent races in class_attach/setup/cleanup/detach (5260)
5971 - don't dereference de->d_inode after l_dput of de (5458)
5972 - use "int" for stripe value returned from lock_to_stripe (5544)
5973 - mballoc allocation and error-checking fixes in 2.6 (5504)
5974 - block device patches to fix I/O request sizes in 2.6 (5482)
5975 - look up hostnames for IB nals (5602)
5976 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5977 - don't start multiple acceptors for the same port (5277)
5978 - fix incorrect LASSERT in mds_getattr_name (5635)
5979 - export a proc file for general "ping" checking (5628)
5980 - fix "lfs check" to not block when the MDS is down (5628)
5982 - service request history (4965)
5983 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5984 - create an "evict_client" /proc entry on OSTs, like the MDS has
5985 - fix mount usage message, return errors per mount(8) (5168)
5986 - change grep [] to grep "[]" in tests so they work in more UMLs
5987 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5988 - remove ancient LOV_MAGIC_V0 EA support (5047)
5989 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5990 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5991 - allow readahead allocations to fail when low on memory (5383)
5992 - mmap locking landed again, after considerable improvement (2828)
5993 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5995 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5998 - send OST transaction number in read/write reply to free req (4966)
5999 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6000 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6001 - the watchdog thread now runs as interruptible (5246)
6002 - flock/lockf fixes (but it's still disabled, pending 5135)
6003 - don't use EXT3 constants in llite code (5094)
6004 - memory shortage at startup could cause assertion (5176)
6006 - reorganization of lov code
6007 - single portals codebase
6009 - add extents/mballoc support (5025)
6010 - direct I/O reads in the obdfilter (4048)
6011 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6013 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6016 - send OST transaction number in read/write reply to free req (4966)
6017 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6018 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6019 - the watchdog thread now runs as interruptible (5246)
6020 - handle missing objects in filter_preprw_read properly (5265)
6021 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6022 - drop import inflight refcount on signal_completed_replay error (5255)
6023 - unlock page after async write error during send (3677)
6024 - return original error code on reconstructed replies (3761)
6025 - no transno return for symlink open, don't save no-trasno open (3440)
6027 - add pid to ldlm debugging output (4922)
6028 - bump the watchdog timeouts -- we can't handle 30sec yet
6029 - extra debugging for orphan dentry/inode bug (5259)
6031 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6034 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6035 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6036 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6037 - fix NULL dereference in /proc/sys/portals/routes (4827)
6038 - allow failed mdc_close() operations to be interrupted (4561)
6039 - stop precreate on OST before MDS would time out on it (4778)
6040 - don't send partial-page writes before EOF from client (4410)
6041 - discard client grant for sub-page writes on large-page clients (4520)
6042 - don't free dentries not owned by NFS code, check generation (4806)
6043 - fix lsm leak if mds_create_objects() fails (4801)
6044 - limit debug_daemon file size, always print CERROR messages (4789)
6045 - use transno after validating reply (3892)
6046 - process timed out requests if import state changes (3754)
6047 - update mtime on OST during writes, return in glimpse (4829)
6048 - add mkfsoptions to LDAP (4679)
6049 - use ->max_readahead method instead of zapping global ra (5039)
6050 - don't interrupt __l_wait_event() during strace
6052 - add software watchdogs to catch hung threads quickly (4941)
6053 - make lustrefs init script start after nfs is mounted
6054 - fix CWARN/ERROR duplication (4930)
6055 - return async write errors to application if possible (2248)
6056 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6057 - print NAL number in %x format (4645)
6058 - update barely-supported suse-2.4.21-171 series (4842)
6059 - support for sles 9 %post scripts
6060 - support for building 2.6 kernel-source packages
6061 - support for sles km_* packages
6063 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6066 - ignore -ENOENT errors in osc_destroy (3639)
6067 - notify osc create thread that OSC is being cleaned up (4600)
6068 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6069 - reconstruct ost_handle() like mds_handle() (4657)
6070 - create a new thread to do import eviction to avoid deadlock (3969)
6071 - let lconf resolve symlinked-to devices (4629)
6072 - don't unlink "objects" from directory with default EA (4554)
6073 - hold socknal file ref over connect in case target is down (4394)
6074 - allow more than 32000 subdirectories in a single directory (3244)
6075 - fix blocks count for O_DIRECT writes (3751)
6076 - OST returns ENOSPC from object create when no space left (4539)
6077 - don't send truncate RPC if file size isn't changing (4410)
6078 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6079 - bind to privileged port in socknal and tcpnal (3689)
6081 - rate limit CERROR/CWARN console message to avoid overload (4519)
6082 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6083 - basic mmap support (3918)
6084 - kernel patch series update from b1_4 (4711)
6086 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6089 - avoid crash during MDS cleanup with OST shut down (2775)
6090 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6091 - don't use bad inodes on the MDS (3744)
6092 - dynamic object preallocation to improve recovery speed (4236)
6093 - don't hold spinlock over lock dumping or change debug flags (4401)
6094 - don't zero obd_dev when it is force cleaned (3651)
6095 - print grants to console if they go negative (4431)
6096 - "lctl deactivate" will stop automatic recovery attempts (3406)
6097 - look for existing locks in ldlm_handle_enqueue() (3764)
6098 - don't resolve lock handle twice in recovery avoiding race (4401)
6099 - revalidate should check working dir is a directory (4134)
6101 - don't always mark "slow" obdfilter messages as errors (4418)
6103 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6106 - don't close LustreDB during write_conf until it is done (3860)
6107 - fix typo in lconf for_each_profile (3821)
6108 - allow dumping logs from multiple threads at one time (3820)
6109 - don't allow multiple threads in OSC recovery (3812)
6110 - fix debug_size parameters (3864)
6111 - fix mds_postrecov to initialize import for llog ctxt (3121)
6112 - replace config semaphore with spinlock (3306)
6113 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6114 - don't allow enqueue to complete on a destroyed export (3822)
6115 - down write_lock before checking llog header bitmap (3825)
6116 - recover from lock replay timeout (3764)
6117 - up llog sem before sending rpc (3652)
6118 - reduce ns lock hold times when setting kms (3267)
6119 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6120 - fix NULL deref and obd_dev leak on setup error (3312)
6121 - replace some LBUG about llog ops with error handling (3841)
6122 - don't match INVALID dentries from d_lookup and spin (3784)
6123 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6124 - fix invalid assertion in ptlrpc_set_wait (3880)
6126 - add libwrap support for the TCP acceptor (3996)
6127 - add /proc/sys/portals/routes for non-root route listing (3994)
6128 - allow setting MDS UUID in .xml (2580)
6129 - print the stack of a process that LBUGs (4228)
6131 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6134 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6135 - make sure to unset replay flag from failed open requests (3440)
6136 - if default stripe count is 0, use OST count for inode size (3636)
6137 - update parent mtime/ctime on client for create/unlink (2611)
6138 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6139 - free recovery state on server during a forced cleanup (3571)
6140 - unregister_reply for resent reqs (3063)
6141 - loop back devices mounting and status check on 2.6 (3563)
6142 - fix resource-creation race that can provoke i_size == 0 (3513)
6143 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6144 - more debugging for page-accounting assertion (3746)
6145 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6146 - avoid deadlock after precreation failure (3758)
6147 - fix race and lock order deadlock in orphan handling (3450, 3750)
6148 - add validity checks when grabbing inodes from l_ast_data (3599)
6150 - add /proc/.../recovery_status to obdfilter (3428)
6151 - lightweight CDEBUG infrastructure, debug daemon (3668)
6152 - change default OSC RPC parameters to be better on small clusters
6153 - turn off OST read cache for files smaller than 32MB
6154 - install man pages and include them in rpms (3100)
6155 - add new init script for (un)mounting lustre filesystems (2593)
6156 - run chkconfig in %post for init scripts (3701)
6157 - drop scimac NAL (unmaintained)
6159 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6162 - clean kiobufs before and after use (3485)
6163 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6164 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6165 - change init script to fail more clearly if not run as root (1528)
6166 - allow clients to reconnect during replay (1742)
6167 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6168 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6169 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6170 - fix cond_resched (3554)
6171 - only evict unfinished clients after recovery (3515)
6172 - allow bulk resend, prevent data loss (3570)
6173 - dynamic ptlrpc request buffer allocation (2102)
6174 - don't allow unlinking open directory if it isn't empty (2904)
6175 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6176 - remove extraneous obd dereference causing LASSERT failure (3334)
6177 - don't use get_cycles() when creating temp. files on the mds (3156)
6178 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6179 - handle EEXIST for set-stripe, set proper directory name (3336)
6181 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6182 - fix ksocknal_fmb_callback() error messages (2918)
6184 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6187 - don't copy lvb into (possibly NULL) reply on error (2983)
6188 - don't deref dentry after dput, don't free lvb on error (2922)
6189 - use the kms to determine writeback rpc length (2947)
6190 - increment oti_logcookies when osc is inactive (2948)
6191 - update client's i_blocks count via lvb messages (2543)
6192 - handle intent open/close of special files properly (1557)
6193 - mount MDS with errors=remount-ro, like obdfilter (2009)
6194 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6195 - don't use cancelling-locks' kms values (2947)
6196 - use highest lock extent for kms, not last one (2925)
6197 - don't dereference ERR_PTR() dentry in error handling path (3107)
6198 - fix thread race in portals_debug_dumplog() (3122)
6199 - create lprocfs device entries at setup instead of at attach (1519)
6200 - common AST error handler, don't evict client on completion race (3145)
6201 - zero nameidata in detach_mnt in 2.6 (3118)
6202 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6203 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6204 - initialize RPC timeout timer earlier for 2.6 (3219)
6205 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6206 - print nal/nid for unknown nid (3258)
6207 - additional checks for oscc recovery before doing precreate (3284)
6208 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6209 - don't crash in mdc_close for bad permissions on open (3285)
6210 - zero i_rdev for non-device files (3147)
6211 - clear page->private before handing to FS, better assertion (3119)
6212 - tune the read pipeline (3236)
6213 - fix incorrect decref of invalidated dentry (2350)
6214 - provide read-ahead stats and refine rpc in flight stats (3328)
6215 - don't hold journal transaction open across create RPC (3313)
6216 - update atime on MDS at close time (3265)
6217 - close LDAP connection when recovering to avoid server load (3315)
6218 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6219 - don't leak open file on MDS after open resend (3325)
6220 - serialize filter_precreate and filter_destroy_precreated (3329)
6221 - loop device shouldn't call sync_dev() for nul device (3092)
6222 - clear page cache after eviction (2766)
6223 - resynchronize MDS->OST in background (2824)
6224 - refuse to mount the same filesystem twice on same mountpoint (3394)
6225 - allow llmount to create routes for mounting behind routers (3320)
6226 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6227 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6228 - fix rare problem with rename on htree directories (3417)
6230 - allow default OST striping configuration per directory (1414)
6231 - fix compilation for qswnal for 2.6 kernels (3125)
6232 - increase maximum number of MDS request buffers for large systems
6233 - change liblustreapi to be useful for external progs like lfsck (3098)
6234 - increase local configuration timeout for slow disks (3353)
6235 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6237 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6240 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6241 - don't overwrite extent policy data in reply if lock was blocked (2901)
6242 - drop filter export grants atomically with removal from device (2663)
6243 - del obd_self_export from work_list in class_disconnect_exports (2908)
6244 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6245 - swab reply message in mdc_close, other PPC fixes (2464)
6246 - fix destroying of named logs (2325)
6247 - overwrite old logs when running lconf --write_conf (2264)
6248 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6249 - fix race in target_handle_connect (2898)
6250 - mds_reint_create() should take same inode create lock (2926)
6251 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6252 - don't close files for self_export to avoid uninitialized obd (2936)
6253 - allow MDS with the same name as client node (2939)
6254 - hold dentry reference for closed log files for unlink (2325)
6255 - reserve space for all logs during transactions (2059)
6256 - don't evict page beyond end of stripe extent (2925)
6257 - don't oops on a deleted current working directory (2399)
6258 - handle hard links to targets without a parent properly (2517)
6259 - don't dereference NULL lock when racing during eviction (2867)
6260 - don't grow lock extents when lots of conflicting locks (2919)
6262 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6265 - account for cache space usage on clients to avoid data loss (974)
6266 - lfsck support in lustre kernel code (2349)
6267 - reduce journal credits needed for BRW writes (2370)
6268 - orphan handling to avoid losing space on client/server crashes
6269 - ptlrpcd can be blocked, stopping ALL progress (2477)
6270 - use lock value blocks to assist in proper KMS, faster stat (1021)
6271 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6272 - recovery for initial connections (2355)
6273 - fixes for mds_cleanup_orphans (1934)
6274 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6275 - block all file creations until orphan recovery completes (1901)
6276 - client remove rq_connection from request struct (2423)
6277 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6278 - recovery timer race (2670)
6279 - mdc_close recovey bug (2532)
6280 - ptlrpc cleanup bug (2710)
6281 - mds timeout on local locks (2588)
6282 - namespace lock held during RPCs (2431)
6283 - handle interrupted sync write properly (2503)
6284 - don't try to handle a message that hasn't been replied to (2699)
6285 - client assert failure during cleanup after abort recovery (2701)
6286 - leak mdc device after failed mount (2712)
6287 - ptlrpc_check_set allows timedout requests to complete (2714)
6288 - wait for inflight reqs when ptlrpcd finishes (2710)
6289 - make sure unregistered services are removed from the srv_list
6290 - reset bulk XID's when resending them (caught by 1138 test)
6291 - unregister_bulk after timeout
6292 - fix lconf error (2694)
6293 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6294 - readahead locks pages, leaves pending causing memory pressure (2673)
6295 - increase OST request buffers to 4096 on large machines (2729)
6296 - fix up permission of existing directories in simple_mkdir (2661)
6297 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6298 - don't assign transno to errored transactions (2742)
6299 - don't delete objects on OST if given a bogus objid from MDS (2751)
6300 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6301 - if rq_no_resend, then timeout request after recovery (2432)
6302 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6303 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6304 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6305 - MDS recovery shouldn't delete objects that it has given out (2730)
6306 - if enqueue arrives after completion, don't clobber LVB (2819)
6307 - don't unlock pages twice when trigger_group_io returns error (2814)
6308 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6309 - don't write pages to disk if there was an error (1450)
6310 - don't ping imports that have recovery disabled (2676)
6311 - take buffered bytes into account when balancing socknal conn (2817)
6312 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6313 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6314 - remove little-endian swabbing of llog records (1987)
6315 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6316 - retry reposting request buffers if they fail (1191)
6317 - grow extent at grant time to avoid granting a revoked lock (2809)
6318 - lock revoke doesn't evict page if covered by a second lock (2765)
6319 - disable VM readahead to avoid reading outside lock extents (2805)
6321 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6322 - updated kernel patches for hp-2.4.20 kernel (2681)
6324 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6327 - fix truncated write corruption (2366)
6328 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6330 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6331 - ensure len(uuid) < 37 in lmc (1171)
6332 - fix ia64 OOPS in llog_test (2255)
6333 - zero end of page at obdfilter for partial page writes (2648)
6334 - don't leave stale dentries around after renames (bug 2428)
6335 - fix timeouts when evicting a client with a single lock held (2642)
6336 - set deadline for the initial HELLO message to drain (2634)
6337 - print out dotted-quad IP addresses in the socknal (2302)
6339 - additional debugging for MDS client eviction problem (2443)
6340 - fix mkfsoptions support for osts (2603, 2604)
6342 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6345 - add series for the vanilla 2.6.0 kernel
6346 - add series for the vanilla 2.4.24 kernel
6347 - add series for a cray x86/64 UL kernel drop
6348 - fix xattr patches for the vanilla 2.4.19 series
6350 - generate true UUIDs in lmc (1171)
6351 - have portals stack dumping break in UML (2466)
6352 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6353 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6354 - wake callback waiting threads on client eviction (2460)
6355 - Add --ptldebug and --subsystem to lmc (1719)
6356 - update assertion to allow safe interrupt allocation
6357 - set rq_no_resend for cancel requests (2432)
6358 - recalculate ptlrpcd timeout after resend (2494)
6359 - call vfs_rmdir when removing pending directories (2368)
6360 - fix renaming a file to itself (2429)
6361 - lmc creates a default one-stripe lov (2454)
6362 - expand procfs space to handle large clusters (2326)
6363 - increase UML stack to avoid overflow
6364 - update lconf's list of debug and subsystem masks
6365 - fix lfs find --obd (2510)
6366 - /proc tunable for disabling filter read caching (2591)
6367 - stop rpm packages from altering slapd.conf (2301)
6368 - disable nagle in the socknal under 0conf (2578)
6369 - choose mds inode size based on stripe count (2572)
6370 - fix kernel-source rpm problems (2516)
6372 - add --disable-doc to avoid pdf generation (2421)
6373 - update documentation, tests, type-os, comments
6374 - avoid format warnings on ia64
6375 - remove the TOE NAL
6376 - tiny code cleanups by removing unused fields
6378 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6381 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6382 - remove the most blatant lies from BUILDING file (2371)
6383 - change default debug level to reasonable production setting
6384 - reduce client side cache size to reduce cache flush time
6385 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6386 - make TCP zerocopy and pinger support enabled by default (2476)
6387 - sync writes completed after process exits caused crashes (2319)
6388 - maintain correct mount count on the MDS (2356)
6389 - backout 1557, because 2316 wasn't really fixed
6390 - better file I/O statistics gathering in /proc
6391 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6392 - another kernel patch to fix zero-copy TCP function export
6393 - don't take duplicate lock when processing re-sent getattr (2420)
6394 - lctl uses obd_self_export instead of creating new conn (2353)
6395 - MDS/OST recovery case which requires object creation asserted (2425)
6396 - move lfs from /usr/sbin to /usr/bin in packages
6397 - fix race between mds_client_add and mds_client_free (2417)
6398 - use kmalloc instead of slabs in portals (2430)
6399 - don't create duplicate records when a failover MDS is present (2442)
6400 - remove unnecessary mount age check (2332)
6401 - don't remove directory inodes from locks prematurely (2451)
6402 - don't break if MDS service name is the same as hostname (2103)
6403 - fix races in client write RPC generation when cache full (2482)
6405 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6408 - remove now-unused request->rq_obd (278)
6409 - if an allocation fails, print out how much memory we've used (1933)
6410 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6411 - add the "configurable stack size" patch to most series files (1256)
6412 - ability to write large log records, for 100+ OST configs (2306)
6413 - fix NULL deref when filter_prep fails (2314)
6414 - fix operator precedence error in filter_sync
6415 - dynamic allocation of socknal TX descriptors (2315)
6416 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6417 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6418 - ensure that CWARN messages in Portals always get to the syslog
6419 - __init/__exit are not for prototype decls (ldlm_init/exit)
6420 - x86-64 compile warning fixes
6421 - fix gateway LMC keyword conflict (2318)
6422 - fix MDS lock inversions in getattr/reint paths (1844)
6423 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6424 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6425 - DLM race condition prevented some lock evictions (2328)
6426 - ENOMEM detection and retry on socknal sends (2230)
6427 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6428 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6429 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6430 - better error messages when a client is rejected during recovery (1505)
6431 - avoid cancelling locks which were never granted, after failure (2330)
6432 - fix i_sem/journal inversion in mds_client_add (2333)
6433 - fix truncate/getattr lock cycle deadlock (2334)
6434 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6435 - fix two rare exit paths which could leak an l_lock() ref (2321)
6436 - fencepost error in MDS/OST orphan recovery (2226)
6437 - make log record alignment 8 bytes (1988)
6438 - lstripe now fails when requested offset > ost_count (2237)
6439 - ensure that all kernel series have a complete list.h (1607)
6440 - fix crashes in special-file operations (2316)
6441 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6443 - allow configurable automake binary, for testing new versions
6444 - small update to the lfs documentation
6446 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6448 * fix negative export reference count in fsfilt_sync (2312)
6450 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6451 * release candidate 0.9.1
6453 - orphans are moved into the PENDING directory for possible recovery
6454 - replayed opens now open by fid for orphan/rename safety (1042)
6455 - last close of an orphan inode generates a transno (683)
6456 - chdir() and mount() now pin the directory entry (1020)
6457 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6458 - discard very old requests without processing them (1502)
6459 - remove some common, well-understood CERRORs (1505)
6460 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6461 - clear "grant" flags in OST replies until OST grant code lands (1644)
6462 - fix read performance by not clobbering i_blksize on client (1598)
6463 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6464 - make LRU size a /proc tunable, clears locks when reduced (707)
6465 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6466 - server locks take references on exports now (1558)
6467 - build fixes for 2.4.20-rh trees (1663)
6468 - return an error from lov_create if all OSCs are inactive (1751)
6469 - fix import levels when a reconnect happens without a timeout (1597)
6470 - exit early from mds_open if we get a lookup error (1749)
6471 - partial page read at EOF wouldn't wait for disk before sending (1642)
6472 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6473 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6474 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6475 - write/truncate lock inversion (1639)
6476 - Don't auto-load obdclass, portals modules during cleanup (1495)
6477 - fix timestamps from jumping to "now" (1763)
6478 - extra journal assertions (1648)
6479 - add an extra multiunlink test (1771)
6480 - fix read_record/write_record API (1776)
6481 - fix leak of offset_extent, possible incorrect i_size later (1772)
6482 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6483 - Debugging for the kqswnal_get_idle_tx problems (1820)
6484 - Allow recovery to be attempted multiple times (1536)
6485 - Write out MDS last_rcvd file after it is first created (1600)
6486 - Fix tx_descriptor leak in failed transmit situations (1827)
6487 - ext3 journaling fixes for assertion failure after IO error (1871)
6488 - class_export_put() on freed export after completion AST error (1896)
6489 - Fix revalidate looping in VFS (1322)
6490 - Don't access a freed export during MDS_REINT timeout (1521)
6491 - Add open-unlink recovery support on the MDS (1673,1764)
6492 - Return an error if no MDS data was read from last_rcvd (1946)
6493 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6494 - Don't setup a disk device that doesn't match exported UUID (317)
6495 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6496 - avoid committing NULL handle in force close
6497 - local.sh is now a one-stripe LOV configuration
6498 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6499 - fix ext3 htree duplicate directory entry corruption (1516)
6500 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6501 - update to new LOV EA format (2097)
6502 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6504 2003-06-15 Phil Schwan <phil@clusterfs.com>
6507 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6508 - per-import/export recovery handling (958, 931, 959)
6509 - multiple last-rcvd slots, for serving multiple FSes (949)
6510 - connections are again shared between multiple imp/exports (963, 964)
6511 - "umount -f" would hang if any requests needed to be sent (393, 978)
6512 - avoid pinning large req buffer by copying for queued messages (989)
6513 - add "uuid" to "lctl device" command to help upcalls (991)
6514 - "open" RPCs with transnos would confuse recovery counters (1037)
6515 - do proper endian conversion of all wire messages (288, 340, 891)
6516 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6517 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6518 - fix ptlrpc_connection leak in target_handle_connect (1174)
6519 - fix import refcounting bug in OST and MDS cleanup (1134)
6520 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6521 - fix very unlikely obd_types race condition (501)
6522 - remove osc_open hack for echo_client (1187)
6523 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6524 - a failure in read_inode2 leads to deadlock (1139)
6525 - cancel ack-locks as soon as transaction is committed (1072)
6526 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6527 - make sure to commitrw after any preprw to avoid deadlock (1162)
6528 - failing to execute a file in a lustre FS would lock inode (1203)
6529 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6530 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6531 - obd_uuid2tgt would walk off the end of the list (1255)
6532 - on IA64 the osc would give portals incorrect bulk size (1258)
6533 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6534 - fix lock inversion caused by new llite matching code (1282)
6535 - limit the number of dirty pages on a client to 10MB (1286)
6536 - timed out locks were not being corrected cancelled (1289)
6537 - fix O_DIRECT above 4GB on IA-32 (1292)
6538 * major user-visible changes
6539 - fail out/fail over policy now controlled by the upcall (993)
6541 - add OBD_PING to check server availability and failure (954)
6542 - lustre messages are now sent in sending host order (288, 340, 891)
6543 - add eadatalen to MDS getattr reply (340)
6544 - OST read replies may contain second buffer, with per-page status (593)
6546 2003-03-11 Phil Schwan <phil@clusterfs.com>
6549 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6550 - failed open()s could cause deadlock; fixed (867, 869)
6551 - stop cancelling OST locks when files are closed (481)
6552 - overlapping XID spaces caused network corruption (851, 853)
6553 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6554 - setattr_raw added, to do single-RPC, server-side setattrs
6555 - lmc/lconf syntax change for OST UUIDs
6556 - fix crashy race condition between ptlrpc_free_req and osc_close
6557 - don't use request in mdc_enqueue if we hit a timeout (889)
6558 - don't set the inode i_size for regular files from the MDS (896)
6559 - handle out of order completion AST (842)
6560 - don't LBUG if a lock request times out after receiving AST (913)
6561 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6562 - if a bad lock AST arrives, send an error instead of dropping entirely
6563 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6564 - fix leak in bulk IO when only partially completed (899, 900, 926)
6565 - fix O_DIRECT for ia64 (55)
6566 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6567 - C-z after timeout could hang a process forever; fixed (977)
6569 - client-side I/O cache (678, 924, 929, 941, 970)
6571 - READPAGE and SETATTRs which don't take server-side locks get
6574 2003-02-11 Phil Schwan <phil@clusterfs.com>
6577 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6578 - Fix fsfilt_extN_readpage() to read a full page of directory
6579 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6580 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6581 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6582 - Most intent operations were removed, in favour of a new RPC mode
6583 that does a single RPC to the server and bypasses most of the VFS
6584 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6585 - Aggressively cancel local locks on DLM servers
6586 - mds_reint_unlink sends EA to the client if it's the last nlink.
6587 client uses that EA to unlink OST objects.
6588 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6589 - recursive symlinks were fixed (439)
6590 - fixed NULL deref in DEBUG_REQ
6591 - filter_update_lastobjid no longer calls sync, which annoyed extN
6592 - fixed multi-client small-writes to a single file problem (445)
6593 - fixed mtime updates during file writes (607)
6594 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6595 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6596 - fixed Linux OST size reporting problem (444, 656)
6597 - OST now updates object mtime with writes or setattr (607, 619)
6598 - client verifies file size before zeroing page past EOF (445)
6599 - OST now writes last allocated objid to disk with allocation (108)
6600 - LOV on echo now works (409)
6602 - mds_reint_unlink sends a new buffer, with the EA included. this
6603 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6604 set if a regular file was being unlinked, and it was the last link
6605 - use PtlGet from the target for bulk writes (315)
6606 - OST now updates object mtime with writes or setattr (607, 619)
6607 - LDLM now has a grant-time callback to revalidate locked items, if
6609 - Many MDS operations were reorganized to combat race conditions
6611 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6612 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6614 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6617 - Fully reactivate OST imports after reconnection (512, others)
6618 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6619 - More graceful error handling for truncating on dead OST (515)
6620 - Don't error out unless we're actually accessing dead stripes (474)
6621 - Fix garbage sizes when stripes are missing (410)
6622 - LRU counters were broken, causing constant lock purge (433, 432)
6623 - garbage on read from stripes with failed OSTs (441)
6624 - mark OSCs as active before reconnecting during recovery (438)
6625 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6626 - lfind did not preserve OST order in output (443)
6627 - symlinks cause hung clients, incorrect data (439)
6628 - stop dereferencing request after dropping refcount (457)
6629 - don't LASSERT(spin_is_locked) on non-SMP (455)
6630 - fixes for many rename() bugs
6631 - fstat didn't correctly synchronize attributes (399)
6632 - server must handle lock cancellation during blocking AST prep (487)
6633 - bulk descriptors were free()d too soon (511)
6634 - fix paths in lconf, which would load incorrect modules (451, 507)
6635 - fix confusing lconf 'host not found' error message (386)
6636 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6637 - fix race condition in mdc_blocking_ast() for inode access (526)
6638 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6639 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6640 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6641 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6642 - fix lstripe in several ways (kernel side) (527)
6643 - fix request leak in ldlm_cli_enqueue (262)
6644 - incorrect OSC was marked inactive after OST failure
6645 - call mds_fs_cleanup before unmounting filesystem (524)
6646 - fix races between taking ns_lock and ldlm_lock_change_resource
6647 - fix races updating LOV export open file list
6648 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6649 - fix recovery NULL deref in ldlm_cli_cancel_unused
6650 - fix some DLM races by using new hash table for lock handles (419)
6651 - permit the client to specify desired inodes, at replay
6652 - duplicate requests when we queue them for replay reintegration
6653 - fix last_rcvd offset calculation
6654 - sync after each recovered transaction, so we always make progress
6655 - never, not always, ERESTART requests without transnos
6656 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6658 - skip replay if the MDS says that the client is already connected
6659 - don't check for a recovery-enabled export to match lctl's UUID
6660 - don't INC_USE_COUNT for phantom exports
6661 - don't crash when cleaning up phantom exports (567)
6662 - don't double-finish or set replay data for errored mdc_open requests
6663 - abort requests when they time out, so we don't get old replies
6664 - send/receive replies for AST messages again
6665 - if the client says that it doesn't have the lock, cancel it on the
6667 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6668 mark it as destroyed, it will all work out when decref is called
6669 - fix module use counts (22, 581)
6671 - ASTs now expect a reply (server cancels lock on error reply)
6673 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6676 - fix many simultaneous client startup (392)
6677 - fix dentry->d_it clobbering
6678 - credentials weren't being shipped for readdir/getattr operations
6679 - remove invalid assertions triggered during some concurrent MD
6681 - proper Lustre versions added (336, 389)
6682 - fix memory leak for create error case (398)
6683 - fix LOV locking bug that would get cli/srv out of sync
6684 - fix echo client over LOV (409)
6685 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6686 - fix double-O_EXCL intent crash (424)
6687 - avoid sending multiple lock CANCELs (352)
6689 - MDS can do multi-client recovery (modulo bugs in new code)
6691 - many updates, edits, cleanups
6693 2002-11-18 Phil Schwan <phil@clusterfs.com>
6696 - fix null d_it dereference (346)
6697 - fix full OST/dbench hang (333)
6698 - fix permission problem with file removal (286)
6699 - fix removal of OSCs from LOV when they fail
6700 - fix NULL deref during bulk timeout (214)
6701 - fix problems related to multiple filesystems on one MDS (241)
6702 - fixed serious subtle metadata locking bugs
6703 - free locks on clients when inodes are removed due to memory
6705 - fix inode pointer in lock data (285)
6706 - partial support for multiple MDS on a single host (241)
6707 - data locks weren't cancelled at clear_inode time (290, 311)
6708 - intent locks could lead to unbounded lock growth (205)
6709 - added a maximum lock count, an LRU list, and a flusher
6710 - fix multiple rename (365)
6711 - properly abstracted the echo client
6712 - OSC locked 1 byte too many; fixed
6713 - rewrote brw callback code:
6714 - fixed recovery bugs related to LOVs (306)
6715 - fixed too-many-pages-in-one-write crash (191)
6716 - fixed (again) crash in sync_io_timeout (214)
6717 - probably fixed callback-related race (385)
6719 - Add capability to MDS protocol
6720 - LDLM cancellations and callbacks on different portals
6722 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6725 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6726 - defer open object destruction to close time (601981, 138)
6727 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6728 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6729 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6730 - don't call obd_setattr in ll_file_release for destroyed objects
6731 * protocol change to lustre_msg: move |version| and add |flags|
6732 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6733 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6734 * added replay of create, unlink, link and rename operations during
6735 MDS failover; recovery should be much more robust now
6736 * remove failed OSCs from LOVs (only lov_create uses this so far)
6737 * the lustre-HOWTO was brought (more) up to date (582544)
6739 2002-10-23 Phil Schwan <phil@clusterfs.com>
6742 - in-use dentries weren't being reused properly (617851)
6743 - prevent multiple LDLM setup (599178)
6744 - fix LOV size calculations for truncate (617853)
6745 - fix client handling of MDS intent errors (POSIX)
6746 - fix permission bug in lovstripe.c test (624321)
6747 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6748 - truncate past end of file could corrupt data
6749 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6750 - a race in recovery could return ETIMEDOUT to apps (623947)
6751 - building outside the source directory was fixed
6752 * the lustre-HOWTO was brought (more) up to date (582544)
6753 * major progress was made on recovery functionality
6755 2002-10-10 Phil Schwan <phil@clusterfs.com>
6758 - recovery deadlock fix
6759 - rm -rf causes LBUG fix (617817)
6760 - file open by multiple tasks fix (618962)
6761 - directory permissions bugs (602707 and 620007)
6762 - journal_stop fixed with locking (611313)
6763 - O_APPEND failures resolved (618273, perhaps 614459)
6764 - lconf PATH fix (619770)
6765 - IA64 build fix (621450)
6766 - RPC buffer sizes scale with amount of memory
6768 2002-10-01 Phil Schwan <phil@clusterfs.com>
6771 - locks would be cancelled without throwing away data pages,
6772 resulting in inconsistent data (605627)
6773 - inode attributes were not always being refreshed (605627, 612449)
6774 - lconf now continues to cleanup after lctl reports an error
6775 - MDS now enforces user permissions (602707)
6776 - lprocfs cleanup fixed, but not yet enabled (614157)
6777 - fixed infinite server hang, should a client not respond to an AST
6778 - avoid going into recovery if user calls readlink() with a buffer
6779 that's too small (613941)
6780 - AST RPCs no longer require replies (614867) -- this may be changed
6781 - don't crash server if client sends an IOV that's too big (611336)
6782 - fixed lock conversion deadlock (611892)
6783 - fixed the following of symlinks (614622)
6784 * recovery: the server can remove locks from a client that dies, other
6785 clients can make progress
6786 * more extN patch fixes
6787 * compile-time configurable ptlrpc buffer allocations
6789 - collaborative read cache document
6790 - Lustre Lite Performance CDR document-in-progress
6792 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6795 - fix typo in patch-2.4.18
6797 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6800 - clear ptlrpc request each time in handle_incoming_request()
6801 - unlink of files now destroys the object on the OST
6803 2002-09-19 Peter Braam <braam@clusterfs.com>
6805 * add hard link support
6806 * change obdfile creation method
6807 * kernel patch changed
6809 2002-09-19 Peter Braam <braam@clusterfs.com>
6812 - stack overflow bug in extN fixed
6814 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6816 * documentation updates
6817 - add man pages for config tools
6818 - update tests/README to describe testing with new config tools
6819 - finish metadata API descriptions
6820 * bug fixes and cleanups
6821 - statfs workaround for 16TB limit
6822 - LOV stripe allocation improved, can stripe on subset of OSTs
6823 - LOV file size/IO offset was wrong for files > 4GB in size
6824 - object EA data was being dropped, caused files to be unreadable
6825 - memory overflow with non-LOV OST caused memory corruption
6826 - fixed regression tests to work with new config tools, obdfilter
6827 - fixed bug when directory size became larger than 1 block
6828 - fixed bug (for single client case) when PWD was deleted
6829 - invalidate local directory pages when doing intent-based ops
6830 - avoid LDLM oops when lock callback contained bad data
6832 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6834 * documentation updates
6835 * bug fixes and cleanups
6836 - configuration tools
6839 - 64-bit compile warnings
6840 - 64-bit internal statfs data
6842 * test_brw on persistent OST devices
6844 * lprocfs (disabled)
6846 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6848 * documentation updates
6849 * bug fixes and cleanups
6850 * configuration tools
6852 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6855 * many small fixes to 0_5_4
6856 * io/network handling
6857 * thinkos in MDS operations
6859 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6862 * crucial basic fixes to 0.5.3
6863 * IOR, Iozone work over Elan
6866 2002-08-07 Phil Schwan <phil@clusterfs.com>
6867 * version 0_5_3, our first alpha
6868 * we use the new Portals iovs
6869 * documentation updates
6870 * bug fixes and cleanups
6871 * small changes in the DLM wire protocol
6873 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6874 * version 0_5_1 with some initial stability,
6875 * locking on MD and file I/O.
6876 * documentation updates
6877 * several bug fixes since 0.5.0
6878 * small changes in wire protocol
6880 2002-07-18 Phil Schwan <phil@clusterfs.com>
6882 * delivered as Lustre Light Alpha
6883 * fixed a crash after handling invalid MDS requests
6884 * fixed directory pages for architectures with non-4k pages sizes
6886 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6887 * release version v0_4_4
6888 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6890 * Fixes a number of interruption problems with OST operations.
6891 * Update documentation for portals header changes
6892 * Move all wire protocol structs/defines to lustre_idl.h
6893 * Fixes symlink length bug.
6894 * Add tcpdump to repository.
6896 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6897 * release version v0_4_3
6898 * Fixes statfs for inodes on extN.
6899 * Fixes bug in runtests which would delete /etc/hosts.
6900 * Use 64-bit object IDs wherever possible (not into VFS though)
6901 Remove ost_get_info, which is unused by lustre, and out of date.
6903 2002-07-03 Peter Braam <braam@clusterfs.com>
6904 * release version v0_4_2 Fixes a lookup error (type not passed)
6905 * move forward to head of Portals
6906 * move forward to latest Lustre kernel
6908 2002-06-25 Peter Braam <braam@clusterfs.com>
6909 * release version v0_4_1. Hopefully stable on single node use.