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 5.14.0-362.24.1.el9 (RHEL9.3)
7 5.14.0-284.30.1.el9 (RHEL9.2)
8 5.14.0-162.23.1.el9 (RHEL9.1)
9 4.18.0-553.el8 (RHEL8.10)
10 4.18.0-513.24.1.el8 (RHEL8.9)
11 4.18.0-477.27.1.el8 (RHEL8.8)
12 * Other server kernels known to build and work at some point (others may also work):
13 3.10.0-862.14.4.el7 (RHEL7.5)
14 3.10.0-957.27.2.el7 (RHEL7.6)
15 3.10.0-1062.18.1.el7 (RHEL7.7)
16 3.10.0-1127.19.1.el7 (RHEL7.8)
17 3.10.0-1160.88.1.el7 (RHEL7.9)
18 4.18.0-80.11.2.el8 (RHEL8.0)
19 4.18.0-147.8.1.el8 (RHEL8.1)
20 4.18.0-193.28.1.el8 (RHEL8.2)
21 4.18.0-240.22.1.el8 (RHEL8.3)
22 4.18.0-305.25.1.el8 (RHEL8.4)
23 4.18.0-348.23.1.el8 (RHEL8.5)
24 4.18.0-372.32.1.el8 (RHEL8.6)
25 4.18.0-425.10.1.el8 (RHEL8.7)
26 4.4.120-92.70 (SLES12 SP2)
27 4.4.180-94.100 (SLES12 SP3)
28 4.4.0-131 (Ubuntu 16.04)
29 4.15.0-32 (Ubuntu 18.04)
30 5.4.0-48 (Ubuntu 20.04)
31 vanilla linux 5.4.0 (ZFS + ldiskfs)
32 vanilla linux 5.4.21 (ZFS + ldiskfs)
33 vanilla linux 5.4.136 (ZFS + ldiskfs)
34 5.10.0-60.94.0.118.oe2203 (openEuler 22.03 LTS)
35 5.10.0-136.32.0.108.oe2203sp1 (openEuler 22.03 LTS SP1)
36 5.10.0-153.19.0.95.oe2203sp2 (openEuler 22.03 LTS SP2)
37 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
38 * Client primary kernels built and tested during release cycle:
39 5.14.0-362.24.1.el9 (RHEL9.3)
40 5.14.0-284.30.1.el9 (RHEL9.2)
41 4.18.0-553.el8 (RHEL8.10)
42 4.18.0-513.24.1.el8 (RHEL8.9)
43 4.18.0-477.27.1.el8 (RHEL8.8)
44 5.4.0-37 (Ubuntu 20.04)
45 5.14.21-150500.55.65 (SLES15 SP5)
46 * Other clients known to build on these kernels at some point (others may also work):
47 3.10.0-862.14.4.el7 (RHEL7.5)
48 3.10.0-957.27.2.el7 (RHEL7.6)
49 4.14.0-49.13.1.el7a (RHEL7.5)
50 4.14.0-115.2.2.el7a (RHEL7.6)
51 3.10.0-1062.18.1.el7 (RHEL7.7)
52 3.10.0-1127.19.1.el7 (RHEL7.8)
53 3.10.0-1160.88.1.el7 (RHEL7.9)
54 4.18.0-80.11.2.el8 (RHEL8.0)
55 4.18.0-147.8.1.el8 (RHEL8.1)
56 4.18.0-193.28.1.el8 (RHEL8.2)
57 4.18.0-240.22.1.el8 (RHEL8.3)
58 4.18.0-305.25.1.el8 (RHEL8.4)
59 4.18.0-348.23.1.el8 (RHEL8.5)
60 4.18.0-372.32.1.el8 (RHEL8.6)
61 4.18.0-425.10.1.el8 (RHEL8.7)
62 5.14.0-70.30.1.el9 (RHEL9.0)
63 5.14.0-162.23.1.el9 (RHEL9.1)
64 4.4.120-92.70 (SLES12 SP2)
65 4.4.180-94.100 (SLES12 SP3)
66 4.12.14-95.48 (SLES12 SP4)
67 4.12.14-122.133 (SLES12 SP5)
68 4.12.14-197.75 (SLES15 SP1)
69 5.3.18-24.96 (SLES15 SP2)
70 5.3.18-150300.59.93 (SLES15 SP3)
71 5.14.21-150400.24.100 (SLES15 SP4)
72 4.4.0-131 (Ubuntu 16.04)
73 4.15.0-48 (Ubuntu 18.04)
74 5.8.0-53 (Ubuntu 20.04.2 HWE)
75 5.11.0-31 (Ubuntu 20.04.3 HWE)
76 5.11.0 (vanilla kernel.org)
77 5.10.0 (openEuler 22.03 LTS)
78 * Recommended e2fsprogs version: 1.47.0-wc6 or newer
79 * Recommended ZFS version: 2.1.15
80 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
81 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
82 stack overflow. For more information, please refer to bugzilla 17630.
83 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
84 the re-exporting nodes to be mounted with "32bitapi" mount option
86 --------------------------------------------------------------------------------
91 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
92 for currently supported client and server kernel versions.
93 * Server primary kernels built and tested during release cycle:
94 4.18.0-348.2.1.el8 (RHEL8.5)
95 * Other server kernels known to build and work at some point (others may also work):
96 3.10.0-862.14.4.el7 (RHEL7.5)
97 3.10.0-957.27.2.el7 (RHEL7.6)
98 3.10.0-1062.18.1.el7 (RHEL7.7)
99 3.10.0-1127.19.1.el7 (RHEL7.8)
100 3.10.0-1160.49.1.el7 (RHEL7.9)
101 4.18.0-80.11.2.el8 (RHEL8.0)
102 4.18.0-147.8.1.el8 (RHEL8.1)
103 4.18.0-193.28.1.el8 (RHEL8.2)
104 4.18.0-240.22.1.el8 (RHEL8.3)
105 4.18.0-305.25.1.el8 (RHEL8.4)
106 4.4.120-92.70 (SLES12 SP2)
107 4.4.180-94.100 (SLES12 SP3)
108 4.4.0-131 (Ubuntu 16.04)
109 4.15.0-32 (Ubuntu 18.04)
110 5.4.0-48 (Ubuntu 20.04)
111 vanilla linux 5.4.0 (ZFS + ldiskfs)
112 vanilla linux 5.4.21 (ZFS + ldiskfs)
113 vanilla linux 5.4.136 (ZFS + ldiskfs)
114 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
115 * Client primary kernels built and tested during release cycle:
116 4.18.0-348.2.1.el8 (RHEL8.5)
117 5.4.0-37 (Ubuntu 20.04)
118 5.3.18-59.27 (SLES15 SP3)
119 * Other clients known to build on these kernels at some point (others may also work):
120 3.10.0-862.14.4.el7 (RHEL7.5)
121 3.10.0-957.27.2.el7 (RHEL7.6)
122 4.14.0-49.13.1.el7a (RHEL7.5)
123 4.14.0-115.2.2.el7a (RHEL7.6)
124 3.10.0-1062.18.1.el7 (RHEL7.7)
125 3.10.0-1127.19.1.el7 (RHEL7.8)
126 3.10.0-1160.49.1.el7 (RHEL7.9)
127 4.18.0-80.11.2.el8 (RHEL8.0)
128 4.18.0-147.8.1.el8 (RHEL8.1)
129 4.18.0-193.28.1.el8 (RHEL8.2)
130 4.18.0-240.22.1.el8 (RHEL8.3)
131 4.18.0-305.25.1.el8 (RHEL8.4)
132 4.4.120-92.70 (SLES12 SP2)
133 4.4.180-94.100 (SLES12 SP3)
134 4.12.14-95.48 (SLES12 SP4)
135 4.12.14-122.91 (SLES12 SP5)
136 4.12.14-197.75 (SLES15 SP1)
137 5.3.18-24.96 (SLES15 SP2)
138 4.4.0-131 (Ubuntu 16.04)
139 4.15.0-48 (Ubuntu 18.04)
140 5.8.0-53 (Ubuntu 20.04.2 HWE)
141 5.11.0-31 (Ubuntu 20.04.3 HWE)
142 5.11.0 (vanilla kernel.org)
143 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
144 * Recommended ZFS version: 2.0.7
145 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
146 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
147 stack overflow. For more information, please refer to bugzilla 17630.
148 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
149 the re-exporting nodes to be mounted with "32bitapi" mount option
151 --------------------------------------------------------------------------------
155 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
156 for currently supported client and server kernel versions.
157 * Server primary kernels built and tested during release cycle:
158 3.10.0-1062.18.1.el7 (RHEL7.7)
159 3.10.0-1127.19.1.el7 (RHEL7.8)
160 3.10.0-1160.6.1.el7 (RHEL7.9)
161 * Other server kernels known to build and work at some point (others may also work):
162 3.10.0-862.14.4.el7 (RHEL7.5)
163 3.10.0-957.27.2.el7 (RHEL7.6)
164 4.18.0-80.11.2.el8 (RHEL8.0)
165 4.18.0-147.8.1.el8 (RHEL8.1)
166 4.18.0-193.28.1.el8 (RHEL8.2)
167 4.18.0-240.1.1.el8 (RHEL8.3)
168 4.4.120-92.70 (SLES12 SP2)
169 4.4.180-94.100 (SLES12 SP3)
170 4.4.0-131 (Ubuntu 16.04)
171 4.15.0-32 (Ubuntu 18.04)
172 5.4.0-48 (Ubuntu 20.04)
173 vanilla linux 5.4.0 (ZFS + ldiskfs)
174 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
175 * Client primary kernels built and tested during release cycle:
176 3.10.0-1062.18.1.el7 (RHEL7.7)
177 3.10.0-1127.19.1.el7 (RHEL7.8)
178 3.10.0-1160.6.1.el7 (RHEL7.9)
179 4.12.14-95.48 (SLES12 SP4)
180 4.12.14-122.20 (SLES12 SP5)
181 4.12.14-197.75 (SLES15 SP1)
182 5.3.18-24.24 (SLES15 SP2)
183 4.15.0-48 (Ubuntu 18.04)
184 * Other clients known to build on these kernels at some point (others may also work):
185 3.10.0-862.14.4.el7 (RHEL7.5)
186 3.10.0-957.27.2.el7 (RHEL7.6)
187 4.14.0-49.13.1.el7a (RHEL7.5)
188 4.14.0-115.2.2.el7a (RHEL7.6)
189 4.18.0-80.11.2.el8 (RHEL8.0)
190 4.18.0-147.8.1.el8 (RHEL8.1)
191 4.18.0-193.28.1.el8 (RHEL8.2)
192 4.18.0-240.1.1.el8 (RHEL8.3)
193 4.4.120-92.70 (SLES12 SP2)
194 4.4.180-94.100 (SLES12 SP3)
195 4.4.0-131 (Ubuntu 16.04)
196 5.4.0-37 (Ubuntu 20.04)
197 5.4.0 (vanilla kernel.org)
198 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
199 * Recommended ZFS version: 2.0.0
200 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
201 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
202 stack overflow. For more information, please refer to bugzilla 17630.
203 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
204 the re-exporting nodes to be mounted with "32bitapi" mount option
206 --------------------------------------------------------------------------------
210 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
211 for currently supported client and server kernel versions.
212 * Server primary kernels built and tested during release cycle:
213 3.10.0-1062.1.1.el7 (RHEL7.7)
214 * Other server kernels known to build and work at some point (others may also work):
215 3.10.0-862.14.4.el7 (RHEL7.5)
216 3.10.0-957.27.2.el7 (RHEL7.6)
217 4.18.0-80.7.1.el8 (RHEL8.0)
218 4.4.120-92.70 (SLES12 SP2)
219 4.4.180-94.100 (SLES12 SP3)
220 4.4.0-131 (Ubuntu 16.04)
221 4.15.0-32 (Ubuntu 18.04)
222 vanilla linux 4.6.7 (ZFS only)
223 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
224 * Client primary unpatched kernels built and tested during release cycle:
225 3.10.0-1062.1.1.el7 (RHEL7.7)
226 4.12.14-95.29 (SLES12 SP4)
227 4.15.0-48 (Ubuntu 18.04)
228 * Other clients known to build on unpatched kernels at some point (others may also work):
229 3.10.0-862.14.4.el7 (RHEL7.5)
230 3.10.0-957.27.2.el7 (RHEL7.6)
231 4.14.0-49.13.1.el7a (RHEL7.5)
232 4.14.0-115.2.2.el7a (RHEL7.6)
233 4.18.0-80.7.1.el8 (RHEL8.0)
234 4.4.120-92.70 (SLES12 SP2)
235 4.4.180-94.100 (SLES12 SP3)
236 4.4.0-131 (Ubuntu 16.04)
237 vanilla linux 4.15.0 (ZFS only)
238 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
239 * Recommended ZFS / SPL version: 0.7.13
240 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
241 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
242 stack overflow. For more information, please refer to bugzilla 17630.
243 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
244 the re-exporting nodes to be mounted with "32bitapi" mount option
245 * Note for Ubuntu 14.04.5:
246 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
247 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
249 --------------------------------------------------------------------------------
253 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
254 for currently supported client and server kernel versions.
255 * Server known to build on the following kernels (others may also work):
256 2.6.32-696.18.7.el6 (RHEL6.9)
257 3.10.0-862.14.4.el7 (RHEL7.5)
258 3.10.0-957.el7 (RHEL7.6)
259 4.4.120-92.70 (SLES12 SP2)
260 4.4.162-94.69 (SLES12 SP3)
261 4.4.0-131 (Ubuntu 16.04)
262 4.15.0-32 (Ubuntu 18.04)
263 vanilla linux 4.6.7 (ZFS only)
264 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
265 * Client known to build on unpatched kernels (others may also work):
266 2.6.32-696.18.7.el6 (RHEL6.9)
267 2.6.32-754.6.3.el6 (RHEL6.10)
268 3.10.0-862.14.4.el7 (RHEL7.5)
269 3.10.0-957.el7 (RHEL7.6)
270 4.14.0-49.13.1.el7a (RHEL7.5)
271 4.14.0-115.2.2.el7a (RHEL7.6)
272 4.4.120-92.70 (SLES12 SP2)
273 4.4.162-94.69 (SLES12 SP3)
274 4.4.0-131 (Ubuntu 16.04)
275 4.15.0-32 (Ubuntu 18.04)
276 vanilla linux 4.15.0 (ZFS only)
277 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
278 * Recommended ZFS / SPL version: 0.7.9
279 * Tested with ZFS / SPL version: 0.7.9
280 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
281 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
282 stack overflow. For more information, please refer to bugzilla 17630.
283 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
284 the re-exporting nodes to be mounted with "32bitapi" mount option
285 * Note for Ubuntu 14.04.5:
286 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
287 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
289 --------------------------------------------------------------------------------
291 2018-04-03 Intel Corporation
293 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
294 for currently supported client and server kernel versions.
295 * Server known to build on patched kernels:
296 2.6.32-431.29.2.el6 (RHEL6.5)
297 2.6.32-504.30.3.el6 (RHEL6.6)
298 2.6.32-573.26.1.el6 (RHEL6.7)
299 2.6.32-642.15.1.el6 (RHEL6.8)
300 2.6.32-696.18.7.el6 (RHEL6.9)
301 3.10.0-693.21.1.el7 (RHEL7.4)
302 3.0.101-0.47.71 (SLES11 SP3)
303 3.0.101-107 (SLES11 SP4)
304 3.12.74-60.64.40 (SLES12 SP1)
305 4.4.114-92.64 (SLES12 SP2)
306 4.4.114-94.11 (SLES12 SP3)
307 3.13.0-101 (Ubuntu 14.04, ZFS only)
308 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
309 4.4.0-116 (Ubuntu 16.04)
310 vanilla linux 4.6.7 (ZFS only)
311 * Client known to build on unpatched kernels:
312 2.6.32-431.29.2.el6 (RHEL6.5)
313 2.6.32-504.30.3.el6 (RHEL6.6)
314 2.6.32-573.26.1.el6 (RHEL6.7)
315 2.6.32-642.15.1.el6 (RHEL6.8)
316 2.6.32-696.18.7.el6 (RHEL6.9)
317 3.10.0-693.21.1.el7 (RHEL7.4)
318 3.0.101-0.47.71 (SLES11 SP3)
319 3.0.101-107 (SLES11 SP4)
320 3.12.74-60.64.40 (SLES12 SP1)
321 4.4.114-92.64 (SLES12 SP2)
322 4.4.114-94.11 (SLES12 SP3)
323 3.13.0-101 (Ubuntu 14.04)
324 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
325 4.4.0-116 (Ubuntu 16.04)
327 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
328 * Recommended ZFS / SPL version: 0.7.6
329 * Tested with ZFS / SPL version: 0.7.6
330 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
331 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
332 stack overflow. For more information, please refer to bugzilla 17630.
333 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
334 the re-exporting nodes to be mounted with "32bitapi" mount option
335 * Note for Ubuntu 14.04.5:
336 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
337 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
339 --------------------------------------------------------------------------------
341 2017-07-05 Intel Corporation
343 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
344 for currently supported client and server kernel versions.
345 * Server known to build on patched kernels:
346 2.6.32-431.29.2.el6 (RHEL6.5)
347 2.6.32-504.30.3.el6 (RHEL6.6)
348 2.6.32-573.26.1.el6 (RHEL6.7)
349 2.6.32-642.15.1.el6 (RHEL6.8)
350 2.6.32-696.3.1.el6 (RHEL6.9)
351 3.10.0-514.21.1.el7 (RHEL7.3)
352 3.0.101-0.47.71 (SLES11 SP3)
353 3.0.101-100 (SLES11 SP4)
354 3.12.74-60.64.40 (SLES12 SP1)
355 4.4.59-92.17 (SLES12 SP2)
356 vanilla linux 4.6.7 (ZFS only)
357 * Client known to build on unpatched kernels:
358 2.6.32-431.29.2.el6 (RHEL6.5)
359 2.6.32-504.30.3.el6 (RHEL6.6)
360 2.6.32-573.26.1.el6 (RHEL6.7)
361 2.6.32-642.15.1.el6 (RHEL6.8)
362 2.6.32-696.3.1.el6 (RHEL6.9)
363 3.10.0-514.21.1.el7 (RHEL7.3)
364 3.0.101-0.47.71 (SLES11 SP3)
365 3.0.101-100 (SLES11 SP4)
366 3.12.74-60.64.40 (SLES12 SP1)
367 4.4.59-92.17 (SLES12 SP2)
369 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
370 * Recommended ZFS / SPL version: 0.6.5.9
371 * Tested with ZFS / SPL version: 0.6.5.9
372 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
373 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
374 stack overflow. For more information, please refer to bugzilla 17630.
375 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
376 the re-exporting nodes to be mounted with "32bitapi" mount option
378 --------------------------------------------------------------------------------
380 12-07-2016 Intel Corporation
382 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
383 for currently supported client and server kernel versions.
384 * Server known to build on patched 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)
393 vanilla linux 4.5.7 (ZFS only)
394 * Client known to build on unpatched kernels:
395 2.6.32-431.29.2.el6 (RHEL6.5)
396 2.6.32-504.30.3.el6 (RHEL6.6)
397 2.6.32-573.26.1.el6 (RHEL6.7)
398 2.6.32-642.11.1.el6 (RHEL6.8)
399 3.10.0-514.el7 (RHEL7.3)
400 3.0.101-0.47.71 (SLES11 SP3)
401 3.0.101-80 (SLES11 SP4)
402 3.12.62-60.64.8 (SLES12 SP1)
404 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
405 * Recommended ZFS / SPL version: 0.6.5.8
406 * Tested with ZFS / SPL version: 0.6.5.8
407 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
408 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
409 stack overflow. For more information, please refer to bugzilla 17630.
410 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
411 the re-exporting nodes to be mounted with "32bitapi" mount option
413 --------------------------------------------------------------------------------
415 02-29-2016 Intel Corporation
417 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
418 for currently supported client and server kernel versions.
419 * Server known to build on patched kernels:
420 2.6.32-431.29.2.el6 (RHEL6.5)
421 2.6.32-504.30.3.el6 (RHEL6.6)
422 2.6.32-573.12.1.el6 (RHEL6.7)
423 3.10.0-327.3.1.el7 (RHEL7.2)
424 3.0.101-0.47.71 (SLES11 SP3)
425 3.0.101-68 (SLES11 SP4)
426 vanilla linux 4.2.1 (ZFS only)
427 * Client known to build on unpatched kernels:
428 2.6.32-431.29.2.el6 (RHEL6.5)
429 2.6.32-504.30.3.el6 (RHEL6.6)
430 2.6.32-573.12.1.el6 (RHEL6.7)
431 3.10.0-327.3.1.el7 (RHEL7.2)
432 3.0.101-0.47.71 (SLES11 SP3)
433 3.0.101-68 (SLES11 SP4)
436 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
437 * Recommended ZFS / SPL version: 0.6.4.2
438 * Tested with ZFS / SPL version: 0.6.4.2
439 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
440 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
441 stack overflow. For more information, please refer to bugzilla 17630.
442 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
443 the re-exporting nodes to be mounted with "32bitapi" mount option
445 --------------------------------------------------------------------------------
447 03-10-2015 Intel Corporation
449 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
450 for currently supported client and server kernel versions.
451 * Server known to build on patched kernels:
452 2.6.32-431.29.2.el6 (RHEL6.5)
453 2.6.32-504.8.1.el6 (RHEL6.6)
454 3.0.101-0.46 (SLES11 SP3)
455 * Client known to build on unpatched kernels:
456 2.6.32-431.29.2.el6 (RHEL6.5)
457 2.6.32-504.8.1.el6 (RHEL6.6)
458 3.10.0-123.20.1.el7 (RHEL7)
459 3.0.101-0.46 (SLES11 SP3)
460 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
461 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
462 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
463 stack overflow. For more information, please refer to bugzilla 17630.
464 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
465 the re-exporting nodes to be mounted with "32bitapi" mount option
467 Severity : enhancement
469 Description: control OST-index in IDIF via ROCOMPAT flag.
470 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
471 last_rcvd file. For new formatted OST device, it will be auto set;
472 for the case of upgrading from old OST device, you can enable it
473 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
474 enabled, for new created OST-object, its IDIF-in-LMA will contain
475 the OST-index; for the existing OST-object, the OSD will convert
476 old format IDIF as new format IDIF with OST-index stored in the
477 LMA EA when accessing such OST-object or via OI scrub. Once such
478 flag is enabled, it cannot be reverted back, so the system cannot
479 be downgraded to the orignal incompatible version.
481 --------------------------------------------------------------------------------
483 07-30-2014 Intel Corporation
485 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
486 for currently supported client and server kernel versions.
487 * Server known to build on patched kernels:
488 2.6.32-431.20.3.el6 (RHEL6.5)
489 3.0.101-0.31 (SLES11 SP3)
490 * Client known to build on unpatched kernels:
491 2.6.32-431.20.3.el6 (RHEL6.5)
492 3.0.101-0.31 (SLES11 SP3)
495 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
496 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
497 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
498 stack overflow. For more information, please refer to bugzilla 17630.
499 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
500 the re-exporting nodes to be mounted with "32bitapi" mount option
502 --------------------------------------------------------------------------------
504 10-31-2013 Intel Corporation
506 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
507 * Server builds for kernels:
508 2.6.32-358.18.1.el6 (RHEL6.4)
509 * Client builds for unpatched kernels:
510 2.6.32-358.18.1.el6 (RHEL6.4)
511 3.0.92-0.8 (SLES11 SP3)
513 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
514 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
515 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
516 stack overflow. For more information, please refer to bugzilla 17630.
517 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
518 the re-exporting nodes to be mounted with "32bitapi" mount option
520 --------------------------------------------------------------------------------
522 05-30-2013 Intel Corporation
524 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
525 * Server builds for kernels:
526 2.6.32-358.6.2.el6 (RHEL6.4)
527 * Client builds for unpatched kernels:
528 2.6.32-358.6.2.el6 (RHEL6.4)
529 3.0.74-0.6.6 (SLES11 SP2)
531 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
532 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
533 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
534 stack overflow. For more information, please refer to bugzilla 17630.
535 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
536 the re-exporting nodes to be mounted with "32bitapi" mount option
538 --------------------------------------------------------------------------------
540 9-30-2012 Whamcloud, Inc.
542 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
543 * Server support for kernels:
544 2.6.32-279.5.1.el6 (RHEL6)
545 * Client support for unpatched kernels:
546 2.6.32-279.5.1.el6 (RHEL6)
548 2.6.18-238.19.1.el5 (RHEL5)
549 2.6.16.60-0.69.1 (SLES10)
550 * Recommended e2fsprogs version: 1.41.90.wc4
551 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
552 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
553 stack overflow. For more information, please refer to bugzilla 17630.
554 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
555 the re-exporting nodes to be mounted with "32bitapi" mount option
557 Severity : enhancement
559 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
561 Severity : enhancement
563 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
565 Severity : enhancement
567 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
569 Severity : enhancement
571 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
573 Severity : enhancement
575 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
577 --------------------------------------------------------------------------------
579 04-30-2012 Whamcloud, Inc.
581 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
582 * Server support for kernels:
583 2.6.32-220.4.2.el6 (RHEL6)
584 * Client support for unpatched kernels:
585 2.6.32-220.4.2.el6 (RHEL6)
587 2.6.18-238.19.1.el5 (RHEL5)
588 2.6.16.60-0.69.1 (SLES10)
589 * Recommended e2fsprogs version: 1.41.90.wc4
590 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
591 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
592 stack overflow. For more information, please refer to bugzilla 17630.
593 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
594 the re-exporting nodes to be mounted with "32bitapi" mount option
596 Severity : enhancement
598 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
600 Severity : enhancement
602 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
604 Severity : enhancement
606 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
608 Severity : enhancement
610 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
612 --------------------------------------------------------------------------------
614 09-30-2011 Whamcloud, Inc.
616 * Server support for kernels:
617 2.6.18-238.19.1.el5 (RHEL5)
618 2.6.32-131.6.1.el6 (RHEL6)
619 * Client support for unpatched kernels:
620 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
621 2.6.18-238.19.1.el5 (RHEL5)
622 2.6.32-131.6.1.el6 (RHEL6)
623 2.6.29.4-167.fc11 (FC11)
624 2.6.16.60-0.69.1 (SLES10)
626 * Recommended e2fsprogs version: 1.41.90.wc3
627 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
628 Lustre filesystem with 4K stack may cause a stack overflow. For more
629 information, please refer to bugzilla 17630.
630 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
631 mounted with "32bitapi" mount option
633 Severity : enhancement
634 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
636 Severity : enhancement
637 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
639 Severity : enhancement
640 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
643 Frequency : only with SLES10
645 Description: Use OFED "KMP" provided by Novell
646 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
647 Lustre is now built against this package. That means you need to
648 install the ofed-kmp package from Novell for the patchless client.
649 Note that the ofed-kmp that Novell ships may not exactly match the
650 kernel version but should still be compatible.
652 Severity : enhancement
654 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
655 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
657 --------------------------------------------------------------------------------
659 2010-07-15 Oracle, Inc.
661 * Server support for kernels:
662 2.6.18-164.11.1.el5 (RHEL 5)
663 2.6.18-164.11.1.0.1.el5 (OEL 5)
664 * Client support for unpatched kernels:
665 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
666 2.6.18-164.11.1.el5 (RHEL 5),
667 2.6.18-164.11.1.0.1.el5 (OEL 5)
668 2.6.16.60-0.42.8 (SLES 10),
670 2.6.29.4-167.fc11 (FC11)
671 * Recommended e2fsprogs version: 1.41.10-sun2
673 Severity : enhancement
675 Description: Distribute lfs_migrate script with lustre
676 Details : The lfs_migrate mentioned in the manual is now included in the
677 lustre tree in lustre/scripts and will distributed in lustre RPMs.
679 Severity : enhancement
681 Description: Failover nid list added to import proc file.
683 Severity : enhancement
685 Description: "som_preview" mount option required to use SOM
686 Details : The clients will need to use som_preview mount option
690 Description: Lustre does not do 1MB IOs to HW RAID
691 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
693 Severity : enhancement
695 Description: lfs quota output is not very convenient for awk/sed-parsing
696 Details : Some positions in lfs quota output table could be empty or
697 non-empty which made it hard to parse it with scripts, now
698 a dash is put instead of space where there is not supposed
703 Description: Update to OFED 1.5.1
704 Details : OFED 1.5.1 will be used in environments that don't use
705 vendor supplied OFED.
709 Description: Remove lnet. prefix from lctl params display
710 Details : Also renamed lnet.memused to lnet_memused to
715 Description: Use conf_param -d to remove permanant settings.
719 Description: Add list_param -R
720 Details : lctl list_param now supports recursive mode.
723 Bugzilla : 21829, 22312
724 Description: Fix llobdstat.
725 Details : 1. Now updating the write_bytes counter
726 2. Port script fixes from 1.8.
730 Description: decrease the usage of memory on clients.
731 Details : 1. On clients, recycle dentries and inodes unused.
732 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
737 Description: llobdstat fix and enhancement
738 Details : add a counter to set a limit to how many samples will be returned
739 fix a wildcard in the path to limit to obdfilter stats only
743 Description: lfs check only allowed for root
744 Details : Code cleanups and added a regresion test of non-root use of
749 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
750 Details : Fix hr->hr_index overload.
754 Description: Fix FIEMAP functionality.
758 Description: FC11 patchless client support added
762 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
763 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
767 Description: New acc-sm script and various test-framework enhancments.
771 Description: Previously landed patch was reverted.
775 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
779 Description: Add ldiskfs maxdirsize mount option.
783 Description: Fix mkfs.lustre support for 16TB LUNs.
787 Description: Print mmp_check_interval
788 Details : print mmp_check_interval and make it possible to abort
789 mount operation in case it takes too long.
793 Description: Add cascading_rw to lustre mpi tests.
797 Description: Store SOM data in EA.
798 Details : Also several SOM related cleanups.
802 Description: Add refcount to md_open_data
806 Description: Use refcount on lov_request_set to prevent memory corruption.
810 Description: Statahead: use parent dir's i_mutex for synchronization.
814 Description: loadgen improvements.
818 Description: Enable peer health detection in ksocklnd and ko2iblnd.
822 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
826 Description: Fix race on obd_devcie:md_stats
830 Description: Fix SLES 11 client builds.
831 Details : Resolved uninitalized pointer gcc warnings.
835 Description: Allow non-root access for "lfs check".
836 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
840 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
842 Severity : enhancement
844 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
845 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
849 Description: Sanity test_27p fails.
850 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
851 Sanity tests 27 were cleaned up and improved.
855 Description: Unify req format on client/servers
856 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
857 buffers and swabbers approach..
862 Description: drop join file stuff.
864 Severity : enhancement
866 Description: Add 2.6.27 and SLES11 (patchless client) support
870 Description: Fix in ptlrpc_expire_one_request() to print the signed time
873 Severity : enhancement
875 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
878 Severity : enhancement
880 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
882 Severity : enhancement
884 Description: correctly handle big reply message.
885 Details : send LNet event if reply is bigger than buffer and adjust this buffer
890 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
892 Severity : enhancement
894 Description: Add support for OEL5.
898 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
900 Severity : enhancement
902 Description: "weak-modules" support
903 Details : Implement "weak-modules" support which enables kernel modules
904 to be used with any kernel that implements the same kABI. In
905 order to achieve this modules are now installed in
906 /lib/modules/$(uname -r)/updates/kernel on all distributions.
910 Description: client could not reconnect to OST because of an active request.
911 Details : The client now retries to reconnect to the same server, if a
912 connect request failed with EBUSY or -EAGAIN. Test case 35b
913 conf-sanity.sh has been created.
917 Description: Don't allow make backward step on assignin osc next id.
918 Details : race between allocation next id and ll_sync thread can be cause
919 of set wrong osc next id and can be kill valid ost objects.
923 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
927 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
931 Description: Error handling in osc_statfs_interpret() has been improved.
932 Details : Check in osc_statfs_interpret() for EBADR.
936 Description: Do not send statfs() requests to OSTs disabled by administrator.
937 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
939 Severity : enhancement
941 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
943 Severity : enhancement
945 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
947 Severity : enhancement
949 Description: Enhancement for lfs(1) command to use numeric uid/gid.
953 Description: Improved cleanup procedure for conf-sanity test_46a.
957 Description: Do not start bulk transfer, if client tried to reconnect.
958 Details : Do not start bulk transfer, if client tried to reconnect,
959 but failed due to active rpcs. Check for exp_abort_active_req
960 has been implemented.
964 Description: Prevent inconsistences between linux and lustre mount structures.
965 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
966 Make the sleep interruptible.
968 Severity : enhancement
970 Description: provide server to client comms path
971 Details : server to client communications path via new LDLM_SET_INFO rpc
975 Description: truncate starts GFP_FS allocation under transaction causing deadlock
976 Details : ldiskfs_truncate calls grab_cache_page which may start page
977 allocation under an open transaction. This may lead to
978 calling prune_icache with consequent lustre reentrance.
982 Description: send correctly lsm on open replay
983 Details : MDS is trust to LSM size on replay open, but client can set wrong size
988 Description: enable client interface failover
989 Details : When a child reconnects from another NID, properly update export
990 nid hash position and ldlm reverse import.
992 Severity : enhancement
994 Description: Communicate OST degraded/readonly state via statfs to MDS
995 Details : Flags in the statfs returned from OSTs indicate whether the
996 OST is in a degraded RAID state, or if the filesystem has
997 turned read-only after a filesystem error is detected.
1002 Description: don't panic if EPROTO was hit when reading symlink
1003 Details : correctly handling request reference in error cases.
1005 Severity : enhancement
1007 Description: Add LustreNetLink, a kernel-userspace communcation path.
1009 Severity : enhancement
1011 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1016 Description: access to llog context before init.
1017 Details : move handling CATALOGS file at osc layer and forbid access to llog
1018 context before init.
1021 Frequency : rare, only when client is writing to OST on local node
1023 Description: Avoid deadlock for local client writes
1024 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1025 memory freeing context. This allows OST threads to set the
1026 PF_MEMALLOC flag on task structures in order to allocate memory
1027 from reserved pools and complete IO.
1028 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1029 so that the OST threads generate memory pressure and allow
1030 inactive pages to be reclaimed.
1032 Severity : enhancement
1034 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1039 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1040 Details : move ldlm namespace creation in setup phase to avoid grab
1041 _lprocfs_lock with cli_sem held.
1045 Description: Temporarily disable grant shrink.
1046 Details : Disable the feature for debugging.
1049 Frequency : only when formatting test filesystems
1051 Description: Unable to run several mkfs.lustre on loop devices at the same time
1052 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1053 formatting. The solution is to proper handle the error.
1055 Severity : enhancement
1057 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1059 Severity : enhancement
1061 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1066 Description: resolve race between obd_disconnect and class_disconnect_exports
1067 Details : if obd_disconnect will be called to already disconnected export he
1068 forget release one reference and osc module can't unloaded.
1070 Severity : enhancement
1072 Description: Allow tuning service thread via /proc
1073 Details : For each service a new
1074 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1075 created that can be used to set min/max thread counts, and get the
1076 current number of running threads.
1081 Descriptoin: don't return error if have partially created objects for file.
1082 Details : lov_update_create_set uses set->set_success as index for created
1083 objects, so if some requests will be failed, they will have hole
1084 at end of array and qos_shrink_lsm for allocate correct lsm.
1086 Severity : enhancement
1088 Description: Update OFED support to 1.4.1
1093 Description: Reduce small size read RPC
1094 Details : Set read-ahead limit for every file and only do read-ahead when
1095 available read-ahead pages are bigger than 1M to avoid small size
1098 Severity : enhancement
1100 Description: move AT tunable parameters for more consistent usage
1101 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1103 Severity : enhancement
1105 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1106 Details : allow skip disconnected ost for send statfs request and hide error
1112 Description: fix racy locking of mballoc block bitmaps causing BUG
1113 Details : The locking of the mballoc buddy bitmap and the in-memory
1114 block bitmap was using two different spin locks in some
1115 cases. This made it possible to incorrectly access the
1116 mballoc bitmap while another process was modifying it,
1117 causing a sanity assertion to fail. While no on-disk corruption
1118 was reported, there was some risk of this happening.
1121 Frequency : rare, on llog test 6
1123 Descriptoin: don't allow connect to already connected import
1124 Details : allowing connect to already connected import is hide connecting problem.
1129 Descriptoin: allow kill process which wait statahead result
1130 Details : for some reasons 'ls' can stick in waiting result from statahead,
1131 in this case need way for kill this process.
1133 Severity : enhancement
1135 Description: Add state history info file, enhance import info file
1136 Details : Track import connection state changes in a new osc/mdc proc file;
1137 add overview-type data to the osc/mdc import proc file.
1139 Severity : enhancement
1141 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1143 Severity : enhancement
1145 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1150 Descriptoin: OST grant too much space to client even there are not enough space.
1151 Details : Client will shrink its grant cache to OST if there are no write
1152 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1153 this grant cache if there are already not enough avaible space
1154 (left_space < total_clients * 32M).
1157 Frequency : start MDS on uncleanly shutdowned MDS device
1159 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1160 Details : stay in waiting mds<>ost recovery finished produce random bugs
1161 due race between two ll_sync thread for one lov target. send
1162 ACTIVATE event only if connect realy finished and import have
1166 Frequency : rare, connect and disconnect target at same time
1168 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1169 Details : don't call obd_disconnect under lov_lock. this long time
1170 operation and can block ptlrpcd which answer to connect request.
1175 Descriptoin: don't lose wakeup for imp_recovery_waitq
1176 Details : recover_import_no_retry or invalidate_import and import_close can
1177 both sleep on imp_recovery_waitq, but we was send only one wakeup
1181 Frequency : always with long access acl
1183 Descriptoin: mds can't pack reply with long acl.
1184 Details : mds don't control size of acl but they limited by reint/getattr
1187 Severity : enhancement
1189 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1191 Severity : enhancement
1193 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1196 Frequency : start MDS on uncleanly shutdowned MDS device
1198 Descriptoin: aborting recovery hang on MDS
1199 Details : don't throttle destroy RPCs for the MDT.
1202 Frequency : on remount
1204 Description: external journal device not working after the remount
1205 Details : clear dev_rdonly flag for external journal devices in
1211 Description: shutdown vs evict race
1212 Details : client_disconnect_export vs connect request race.
1213 if client will evicted at this time - we start invalidate
1214 thread without referece to import and import can be freed
1218 Frequency : rare, need acl's on inode.
1220 Description: client can't handle ost additional correctly
1221 Details : if ost was added after client connected to mds client can have
1222 hit lnet_try_match_md ... to big messages to wide striped files.
1223 in this case need teach client to handle config events about add
1224 lov target and update client max ea size at that event.
1226 Severity : enhancement
1228 Description: Changelogs
1229 Details : Changelogs are a lightweight mechanism to track filesystem
1230 metadata and namespace changes. The changelog is recorded
1231 permanently on the MDTs, and is periodically "consumed" / purged
1232 when records are no longer needed.
1234 Severity : enhancement
1236 Description: compact fld format with extents
1237 Details : Store range of seq rather than every seq in FLD. Seq
1238 controller update FLD rather than clients. In Case of CMD, mdt0
1239 has FLD, all other metadata server act as non persistent proxy
1240 for FLD queries and cache fld entries in fld cache.
1245 Description: don't skip ost target if they assigned to file
1246 Details : Drop slow OSCs if we can, but not for requested start idx.
1247 This means "if OSC is slow and it is not the requested
1248 start OST, then it can be skipped, otherwise skip it only
1249 if it is inactive/recovering/out-of-space.
1253 Description: more cleanup in mds_lov
1254 Details : not send LOV EA under replay, we can't know about they size at this
1255 time. Don't allow client connect to mds before any ost connected,
1256 for avoid problems with LOV EA size and returning EIO to client.
1258 Severity : enhancement
1260 Description: Interoperability at server side (Disk interoperability)
1262 Severity : enhancement
1264 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1266 Severity : enhancement
1268 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1270 Severity : enhancement
1272 Description: New client IO stack (CLIO).
1274 Severity : enhancement
1276 Description: Commit on sharing. Eliminate inter-client dependencies between
1277 uncommitted transactions by doing transaction commits.
1278 Thereby clients may recovery independently.
1281 Frequency : Create a symlink file with a very long name
1283 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1284 Details : If there is no extra space in the request for early cancels,
1285 ldlm_req_handles_avail() returns 0 instead of a negative value.
1287 Severity : enhancement
1289 Description: Add /proc entry for import status
1290 Details : The mdc, osc, and mgc import directories now have
1291 an import directory that contains useful import data for debugging
1292 connection problems.
1294 Severity : enhancement
1296 Description: Re-disable certain /proc logging
1297 Details : Enable and disable client's offset_stats, extents_stats and
1298 extents_stats_per_process stats logging on the fly.
1301 Frequency : Only on FC kernels 2.6.22+
1303 Description: oops in statahead
1304 Details : Do not drop reference count for the dentry from VFS when lookup,
1305 VFS will do that by itself.
1307 Severity : enhancement
1309 Description: Generic /proc file permissions
1310 Details : Set /Proc file permissions in a more generic way to enable non-
1311 root users operate on some /proc files.
1315 Description: Hitting mdc_commit_close() ASSERTION
1316 Details : Properly handle request reference release in
1317 ll_release_openhandle().
1321 Description: quota recovery deadlock during mds failover
1322 Details : This patch includes att18982, att18236, att18237 in bz14840.
1324 1. fix osts hang when mds does failover with quotaon
1325 2. prevent watchdog storm when osts threads wait for the
1330 Frequency : only patchless client
1331 Description: add workaround for race between add/remove dentry from hash
1333 Severity : enhancement
1335 Description: Allow OST glimpses to return PW locks
1339 Description: LBUG when llog conf file is full
1340 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1344 Description: Prevent import from entering FULL state when server in recovery
1348 Description: service mount cannot take device name with ":"
1349 Details : Only when device name contains ":/" will mount treat it as
1355 Description: replace ptlrpcd with the statahead thread to interpret the async
1356 statahead RPC callback
1360 Frequency : on recovery
1361 Description: I/O failures after umount during fail back
1362 Details : if client reconnected to restarted server we need join to recovery
1363 instead of find server handler is changed and process self
1364 eviction with cancel all locks.
1366 Severity : enhancement
1368 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1370 Severity : enhancement
1372 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1374 Severity : enhancement
1376 Description: Upcall on Lustre log has been dumped
1377 Details : Allow for a user mode script to be called once a Lustre log has
1378 been dumped. It passes the filename of the dumped log to the
1379 script, the location of the script can be specified via
1380 /proc/sys/lnet/debug_log_upcall.
1385 Description: avoid idr_remove called for id which is not allocated.
1386 Details : Move assigment s_dev for clustered nfs to end of initialization,
1387 to avoid problem with error handling.
1392 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1393 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1394 into CONN_USED_HASH and this prodice warning when put connection
1395 again in unused hash.
1398 Severity : enhancement
1400 Description: File striping can now be set to use an arbitrary pool of OSTs.
1402 Severity : enhancement
1404 Description: Export bytes_read/bytes_write count on OSC/OST.
1408 Description: Early reply size mismatch, MGC loses connection
1409 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1410 the connect flags are properly negotiated.
1415 Description: quotas are not honored with O_DIRECT
1416 Details : all writes with the flag O_DIRECT will use grants which leads to
1417 this problem. Now using OBD_BRW_SYNC to guard this.
1421 Description: add quota statistics
1422 Details : 1. sort out quota proc entries and proc code.
1423 2. add quota statistics
1425 Severity : enhancement
1427 Description: enable quota support for HEAD.
1431 Description: Properly propagate oinfo flags from lov to osc for statfs
1432 Details : restore missing copy oi_flags to lov requests.
1434 Severity : enhancement
1436 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1437 : l_getgroups(8), lst(8), routerstat(8)
1438 Details : included man pages for llobdstat(8), llstat(8),
1439 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1441 Severity : enhancement
1443 Description: configure's --enable-quota should check the
1444 : kernel .config for CONFIG_QUOTA
1445 Details : configure is terminated if --enable-quota is passed but
1446 : no quota support is in kernel
1450 Description: Remove portals compatibility
1451 Details : Remove portals compatibility, not interoperable with releases
1456 Description: Resolve device initialization race
1457 Details : Prevent proc handler from accessing devices added to the
1458 obd_devs array but yet be intialized.
1460 Severity : enhancement
1462 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1464 Severity : enhancement
1466 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1471 Description: Using wrong pointer in osc_brw_prep_request
1472 Details : Access to array[-1] can produce panic if kernel compiled with
1473 CONFIG_PAGE_ALLOC enabled
1477 Description: Client runs out of low memory
1478 Details : Consider only lowmem when counting initial number of llap pages
1482 Description: *optional* service tags registration
1483 Details : if the "service tags" package is installed on a Lustre node
1484 When the filesystem is mounted, a local-node service tag will
1485 be created. See http://inventory.sun.com/ for more information
1486 about the Service Tags asset management system.
1490 Description: Kernel BUG tries to release flock
1491 Details : Lustre does not destroy flock lock before last reference goes
1492 away. So always drop flock locks when client is evicted and
1493 perform unlock regardless of successfulness of speaking to MDS.
1497 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1501 Description: sanity test 65a fails if stripecount of -1 is set
1502 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1507 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1508 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1511 Severity : enhancement
1513 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1517 Description: do not process already freed flock
1518 Details : flock can possibly be freed by another thread before it reaches
1519 to ldlm_flock_completion_ast.
1523 Description: LBUG during stress test
1524 Details : Need properly lock accesses the flock deadlock detection list.
1528 Description: oops in page fault handler
1529 Details : kernel page fault handler can return two special 'pages' in error
1530 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1534 Description: timeout with invalidate import.
1535 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1536 should be handled by ptlrpcd. This produce long age waiting and
1537 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1539 Severity : enhancement
1541 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1545 Description: enable MGS and MDT services start separately
1546 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1547 a co-located MGS without starting the MGS, which is a complement
1548 to 'nosvc' mount option.
1552 Frequency : after recovery
1553 Description: precreate to many object's after del orphan.
1554 Details : del orphan st in oscc last_id == next_id and this triger growing
1555 count of precreated objects. Set flag LOW to skip increase count
1556 of precreated objects.
1560 Frequency : rare, on clear nid stats
1561 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1562 Details : when clean nid stats sometimes try destroy live entry,
1563 and this produce panic in free.
1567 Description: Stack overflow during MDS log replay
1568 ease stack pressure by using a thread dealing llog_process.
1572 Description: wait until IO finished before start new when do lock cancel.
1573 Details : VM protocol want old IO finished before start new, in this case
1574 need wait until PG_writeback is cleared until check dirty flag and
1575 call writepages in lock cancel callback.
1577 Severity : enhancement
1579 Description: using special macro for print time and cleanup in includes.
1583 Description: mds_mfd_close() ASSERTION(rc == 0)
1584 Details : In mds_mfd_close(), we need protect inode's writecount change
1585 within its orphan write semaphore to prevent possible races.
1589 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1593 Frequency : rare, on shutdown ost
1594 Description: don't hit live lock with umount ost.
1595 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1596 use shrink_dcache_sb instead.
1600 Description: don't panic with use echo client
1601 Details : echo client pass NULL as client nid pointer and this produce null
1602 pointer dereference.
1606 Description: fix build on ppc32
1607 Details : compile code with -m64 flag produce wrong object file for ppc32.
1611 Description: add message levels for liblustreapi
1615 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1616 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1617 will result in an -EAGAIN and in the exit call path we call
1618 alloc_rr() which will with increasing aggressiveness attempt to
1619 aquire precreated objects on the minimum number of required OSCs.
1623 Description: Use old size assignment to avoid deadlock
1624 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1625 the scheduling while holding a spinlock. We do not need locking
1626 for size in ll_update_inode() because size is only updated from
1627 the MDS for directories or files without objects, so there is no
1628 other place to do the update, and concurrent access to such inodes
1629 are protected by the inode lock.
1633 Description: resolve "_IOWR redefined" build error on SLES10
1637 Description: dump the memory debugging after all modules are unloaded to
1638 suppress false negative in conf_sanity test 39
1640 Severity : enhancement
1642 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1647 Description: fix SLES kernel versioning
1648 Details : the kernel version for our SLES 10 kernel did not include a "-"
1649 before the "smp" at the end. while this was not a problem in
1650 general, it did mean that software trying to use the kernel
1651 version to try to detect a vendor specific kernel would fail.
1652 this was most evident by the OFED build scripts.
1656 Description: Don't update lov_desc members until making sure they are valid
1657 Details : When updating lov_desc members via proc fs, need fix their
1658 validities before doing the real update.
1662 Description: don't put request into delay list while invalidate in flight.
1663 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1664 import in flight. this produce timeout for invalidate and sometimes
1665 can cause stale data.
1669 Frequency : on ppc only
1670 Description: not convert ost objects for directory because it's not exist.
1671 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1675 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1677 Severity : enhancement
1679 Description: Update to support 2.6.22.14 vanilla kernel.
1683 Frequency : rare, on recovery
1684 Description: read procfs can produce deadlock in some situation
1685 Details : Holding lprocfs lock which send rpc can produce block for destroy
1686 obd objects and this also block reconnect with -EALREADY. This
1687 isn't fix all lprocfs bugs - but make it rare.
1689 Severity : enhancement
1691 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1694 Frequency : frequent on X2 node
1696 Description: mdc_set_open_replay_data LBUG
1697 Details : Set replay data for requests that are eligible for replay.
1701 Description: lustre_mgs: operation 101 on unconnected MGS
1702 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1703 MGC, and late on MGC cannot successfully connect to MGS and a lot
1704 of the error messages complaining that MGS is not connected.
1708 Frequency : on network error
1709 Description: panic with double free request if network error
1710 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1711 but it's true only for synchronus enqueue, for async enqueue
1712 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1715 Severity : enhancement
1717 Description: client-side metadata stat-ahead during readdir(directory readahead)
1718 Details : perform client-side metadata stat-ahead when the client detects
1719 readdir and sequential stat of dir entries therein
1722 Frequency : on start mds
1724 Description: Implement get_info(last_id) in obdfilter.
1727 Frequency : occasional
1729 Description: Correctly check stale fid, not start epoch if ost not support SOM
1730 Details : open with flag O_CREATE need set old fid in op_fid3 because
1731 op_fid2 was overwritten with new generated fid, but mds can answer
1732 with one of these two fids and both is not stale. Setattr
1733 incorrectly started an epoch and assume will be called
1734 done_writing, but without SOM done_writing ever being called.
1737 Frequency : rare, depends on device drivers and load
1739 Description: MDS or OSS nodes crash due to stack overflow
1740 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1741 In some cases, in conjunction with device drivers that use a lot
1742 of stack the MDS (or possibly OSS) service threads could overflow
1743 the stack. One change which was identified to consume additional
1744 stack has been reworked to avoid the extra stack usage.
1747 Frequency : occasional
1749 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1750 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1751 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1752 oscc_has_objects and see OSC has no objests and call
1753 oscc_internal_create to resend create request.
1755 Severity : enhancement
1757 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1759 Severity : enhancement
1761 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1764 Frequency : very rare
1766 Description: Fix replay if there is an un-replied request and open
1767 Details : In some cases, older replay request will revert the
1768 mcd->mcd_last_xid on MDS which is used to record the client's
1769 latest sent request.
1771 Severity : enhancement
1773 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1775 Severity : enhancement
1777 Description: Add rhel5 support to HEAD.
1779 Serverity : enhancement
1781 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1786 Description: Don't allow skipping OSTs if index has been specified.
1787 Details : Don't allow skipping OSTs if index has been specified, make
1788 locking in internal create lots better.
1792 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1793 Details : access to bitfield in structure is always rounded to long
1794 and this produce problem with not atomic change any bit.
1798 Description: Lustre make rpms failed.
1799 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1800 builds Lustre rpms from tarball.
1803 Frequency : rare on shutdown ost
1805 Description: If llog cancel was not send before clean_exports phase, this can
1806 produce deadlock in llog code.
1807 Details : If llog thread has last reference to obd and call class_import_put
1808 this produce deadlock because llog_cleanup_commit_master wait when
1809 last llog_commit_thread exited, but this never success because was
1810 called from llog_commit_thread.
1814 Description: allow userland application know is lost one of stripes.
1815 Details : fill lvb_blocks with error code on ost and return it to
1816 application if error flag found.
1820 Description: NULL lov_tgts causing MDS oops
1821 Details : more safe checks for NULL lov_tgts for avoid oops.
1823 Severity : enhancement
1825 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1829 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1830 Details : If an OST is down the MDS will hang indefinitely in
1831 obd_statfs_rqset() waiting for the statfs data. While for
1832 MDS QOS usage of statfs, it should not stuck in waiting.
1834 Severity : enhancement
1836 Description: remote_acl support
1837 Details : Support ACL-based permission check for remote user.
1838 Support setfacl/getfacl for remote user with the utils
1839 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1840 the system "{s,g}etfacl" utils.
1842 Severity : enhancement
1844 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1846 Severity : enhancement
1848 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1852 Description: make mgs_setparam() handle fsname containing dash
1853 Details : fsname containing a dash does not work with lctl conf_param
1855 Severity : enhancement
1857 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1859 Severity : enhancement
1861 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1865 Frequency : rare, at shutdown
1866 Description: access already free / zero obd_namespace.
1867 Details : if client_disconnect_export was called without force flag set,
1868 and exist connect request in flight, this can produce access to
1869 NULL pointer (or already free pointer) when connect_interpret
1870 store ocd flags in obd_namespace.
1874 Frequency : only at startup
1875 Description: not alloc memory with spinlock held.
1876 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1877 if any spinlock held.
1879 Severity : enhancement
1881 Description: make lustre randomly fail allocating memory
1882 Details : Make lustre randomly failed allocating memory for testing purpose.
1884 Severity : enhancement
1886 Description: lost problems with lov objid file
1887 Details : Fixes some scability and access to not inited memory problems
1888 in work with lov objdid file.
1893 Description: lfs find does not continue on file error
1894 Details : Continue other files processing when a file/dir is absent.
1898 Description: Inconsistent usage of lustre_pack_reply()
1899 Details : Standardize the usage of lustre_pack_reply() such that it
1900 always generate a CERROR on failure.
1903 Frequency : occasional
1905 Description: MDS hang or stay in waiting lock
1906 Details : If client receive lock with CBPENDING flag ldlm need send lock
1907 cancel as separate rpc, to avoid situation when cancel request
1908 can't processed due all i/o threads stay in wait lock.
1912 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1916 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1920 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1924 Description: Update to RHEL4 latest kernel.
1926 Severity : enhancement
1928 Description: Build SLES10 patchless client fails
1929 Details : The configure was broken by run ./configure with
1930 --with-linux-obj=.... argument for patchless client. When the
1931 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1932 can't search header adequately. Use absolute path such as
1933 -I($LINUX)/include instead.
1937 Description: interrupt oig_wait produce painc on resend.
1938 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1939 private set, and this produce situation when rq_ptlrpcd_data not
1940 copyed to new allocated request and triggered LBUG on assert
1941 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1942 ptlrpcd set if request is changed and can be safety changed to use
1947 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1948 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1949 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1950 these macros and replace them with their difinitions instead.
1954 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1955 Details : Update which_patch & target file for SLES10 latest kernel.
1958 Description: Remove client patches from SLES 10 kernel.
1959 Details : This causes SLES 10 clients to behave as patchless clients
1960 even on a Lustre-patched (server) kernel.
1962 Severity : enhancement
1964 Description: self-adjustable client's lru lists
1965 Details : use adaptive algorithm for managing client cached locks lru
1966 lists according to current server load, other client's work
1967 pattern, memory activities, etc. Both, server and client
1968 side namespaces provide number of proc tunables for controlling
1971 Severity : enhancement
1973 Description: light-weight GSS support
1974 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1975 and reduce performance overhead.
1977 Severity : enhancement
1979 Description: Linux keyring support
1980 Details : Support using service of Linux keyring for Lustre GSS internal
1981 context refresh/cache mechanism.
1985 Description: Fix errors in lfs documentation
1986 Details : Fixes man pages
1990 Description: don't use GFP_* in generic Lustre code.
1991 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1995 Description: obdclass is limited by single OBD_ALLOC(idarray)
1996 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2000 Description: warnings with build patchless client with vanila 2.6.19 and up
2001 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2002 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2006 Frequency : Only for SLES
2007 Description: sanity_quota fail test_1
2008 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2009 file, which makes TSTID[2] inunique.
2013 Frequency : Only for SLES9
2014 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2015 Details : Update target/ChangeLog/which_patch .
2019 Frequency : Only for test_10 in sanity-quota.sh
2020 Description: a bug in quota test code
2021 Details : checking a test flag in a wrong place
2025 Frequency : Only for test_13 in sanity-quota.sh
2026 Description: a bug in quota test script
2027 Details : shouldn't check overfull allocated quota
2031 Description: jbd statistics
2032 Details : Port older jbd statistics patch for sles10
2034 Severity : enhancement
2036 Description: Add printing inode info into message about error in writepage.
2041 Description: reply_lock_interpret crash due to race with it and lock cancel.
2042 Details : Do not replay locks that are being cancelled. Do not reference
2043 locks by their address during replay, just by their handle.
2047 Frequency : When flocks are used.
2048 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2049 Details : Flock locks might destroy just granted lock if it could be merged
2050 with another existing flock, this is done in completion handler,
2051 so teach ldlm_cli_enquque_fini that this is a valid case for
2057 Description: Oops in read and write path when failing to allocate lock.
2058 Details : Check if lock allocation failed and return error back.
2062 Description: lstripe command fails for valid OST index
2063 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2064 instead of lov->desc.ld_active_tgt_count.
2068 Description: sanity.sh failed test 103
2069 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2074 Description: df doesn't work properly if diskfs blocksize != 4K
2075 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2079 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2080 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2082 Severity : enhancement
2084 Description: Adaptive timeouts
2085 Details : RPC timeouts adapt to changing server load and network
2086 conditions to reduce resend attempts and improve recovery time.
2090 Description: llapi_file_create() does not allow some changes
2091 Details : add llapi_file_open() that allows specifying the mode and
2092 open flags, and also returns an open file handle.
2096 Description: merge and cleanup kernel patches.
2097 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2101 Description: Add journal checksum support.(Kernel part)
2102 Details : The journal checksum feature adds two new flags i.e
2103 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2104 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2105 indicates that the commit block contains the checksum for
2106 the blocks described by the descriptor blocks. Now commit
2107 record can be sent to disk without waiting for descriptor
2108 blocks to be written to disk. This behavior is controlled
2109 using JBD2_FEATURE_ASYNC_COMMIT flag.
2113 Description: OSS needs mutliple precreate threads
2114 Details : Add ability to start more than one create thread per OSS.
2118 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2119 Details : execute lfs setstripe on client
2123 Description: mds_obd_create error creating tmp object
2124 Details : When the user sets quota on root, llog will be affected and can't
2125 create files and write files.
2128 Frequency : Always on ia64 patchless client, and possibly others.
2130 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2131 Details : This allows the patchless client to be loaded on architectures
2132 without this export.
2136 Description: RedHat Update kernel for RHEL5
2137 Details : Add the kernel config file for RHEL5.
2141 Description: RedHat Update kernel for RHEL5
2142 Details : Modify the kernel config file more closer RHEL5.
2146 Description: Build failure against Centos5 (RHEL5)
2147 Details : Define PAGE_SIZE when it isn't present.
2151 Description: test_7 on sanity-quota.sh can't be used on separate machines
2152 Details : add facet to handle it
2156 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2157 Details : with NFS, the anon dentry's parent was set to itself in
2158 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2159 corresponding dentry other than use rec->ur_name.
2161 Severity : enhancement
2163 Description: lfs setstripe enhancement
2164 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2168 Description: enable data checksumming by default
2169 Details : enable checksum by default, allow --disable-checksum
2170 configure option and "-o nochecksum" mount option. Checksums
2171 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2175 Description: lustre support for RHEL5
2176 Details : Add support for RHEL5.
2180 Description: Client eviction due to failover config
2181 Details : after a connection loss, the lustre client should attempt to
2182 reconnect to the last active server first before trying the
2183 other potential connections.
2187 Description: when mds and osts use different quota unit(32bit and 64bit),
2188 quota will be released repeatly.
2189 Details : void sending multiple quota reqs to mds, which will keep the
2190 status between the reqs.
2194 Description: osts not allocated evenly to files
2195 Details : change the condition to increase offset_idx
2198 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2200 Description: Data corruption for OSTs that are formatted larger than 2TB
2202 Details : When generating the bio request for lustre file writes the
2203 sector number would overflow a temporary variable before being
2204 used for the IO. The data reads correctly from Lustre (which
2205 will overflow in a similar manner) but other file data or
2206 filesystem metadata may be corrupted in some cases.
2210 Description: Tune the kernel for good SCSI performance.
2211 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2212 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2217 Description: rewrite ext2-derived code in obdclass/uuid.c
2218 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2219 from scratch preserving functionality.
2223 Description: "lfs find -obd UUID" prints directories
2224 Details : "lfs find -obd UUID" will return all directory names instead
2225 of just file names. It is incorrect because the directories
2226 do not reside on the OSTs.
2230 Description: use i_size_read and i_size_write in 2.6 port
2231 Details : replace inode->i_size access with i_size_read/write()
2234 Frequency : when removing large files
2236 Description: scheduling issue during removal of large Lustre files
2237 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2238 It causes scheduling issues when removing large files (17TB in the
2242 Frequency : only with liblustre clients on XT3
2244 Description: evictions taking too long
2245 Details : allow llrd to evict clients directly on OSTs
2248 Frequency : only on ppc
2250 Description: /proc/fs/lustre/devices broken on ppc
2251 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2252 We should make sure the type of 'index' is loff_t and then cast
2253 explicitly as needed below. Do not assign an explicitly cast
2257 Frequency : only for rhel5
2259 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2260 Details : Modify the target file & which_kernel.
2262 Severity : enhancement
2264 Description: omit set fsid for export NFS
2265 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2266 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2270 Frequency : Always, for kernels after 2.6.16
2271 Description: Fix warning idr_remove called for id=.. which is not allocated.
2272 Details : Last kernels save old s_dev before kill super and not allow
2273 to restore from callback - restore it before call kill_anon_super.
2277 Description: buffer overruns could theoretically occur
2278 Details : llapi_semantic_traverse() modifies the "path" argument by
2279 appending values to the end of the origin string, and a
2280 overrun may occur. Adding buffer overrun check in liblustreapi.
2284 Description: Fix error on 'ls .' at the top of the Lustre mount.
2285 Details : Don't revalidate dentry if it is a root dentry.
2289 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2290 Details : Modify vm-tunables-rhel4.patch.
2294 Description: Kernel config for 2.6.18-vanilla.
2295 Details : Modify targets/2.6-vanilla.target.in.
2296 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2297 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2298 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2299 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2303 Description: improve handling recoverable errors
2304 Details : if request processig with error which can be recoverable on server
2305 request should be resend, otherwise page released from cache and
2310 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2311 Details : Modify target file & which_patch.
2312 A flaw was found in the IA32 system call emulation provided
2313 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2314 value could be stored in the %RAX register, which could trigger an
2315 out-of-bounds system call table access. An untrusted local user
2316 could exploit this flaw to run code in the kernel
2317 (ie a root privilege escalation). (CVE-2007-4573).
2321 Description: O_DIRECT bypasses client statistics.
2322 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2323 read_ahead_stats were updated. copied the stats section from
2324 osc_send_oap_rpc() into async_internal().
2328 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2332 Description: change order of libsysio includes
2333 Details : '#include sysio.h' should always come before '#include xtio.h'
2337 Description: handle "serious error: objid * already exists" more gracefully
2338 Details : If LAST_ID value on disk is smaller than the objects existing in
2339 the O/0/d* directories, it indicates disk corruption and causes an
2340 LBUG(). If the object is 0-length, then we should use the existing
2341 object. This will help to avoid a full fsck in most cases.
2343 Severity : enhancement
2345 Description: adapt the lustre_config script to support the upgrade case
2346 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2347 server targets to 1.6 in parallel.
2351 Description: To avoid grant space > avaible space when the disk is almost
2352 full. Without this patch you might see the error "grant XXXX >
2353 available" or some LBUG about grant, when the disk is almost
2355 Details : In filter_check_grant, for non_grant cache write, we should
2356 check the left space by if (*left > ungranted + bytes), instead
2357 of (*left > ungranted), because only we are sure the left space
2358 is enough for another "bytes", then the ungrant space should be
2359 increase. In client, we should update cl_avail_grant only there
2360 is OBD_MD_FLGRANT in the reply.
2364 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2365 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2366 described in CVE-2007-4573. This problem could allow untrusted
2367 local users to gain root access.
2370 Frequency : when using O_DIRECT and quotas
2372 Description: Incorrect file ownership on O_DIRECT output files
2373 Details : block usage reported by 'lfs quota' does not take into account
2374 files that have been written with O_DIRECT.
2379 Description: touch file failed when fs is not full
2380 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2381 otherwise we can get ENOSP while fs is not full.
2385 Description: parallel lock callbacks
2386 Details : Instead of sending blocking and completion callbacks as separated
2387 requests, adding them to a set and sending in parallel.
2390 Frequency : only for Cray XT3
2391 Bugzilla : 12829/13455
2392 Description: Changing primary group doesn't change the group lustre assigns to
2394 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2395 overridden with the primary group provided by the group upcall,
2396 whereas the supplied fsgid can be trusted if it is in the list of
2397 supplementary groups returned by the group upcall.
2399 Severity : enhancement
2401 Description: Allow masking D_WARNING, D_ERROR messages from console
2402 Details : Console messages can now be disabled via lnet.printk.
2406 Description: User code with malformed file open parameter crashes client node
2407 Details : Before packing join_file req, all the related reference should be
2408 checked carefully in case some malformed flags cause fake
2409 join_file req on client.
2413 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2414 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2415 mode is equal to the granted mode. However, it should also check
2416 that LDLM_FL_CP_REQD is not set.
2420 Description: Heavy nfs access might result in deadlocks
2421 Details : After ELC code landed, it is now improper to enqueue any mds
2422 locks under och_sem, because enqueue might want to decide to
2423 cancel open locks for same inode we are holding och_sem for.
2427 Description: Client eviction while running blogbench
2428 Details : A lot of unlink operations with concurrent I/O can lead to a
2429 deadlock causing evictions. To address the problem, the number of
2430 oustanding OST_DESTROY requests is now throttled to
2431 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2432 callbacks are processed in priority.
2436 Description: enable ACLs on MDS by default
2437 Details : ACLs must be enabled on MDS by default.
2440 Frequency : PPC/PPC64 only
2442 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2443 Details : fix duplicated definitions between asm-ppc64/types.h and
2444 lustre_types.h on PPC.
2447 Frequency : PPC/PPC64 only
2449 Description: asm-ppc/segment.h does not exist
2450 Details : fix compile issue on PPC.
2454 Description: better handle error messages in extents code
2457 Frequency : RHEL4 only
2459 Description: mkfs is very slow on IA64/RHEL4
2460 Details : A performance regression has been discovered in the MPT Fusion
2461 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2462 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2463 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2465 Severity : enhancement
2467 Description: SNMP support enhancement
2468 Details : Adding total number of sampled request for an MDS node in snmp
2471 Severity : enhancement
2473 Description: Optimize ldlm waiting list processing for PR extent locks
2474 Details : When processing waiting list for read extent lock and meeting read
2475 lock that is same or wider to it that is not contended, skip
2476 processing rest of the list and immediatelly return current
2477 status of conflictness, since we are guaranteed there are no
2478 conflicting locks in the rest of the list.
2482 Description: Time out and refuse to reconnect
2483 Details : When the failover node is the primary node, it is possible
2484 to have two identical connections in imp_conn_list. We must
2485 compare not conn's pointers but NIDs, otherwise we can defeat
2486 connection throttling.
2490 Description: port llog fixes from b1_6 into HEAD
2491 Details : Port llog reference couting and some llog cleanups from b1_6
2492 (bug 10800) into HEAD, for protect from panic and access to already
2493 free llog structures.
2497 Description: Detect stride IO mode in read-ahead
2498 Details : When a client does stride read, read-ahead should detect that and
2499 read-ahead pages according to the detected stride pattern.
2503 Description: data checksumming impacts single node performance
2504 Details : add support for several checksum algorithm. Currently, only CRC32
2505 and Adler-32 are supported. The checksum type can be changed on
2506 the fly via /proc/fs/lustre/osc/*/checksum_type.
2510 Description: use adler32 for page checksums
2511 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2516 Description: build for x2 fails
2517 Details : fix compile issue on Cray systems.
2521 Description: Properly match for duplicate locks
2522 Details : Due to different lock order from skiplists code, we need to
2523 traverse entire list for now
2526 Frequency : only on PPC/SLES10
2528 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2529 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2530 be defined. Add a hack in mkfs_lustre.c to work around this compile
2535 Description: LASSERT on MDS when client holding flock lock dies
2536 Details : ldlm pool logic depends on number of granted locks equal to
2537 number of released locks which is not true for flock locks, so
2538 just exclude such locks from consideration.
2542 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2543 Details : Use fsfilt_sync() for both the whole filesystem sync and
2544 individual file sync to eliminate dangerous inode locking
2545 with I_LOCK that can lead to a deadlock.
2549 Description: performance in 1.6.3
2550 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2551 sure that 1MB requests are not fragmented by the block layer.
2553 Severity : enhancement
2555 Description: organize the server-side client stats on per-nid basis
2556 Details : Change the structure of stats under obdfilter and mds to
2564 The "uuid"s file would list the uuids of _active_ exports.
2565 And the clear entry is to clear all stats and stale nids.
2567 Severity : enhancement
2569 Description: eliminate client locks in face of contention
2570 Details : file contention detection and lockless i/o implementation
2571 for contended files.
2575 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2578 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2580 Description: MDS may lose file striping (and hence file data) in some cases
2581 Details : If there are additional extended attributes stored on the MDS,
2582 in particular ACLs, SELinux, or user attributes (if user_xattr
2583 is specified for the client mount options) then there is a risk
2584 of attribute loss. Additionally, the Lustre file striping
2585 needs to be larger than default (e.g. striped over all OSTs),
2586 and an additional attribute must be stored initially in the
2587 inode and then increase in size enough to be moved to the
2588 external attribute block (e.g. ACL growing in size) for file
2593 Description: skiplist implementation simplification
2594 Details : skiplists are used to group compatible locks on granted list
2595 that was implemented as tracking first and last lock of each
2596 lock group the patch changes that to using doubly linked lists
2600 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2601 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2602 because LASSERT on both the data supplied by a client, and the
2603 data on disk is dangerous and incorrect.
2605 Severity : enhancement
2607 Description: Slow truncate/writes to huge files at high offsets.
2608 Details : Directly associate cached pages to lock that protect those pages,
2609 this allows us to quickly find what pages to write and remove
2610 once lock callback is received.
2614 Description: more ldlm soft lockups
2615 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2616 starve other threads from the resource lock for a long time in
2617 case of long waiting queue, so change the debug level from
2618 D_OTHER to the less frequently used D_INFO.
2620 Severity : enhancement
2622 Description: add -gid, -group, -uid, -user options to lfs find
2626 Description: Hung threads in invalidate_inode_pages2_range
2627 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2628 once one is completed. As a result, some RPCs are stuck in the
2629 queue and are never sent.
2633 Description: filter threads hungs on waiting journal commit
2634 Details : Cleanup filter group llog code, then only filter group llog will
2635 be only created in the MDS/OST syncing process.
2639 Description: Procfs and llog threads access destoryed import sometimes.
2640 Details : Sync the import destoryed process with procfs and llog threads by
2641 the import refcount and semaphore.
2643 Severity : enhancement
2645 Description: openlock cache of b1_6 port to HEAD
2650 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2651 Details : If insertion of an extent fails, then discard the inode
2652 preallocation and free data blocks else it can lead to duplicate
2657 Description: don't always update ctime in ext3_xattr_set_handle()
2658 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2659 In some cases the ctime should not be updated, for example for
2660 2.0->1.8 compatibility it is necessary to delete an xattr and it
2661 should not update the ctime.
2665 Bugzilla : 15713/16362
2666 Description: Assertion in iopen_connect_dentry in 1.6.3
2667 Details : looking up an inode via iopen with the wrong generation number can
2668 populate the dcache with a disconneced dentry while the inode
2669 number is in the process of being reallocated. This causes an
2670 assertion failure in iopen since the inode's dentry list contains
2671 both a connected and disconnected dentry.
2675 Description: assertion failure in ldlm_handle2lock()
2676 Details : fix a race between class_handle_unhash() and class_handle2object()
2677 introduced in lustre 1.6.5 by bug 13622.
2682 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2683 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2688 Description: oops in sd_iostats_seq_show()
2689 Details : unloading/reloading the scsi low level driver triggers a kernel
2690 bug when trying to access the sd iostat file.
2695 Description: Kernel panics during QLogic driver reload
2696 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2697 patch, causing memory corruption.
2702 Description: journal_dev option does not work in b1_6
2703 Details : pass mount option during pre-mount.
2705 Severity : enhancement
2707 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2708 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2709 extent information of a file. It can be used to map logical blocks
2710 in a file to physical blocks in the block device.
2714 Description: LDLM soft lockups - improvement
2715 Details : It is be possible to send the lock handle along with each read
2716 or write request because the client is already doing a lock match
2717 itself so there isn't any reason the OST should have to re-do that
2721 Frequency : only on Cray X2
2723 Description: X2 build failures
2724 Details : fix build failures on Cray X2.
2728 Description: xid & resent requests
2729 Details : Initialize RPC XID from clock at startup (randomly if clock is
2732 Severity : enhancement
2734 Description: Add lustre_start utility to start or stop multiple Lustre servers
2739 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2740 Details : In case of memory pressure, list_del() can be called twice on
2741 req->rq_history_list, causing a kernel oops.
2745 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2746 Details : fix stack overflow in the distributed lock manager by defering
2747 export eviction after a failed AST to the elt thread instead of
2748 handling it in the dlm interpret routine.
2752 Description: Convert some comments to new format.
2753 Details : Update documenting comments to match doxygen conventions.
2757 Description: Grammar fixes.
2758 Details : A couple of trivial spelling fixes.
2762 Description: OSD_COUNTERS-mandatory
2763 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2764 to simplify the code.
2768 Description: simplify cmm_device freeing logic.
2769 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2773 Description: Add lockdep support to dt_object_operations locking interface.
2774 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2775 parameter indicating lock ordering. Update mdd code to use new
2780 Description: Introduce failloc constants for lockless IO tests.
2781 Details : Add two new failloc constants to test lockless IO. Only one of
2782 them in implemented---another is checked in yet to be landed
2787 Description: Add lockdep support for inode mutex.
2788 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2789 in the situations where Lustre has to lock more than one inode
2794 Description: Add optional invariants checking support.
2795 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2796 configure switch. This macro is to be used for consistency and
2797 sanity checks that are too expensive to be left in `production'
2802 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2803 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2808 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2809 Details : Use spin_lock_nested() in (the only) situation where more than
2810 one ldlm_lock is locked simultaneously. Also, fix possible
2811 dead-lock in ldlm_lock_change_resource() by enforcing particular
2816 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2817 Details : Instead of specifying each ldlm_lock call-back through separate
2818 parameter, wrap them into struct ldlm_callback_suite.
2822 Description: Kill join_lru obd method and its callers.
2823 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2824 in memory---a requirement implemented through ->o_join_lru() obd
2825 method in HEAD. Remove this method and its users.
2829 Description: Add asynchronous ldlm ENQUEUE completion handler.
2830 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2831 case that stock ldlm_completion_ast() cannot handle as it waits
2832 until lock is granted. Introduce new ldlm_completion_ast_async()
2833 for this. Also comment ldlm_completion_ast().
2837 Description: ldlm_error <-> errno conversion.
2838 Details : Add functions to map (rather arbitrary) between LDLM error codes
2839 and standard errno values. CLIO needs this to prevent LDLM specific
2840 constants from escaping ldlm and osc.
2844 Description: Kill unused ldlm_handle2lock_ns() function.
2845 Details : Kill unused ldlm_handle2lock_ns() function.
2849 Description: Add lu_ref support to ldlm_lock
2850 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2851 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2852 and ldlm_resource. LDLM interface has to be changed, because code that
2853 releases a reference on a lock, has to "know" what reference this is.
2854 In the most frequent case
2856 lock = ldlm_handle2lock(handle);
2858 LDLM_LOCK_PUT(lock);
2860 no changes are required. When any other reference (received _not_ from
2861 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2862 instead of LDLM_LOCK_PUT().
2864 Arguably, changes are pervasive, and interface requires some discipline
2865 for proper use. On the other hand, it was very instrumental in finding
2866 a few leaked lock references.
2870 Description: Add ldlm_lock_addref_try().
2871 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2872 attempts to addref a lock that might be being canceled
2877 Description: Add ldlm_weigh_callback().
2878 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2879 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2880 hard-coded `number of pages' logic.
2884 Description: Add lockdep annotations to llog code.
2885 Details : Use appropriately tagged _nested() locking calls in the places
2886 where llog takes more than one ->lgh_lock lock.
2890 Description: Add loi_kms_set().
2891 Details : Wrap kms updates into a helper function.
2895 Description: Constify instances of struct lsm_operations.
2896 Details : Constify instances of struct lsm_operations.
2900 Description: lu_conf support.
2901 Details : On a server, a file system object is uniquely identified
2902 by a fid, which is sufficient to locate and load all object
2903 state (inode). On a client, on the other hand, more data are
2904 necessary instantiate an object. Change lu_object_find() and
2905 friends to take additional `lu_conf' argument describing object.
2906 Typically this includes layout information.
2910 Description: lu_context fixes.
2911 Details : Introduce new lu_context functions that are needed on the client
2912 side, where some system threads (ptlrpcd) are shared by multiple
2913 modules, and so cannot be stopped during module shutdown.
2917 Description: Add start and stop methods to lu_device_type_operations.
2918 Details : Introduce two new methods in lu_device_type_operations, that are
2919 invoked when first instance of a given type is created and last one
2920 is destroyed respectively. This is need by CLIO.
2924 Description: Add lu_ref support to struct lu_device.
2925 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2926 track leaked references.
2930 Description: Introduce lu_kmem_descr.
2931 Details : lu_kmem_descr and its companion interface allow to create
2932 and destroy a number of kmem caches at once.
2936 Description: Fix lu_object finalization race.
2937 Details : Fix a race between lu_object_find() finding an object and its
2938 concurrent finalization. This race is (most likely) not possible
2939 on the server, but might happen on the client.
2943 Description: Introduce lu_ref interface.
2944 Details : lu_ref is a debugging module allowing to track references to
2945 a given object. It is quite cpu expensive, and has to be
2946 explicitly enabled with --enable-lu_ref. See usage description
2951 Description: Factor lu_site procfs stats into a separate function.
2952 Details : Separate lu_site stats printing code into a separate function
2953 that can be reused on a client.
2957 Description: Constify instances of struct {lu,dt,md}_device_operations.
2958 Details : Constify instances of struct {lu,dt,md}_device_operations.
2962 Description: Introduce struct md_site and move meta-data specific parts of
2963 struct lu_site here.
2964 Details : Move md-specific fields out of struct lu_site into special struct
2965 md_site, so that lu_site can be used on a client.
2969 Description: Kill mdd_lov_destroy().
2970 Details : Remove unused mdd code.
2974 Description: Add st_block checking to multistat.c.
2975 Details : Add st_block checking to multistat.c.
2979 Description: Add lu_ref support to struct obd_device.
2980 Details : Add lu_ref tracking to obd_device.
2984 Description: Kill obd_set_fail_loc().
2985 Details : Remove unused code.
2989 Description: Add special type for ptlrpc_request interpret functions.
2990 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2991 there. Actual usage will be in CLIO.
2995 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2996 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2997 doesn't work with lockdep.
3001 Description: Add rwv.c test program.
3002 Details : New testing program exercising readv(2) and writev(2) (Qian).
3006 Description: Add sendfile.c test program.
3007 Details : New testing program exercising sendfile(2) (Jay).
3011 Description: Ratelimit a message that can be very frequent.
3012 Details : Ratelimit a memory allocation failure message that can
3017 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3018 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3022 Description: Liblustre build fixes.
3023 Details : Liblustre build fixes.
3027 Description: libcfs: add cfs_{need,cond}_resched() interface.
3028 Details : libcfs: add cfs_{need,cond}_resched() definition and
3029 implementations for Linux, NT, and liblustre.
3031 Severity : enhancement
3033 Description: More exported tunables for mballoc
3034 Details : Add support for tunable preallocation window and new tunables for
3035 large/small requests
3039 Description: Detect corruption of block bitmap and checking for preallocations
3040 Details : Checks validity of on-disk block bitmap. Also it does better
3041 checking of number of applied preallocations. When corruption is
3042 found, it turns filesystem readonly to prevent further corruptions.
3046 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3047 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3048 window to 0. If it does hit cache miss, and read-pattern is still
3049 stride-io mode, does not reset the stride window, but also does not
3050 increase the stride window length in this case.
3054 Frequency : only for big-endian servers
3055 Description: Check if system is big-endian while mounting fs with extents feature
3056 Details : Mounting a filesystem with extents feature will fail on big-endian
3057 systems since ext3-based ldiskfs is not supported on big-endian
3058 systems. This can be over-riden with "bigendian_extents" mount option.
3060 Severity : enhancement
3062 Description: The root squash functionality
3063 Details : A security feature, which is to prevent users from being able
3064 to mount lustre on their desktop, run as root, and delete
3065 all of the files in the filesystem. The goal is accomplished by
3066 remapping user id (UID) and group id (GID) of the root user to
3067 a UID and GID specified by the system administartor via Lustre
3068 configuration management server (MGS). The functionality also
3069 allows to specify sets of clients for which the remapping does
3074 Description: Excessive recovery window
3075 Details : With AT enabled, the recovery window can be excessively long (6000+
3076 seconds). To address this problem, we no longer use
3077 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3078 timeout no longer depends on the service time, it is set to
3079 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3080 time via pb_service_time.
3084 Description: Watchdog triggered on MDS failover
3085 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3086 the OSTs know that the MDS "UUID" can be reused for the same export
3087 from a different NID, so we do not need to wait for the export to be
3091 Frequency : rare, only if using MMP with Linux RAID
3093 Description: MMP doesn't work with Linux RAID
3094 Details : While using HA for Lustre servers with Linux RAID, it is possible
3095 that MMP will not detect multiple mounts. To make this work we
3096 need to unplug the device queue in RAID when the MMP block is being
3097 written. Also while reading the MMP block, we should read it from
3098 disk and not the cached one.
3100 Severity : enhancement
3102 Description: open file using fid
3103 Details : A file can be opened using just its fid, like
3104 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3107 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3110 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3111 Details : add proper locking for creation of jbd_4k slab cache
3115 Description: MMP check in ext3_remount() fails without displaying any error
3116 Details : When multiple mount protection fails during remount, proper error
3119 Severity : enhancement
3121 Description: Allow stripe size to be up to 4G-64k
3122 Details : Fix math logic to allow large stripe sizes.
3126 Description: add check for >8TB ldiskfs filesystems
3127 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3128 Don't allow >8TB ldiskfs filesystems to be mounted without
3129 force_over_8tb mount option
3134 Descriptoin: OSS complains "too many blocks" and crashes
3135 Details : In filter_commitrw_write():
3136 1. lquota_chkquota();
3137 2. obdo_from_inode();
3138 3. lquota_pending_commit();
3139 "2." has changed uid and gid in this case, then caused
3140 lquota_chkquota() and lquota_pending_commit() to use
3141 different uid and gid. That is the root of the bug.
3145 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3146 Details : Detect and print out the 'stripe count = -1' case properly.
3150 Description: Out or order replies might be lost on replay
3151 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3152 request with tid smaller than one currently being added, add it
3153 to the start, not end of the list.
3155 --------------------------------------------------------------------------------
3157 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3159 * Support for kernels:
3160 2.6.5-7.283 (SLES 9),
3161 2.6.9-55.EL (RHEL 4),
3162 2.6.16.46-0.14 (SLES 10),
3163 2.6.18.8 vanilla (kernel.org)
3164 * Client support for unpatched kernels:
3165 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3166 2.6.16 - 2.6.22 vanilla (kernel.org)
3167 * Due to recently discovered recovery problems, we do not recommend
3168 using patchless RHEL 4 clients with this or any earlier release.
3169 * Recommended e2fsprogs version: 1.39.cfs8
3170 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3171 * Starting with this release, the ldiskfs backing filesystem required
3172 by Lustre is now in its own package, lustre-ldiskfs. This package
3173 should be installed. It is versioned separately from Lustre and
3174 may be released separately in future.
3178 Description: block reactivating mgc import until all deactivates complete
3179 Details : Fix race when failing back MDT/MGS to itself (testing)
3182 Frequency : at statup only
3184 Description: mds_lov_synchronize race leads to various problems
3185 Details : simultaneous MDT->OST connections at startup can cause the
3186 sync to abort, leaving the OSC in a bad state.
3188 Severity : enhancement
3190 Description: add optional extra BUILD_VERSION info
3191 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3192 to override the lustre version.
3195 Frequency : 2.6.18 servers only
3197 Description: ll_kern_mount() doesn't release the module reference
3198 Details : The ldiskfs module reference count never drops down to 0
3199 because ll_kern_mount() doesn't release the module reference.
3204 Description: server LBUG when using old ost_num_threads parameter
3205 Details : Accept the old ost_num_threads parameter but warn that it
3206 is deprecated, and fix an off-by-one error that caused an LBUG.
3211 Description: Transient SCSI error results in persistent IO issue
3212 Details : iobuf->dr_error is not reinitialized to 0 between two
3216 Frequency : sometimes when underlying device returns I/O errors
3218 Description: OSTs not going read-only during write failures
3219 Details : OSTs are not remounted read-only when the journal commit threads
3220 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3221 instead of the ext3 wrappers.
3225 Description: poor connect scaling with increasing client count
3226 Details : Don't run filter_grant_sanity_check for more than 100 exports
3227 to improve scaling for large numbers of clients.
3230 Frequency : SLES10 only
3232 Description: sanity-quota.sh quotacheck failed: rc = -22
3233 Details : Quotas cannot be enabled on SLES10.
3236 Frequency : liblustre clients only
3238 Description: getdirentries does not give error when run on compute nodes
3239 Details : getdirentries does not fail when the size specified as an argument
3240 is too small to contain at least one entry
3242 Severity : enhancement
3244 Description: Add LNET router traceability for debug purposes
3245 Details : If a checksum failure occurs with a router as part of the
3246 IO path, the NID of the last router that forwarded the bulk data
3247 is printed so it can be identified.
3252 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3253 Details : Due to a race condition, liblustre clients were occasionally
3254 evicted incorrectly.
3256 Severity : enhancement
3258 Description: lfs setstripe use optional parameters instead of postional
3261 Severity : enhancement
3263 Description: Nanosecond timestamp support for ldiskfs
3264 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3265 resolution timestamps. There is not yet support for this at
3266 the Lustre filesystem level.
3269 Frequency : during server recovery
3271 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3272 Details : request with rq_no_resend flag not awake l_wait_event if they get
3276 Frequency : nfs export on patchless client
3278 Description: connectathon hang when test nfs export over patchless client
3279 Details : Disconnected dentry cannot be found with lookup, so we do not need
3280 to unhash it or make it invalid
3283 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3284 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3285 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3290 Description: open req refcounting wrong on reconnect
3291 Details : If reconnect happened between getting open reply from server and
3292 call to mdc_set_replay_data in ll_file_open, we will schedule
3293 replay for unreferenced request that we are about to free.
3294 Subsequent close will crash in variety of ways.
3295 Check that request is still eligible for replay in
3296 mdc_set_replay_data().
3301 Description: disable writes to filesystem when reading health_check file
3302 Details : the default for reading the health_check proc file has changed
3303 to NOT do a journal transaction and write to disk, because this
3304 can cause reads of the /proc file to hang and block HA state
3305 checking on a healthy but otherwise heavily loaded system. It
3306 is possible to return to the previous behaviour during configure
3307 with --enable-health-write.
3309 Severity : enhancement
3311 Description: 64-bit inode version
3312 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3313 to the inode. This will be required for version-based recovery.
3318 Description: MDS fails to start if a duplicate client export is detected
3319 Details : in some rare cases it was possible for a client to connect to
3320 an MDS multiple times. Upon recovery the MDS would detect this
3321 and fail during startup. Handle this more gracefully.
3323 Severity : enhancement
3325 Description: Add -o localflock option to simulate old noflock
3327 Details : This will achieve local-only flock/fcntl locks
3333 Description: log_commit_thread vs filter_destroy race leads to crash
3334 Details : Take import reference before releasing llog record semaphore
3339 Description: Wrong request locking in request set processing
3340 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3341 imp_delayed_list, in this place should be used imp_lock.
3344 Frequency : when reconnection
3346 Description: Grant Leak when osc reconnect to OST
3347 Details : When osc reconnect ost, OST(filter) should check whether it
3348 should grant more space to client by comparing fed_grant and
3349 cl_avail_grant, and return the granted space to client instead
3350 of "new granted" space, because client will call osc_init_grant
3351 to update the client grant space info.
3354 Frequency : when client reconnect to OST
3356 Description: Grant Leak when osc do resend and replay bulk write
3357 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3358 bulk write request, because the grant info while be sync between
3359 OSC and OST when reconnect, and we should ignore the grant info
3360 these of resend/replay write req.
3365 Description: Grant space more than available left space sometimes.
3366 Details : When then OST is about to be full, if two bulk writing from
3367 different clients came to OST. Accord the avaliable space of the
3368 OST, the first req should be permitted, and the second one
3369 should be denied by ENOSPC. But if the seconde arrived before
3370 the first one is committed. The OST might wrongly permit second
3371 writing, which will cause grant space > available space.
3374 Frequency : when client is evicted
3376 Description: Grant might be wrongly erased when osc is evicted by OST
3377 Details : when the import is evicted by server, it will fork another
3378 thread ptlrpc_invalidate_import_thread to invalidate the
3379 import, where the grant will be set to 0. While the original
3380 thread will update the grant it got when connecting. So if
3381 the former happened latter, the grant will be wrongly errased
3382 because of this race.
3387 Description: Checking Stale with correct fid
3388 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3389 to check whether it is stale, because sometimes, we want the
3390 enqueue happened anyway, and op_data.fid2 will not be initialized.
3392 Severity : enhancement
3394 Description: update patchless client
3395 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3398 Frequency : only with 2.4 kernel
3400 Description: random memory corruption
3401 Details : size of struct ll_inode_info is to big for union inode.u and this
3402 can be cause of random memory corruption.
3407 Description: Memory leak in recovery
3408 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3409 It should also check obd_fail before fsfilt_start, otherwise if
3410 fsfilt_start return -EROFS,(failover mds during mds recovery).
3411 then the req will return with repmsg->transno = 0 and rc = EROFS.
3412 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3413 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3414 be freed no matter whether fsfilt_commit success or not.
3417 Frequency : only with huge count clients
3419 Description: Prevents from taking the superblock lock in llap_from_page for
3421 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3422 for llap_from_page call in ll_removepage prevents from taking the
3423 superblock lock for a soon died page.
3428 Description: Not check open intent error before release open handle
3429 Details : in some rare cases, the open intent error is not checked before
3430 release open handle, which may cause
3431 ASSERTION(open_req->rq_transno != 0), because it tries to release
3432 the failed open handle.
3437 Description: Set cat log bitmap only after create log success.
3438 Details : in some rare cases, the cat log bitmap is set too early. and it
3439 should be set only after create log success.
3443 Description: Accessing a block bevice can re-enable I/O when Lustre is
3444 tearing down a device.
3445 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3450 Description: service threads may hog cpus when there are a lot of requests
3451 Details : Insert cond_resched to give other threads a chance to use some CPU
3456 Description: the cat log was not initialized in recovery
3457 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3458 unlink log on mds will not be initialized until mds post
3459 recovery. And also in mds post recovery, the unlink log will
3460 initialization will be done asynchronausly, so there will be race
3461 between add unlink log and unlink log initialization.
3465 Description: brw_stats were being printed incorrectly
3466 Details : brw_stats were being printed as log2 but all of them were not
3467 recorded as log2. Also remove some code duplication arising from
3468 filter_tally_{read,write}.
3472 Frequency : rare, only in recovery.
3473 Description: ASSERTION(req->rq_type != LI_POISON) failed
3474 Details : imp_lock should be held while iterating over imp_sending_list for
3475 prevent destroy request after get timeout in ptlrpc_queue_wait.
3479 Description: replay-single.sh test 52 fails
3480 Details : A lock's skiplist need to be cleanup when it being unlinked
3481 from its resource list.
3485 Description: Short directio read returns full requested size rather than
3487 Details : Direct I/O operations should return actual amount of bytes
3488 transferred rather than requested size.
3490 Severity : enhancement
3492 Description: metadata RPC reduction (e.g. for rm performance)
3493 Details : decrease the amount of synchronous RPC between clients and servers
3494 by canceling conflicing lock before the operation on the client
3495 and packing thier handles into the main operation RPC to server.
3497 Severity : enhancement
3499 Description: Async OSC create to avoid the blocking unnecessarily.
3500 Details : If an OST has no remaining object, system will block on the
3501 creation when it needs to create a new object on this OST. Now,
3502 ways use pre-created objects when available, instead of blocking on
3503 an empty osc while others are not empty. If we must block, we
3504 block for the shortest possible period of time.
3508 Description: improve handling recoverable errors
3509 Details : if request processig with error which can be recoverable on server
3510 request should be resend, otherwise page released from cache and
3513 Severity : enhancement
3515 Description: refine locking for avoid write wrong info into lov_objid file
3516 Details : fix possible races with add new target and write/update data in
3519 --------------------------------------------------------------------------------
3521 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3526 Frequency : on some architectures
3528 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3529 Details : Uninitialized flags sometimes cause configuration commands to
3533 Frequency : patchless clients only
3535 Description: missing __iget() symbol export
3536 Details : The __iget() symbol export is missing. To avoid the need for
3537 this on patchless clients the deathrow inode reaper is turned
3538 off, and we depend on the VM to clean up old inodes. This
3539 dependency was during via the fix for bug 12181.
3541 --------------------------------------------------------------------------------
3543 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3545 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3546 INTEROPERATE with older versions automatically. In many cases a
3547 special upgrade step is needed. Please read the
3548 user documentation before upgrading any part of a live system.
3549 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3550 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3551 * WARNING: Lustre configuration and startup changes are required with
3552 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3556 Severity : enhancement
3558 Description: MountConf
3559 Details : Lustre configuration is now managed via mkfs and mount
3560 commands instead of lmc and lconf. New obd types (MGS, MGC)
3561 are added for dynamic configuration management. See
3562 https://mail.clusterfs.com/wikis/lustre/MountConf for
3565 Severity : enhancement
3567 Description: dynamic OST addition
3568 Details : OSTs can now be added to a live filesystem
3570 Severity : enhancement
3572 Description: startup order invariance
3573 Details : MDTs and OSTs can be started in any order. Clients only
3574 require the MDT to complete startup.
3576 Severity : enhancement
3578 Description: parallel, asynchronous orphan cleanup
3579 Details : orphan cleanup is now performed in separate threads for each
3580 OST, allowing parallel non-blocking operation.
3582 Severity : enhancement
3584 Description: optimized stripe assignment
3585 Details : stripe assignments are now made based on ost space available,
3586 ost previous usage, and OSS previous usage, in order to try
3587 to optimize storage space and networking resources.
3589 Severity : enhancement
3591 Description: Permanently set tunables
3592 Details : All writable /proc/fs/lustre tunables can now be permanently
3593 set on a per-server basis, at mkfs time or on a live system.
3595 Severity : enhancement
3597 Description: Lustre message v2
3598 Details : Add lustre message format v2.
3600 Severity : enhancement
3602 Description: client OST exclusion list
3603 Details : Clients can be started with a list of OSTs that should be
3604 declared "inactive" for known non-responsive OSTs.
3608 Description: SPEC SFS validation failure on NFS v2 over lustre.
3609 Details : Changes the blocksize for regular files to be 2x RPC size,
3610 and not depend on stripe size.
3612 Severity : enhancement
3614 Description: Multiple MD RPCs in flight.
3615 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3616 To avoid overly-overloading MDS, introduce a limit on number of
3617 MDS RPCs in flight for a single client and add /proc controls
3618 to adjust this limit.
3620 Severity : enhancement
3622 Description: client read/write statistics
3623 Details : Add client read/write call usage stats for performance
3624 analysis of user processes.
3625 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3626 file access. extents_stats shows chunk size distribution.
3627 extents_stats_per_process show chunk size distribution per
3630 Severity : enhancement
3632 Description: mds statistics
3633 Details : Add detailed mds operations statistics in
3634 /proc/fs/lustre/mds/*/stats.
3638 Description: Failure of copying files with lustre special EAs.
3639 Details : Client side always return success for setxattr call for lustre
3640 special xattr (currently only "trusted.lov").
3645 Description: Refcount LNET uuids
3646 Details : The global LNET uuid list grew linearly with every startup;
3647 refcount repeated list entries instead of always adding to
3650 Severity : enhancement
3652 Description: Dynamic service threads
3653 Details : Within a small range, start extra service threads
3654 automatically when the request queue builds up.
3657 Frequency : mixed-endian client/server environments
3659 Description: mixed-endian crashes
3660 Details : The new msg_v2 system had some failures in mixed-endian
3663 Severity : enhancement
3665 Description: Easy OST removal
3666 Details : OSTs can be permanently deactivated with e.g. 'lctl
3667 conf_param lustre-OST0001.osc.active=0'
3669 Severity : enhancement
3671 Description: MGS proc entries
3672 Details : Added basic proc entries for the MGS showing what filesystems
3675 Severity : enhancement
3677 Description: provide MGS failover
3678 Details : Added config lock reacquisition after MGS server failover.
3680 Severity : enhancement
3682 Description: add Linux 2.4 support
3683 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3687 Description: a large application tries to do I/O to the same resource and dies
3688 in the middle of it.
3689 Details : Check the req->rq_arrival time after the call to
3690 ost_brw_lock_get(), but before we do anything about
3691 processing it & sending the BULK transfer request. This
3692 should help move old stale pending locks off the queue as
3693 quickly as obd_timeout.
3696 Frequency : when an incorrect nid is specified during startup
3698 Description: ptlrpc connect to non-existant node causes kernel crash
3699 Details : LNET can't be re-entered from an event callback, which
3700 happened when we expire a message after the export has been
3701 cleaned up. Instead, hand the zombie cleanup off to another
3704 Severity : enhancement
3706 Description: plain/inodebits lock performance improvement
3707 Details : Grouping plain/inodebits in granted list by their request modes
3708 and bits policy, thus improving the performance of search through
3712 Frequency : only if OST filesystem is corrupted
3714 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3715 Details : for a short time RPCs with bulk IO are in the replay list,
3716 but replay of bulk IOs is unimplemented. If the OST filesystem
3717 is corrupted due to disk cache incoherency and then replay is
3718 started it is possible to trip an assertion. Avoid putting
3719 committed RPCs into the replay list at all to avoid this issue.
3722 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3724 Description: System hang on startup
3725 Details : This bug allowed the liblustre (e.g. catamount) client to
3726 return to the app before handling all startup RPCs. This
3727 could leave the node unresponsive to lustre network traffic
3728 and manifested as a server ptllnd timeout.
3730 Severity : enhancement
3732 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3733 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3734 boolean to control whether to print peer debug info when a
3735 client's RPC times out.
3738 Frequency : only for kernels with patches from Lustre below 1.4.3
3740 Description: Remove old rdonly API
3741 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3744 Frequency : only for devices with external journals
3746 Description: Set external device read-only also
3747 Details : During a commanded failover stop, we set the disk device
3748 read-only while the server shuts down. We now also set any
3749 external journal device read-only at the same time.
3752 Frequency : when upgrading from 1.4 while trying to change parameters
3754 Description: The wrong (new) MDC name was used when setting parameters for
3755 upgraded MDT's. Also allows changing of OSC (and MDC)
3756 parameters if --writeconf is specified at tunefs upgrade time.
3759 Frequency : when setting specific ost indicies
3761 Description: QOS code breaks on skipped indicies
3762 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3763 created with --index need not be sequential.
3765 Severity : enhancement
3767 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3768 Details : The uninit_groups feature works in conjunction with the kernel
3769 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3770 up the pass1 processing of e2fsck. This is a read-only feature
3771 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3772 mount filesystems that have had this feature enabled.
3774 Severity : enhancement
3776 Description: Improve multi-block allocation algorithm to avoid fragmentation
3777 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3778 allocation locality and avoid filesystem fragmentation.
3780 ------------------------------------------------------------------------------
3782 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3784 * Support for kernels:
3785 2.4.21-47.0.1.EL (RHEL 3)
3786 2.6.5-7.283 (SLES 9)
3787 2.6.9-42.0.10.EL (RHEL 4)
3788 2.6.12.6 vanilla (kernel.org)
3789 2.6.16.27-0.9 (SLES 10)
3790 * Recommended e2fsprogs version: 1.39.cfs5
3792 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3796 Frequency : occasional, depends on client load and configuration
3797 Bugzilla : 12181, 12203
3798 Description: data loss for recently-modified files
3800 Details : In some cases it is possible that recently written or created
3801 files may not be written to disk in a timely manner (this should
3802 normally be within 30s unless client IO load is very high).
3803 The problem appears as zero-length files or files that are a
3804 multiple of 1MB in size after a client crash or client eviction
3805 that are missing data at the end of the file.
3807 This problem is more likely to be hit on clients where files are
3808 repeatedly created and unlinked in the same directory, clients
3809 have a large amount of RAM, have many CPUs, the filesystem has
3810 many OSTs, the clients are rebooted frequently, and/or the files
3811 are not accessed by other nodes after being written.
3813 The presence of the problem can be detected by looking at
3814 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3815 smaller than the second (nr_unused) then dirty files will not
3816 be flushed automatically to disk. "sync; sleep 10" should be
3817 run several times on the node before unmounting it to update
3818 Lustre (this is also safe to run on nodes without this problem).
3820 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3821 that can cause this same problem, so customers using that kernel
3822 also need to update the kernel in addition to Lustre. In order
3823 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3825 It is normal that files written just before a client crash (less
3826 than 30s) may not yet have been flushed to disk, even for local
3830 Frequency : frequent on thin XT3 nodes
3832 Description: UUID collision on thin XT3 Linux nodes
3833 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3834 randomly, since we relied on an insufficiently-seeded PRNG.
3839 Description: OSS hangs after "All ost request buffers busy"
3840 Details : A deadlock between quota and journal operations caused OSS
3841 hangs after printing "All ost request buffers busy."
3844 Frequency : always on liblustre builds
3846 Description: Cleanup compiler warnings on liblustre
3849 Frequency : always on liblustre builds on XT3
3851 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3855 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3856 > 32000 subdirectories
3857 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3858 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3859 subdirectory count crosses 32000. This will aid e2fsck to
3860 correctly handle more than 32000 subdirectories.
3863 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3865 Description: System hang on startup
3866 Details : This bug allowed the liblustre (e.g. catamount) client to
3867 return to the app before handling all startup RPCs. This
3868 could leave the node unresponsive to lustre network traffic
3869 and manifested as a server ptllnd timeout.
3871 Severity : enhancement
3873 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3874 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3875 boolean to control whether to print peer debug info when a
3876 client's RPC times out.
3881 Description: make O_SYNC working on 2.6 kernels
3882 Details : 2.6 kernels use different method for mark pages for write,
3883 so need add a code to lustre for O_SYNC work.
3888 Description: Failure to close file and release space on NFS
3889 Details : Put inode details into lock acquired in ll_intent_file_open.
3890 Use mdc_intent_lock in ll_intent_open to properly
3891 detect all kind of errors unhandled by mdc_enqueue.
3896 Description: proc file read during shutdown sometimes raced obd removal,
3898 Details : Add lock to prevent obd access after proc file removal.
3901 Frequency : Only for files larger than 4GB on 32-bit clients.
3903 Description: improperly doing page alignment of locks
3904 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3905 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3910 Description: RPCs being resent when they shouldn't be
3911 Details : Some RPCs that should not be resent are being resent. This
3912 can cause inconsistencies in the RPC state machine. Do not
3913 resend such requests.
3916 Frequency : rare, only with NFS export
3918 Description: Crash on NFS re-export node
3919 Details : under very unusual load conditions an assertion is hit in
3920 ll_intent_file_open()
3923 Frequency : only if OST filesystem is corrupted
3925 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3926 Details : for a short time RPCs with bulk IO are in the replay list,
3927 but replay of bulk IOs is unimplemented. If the OST filesystem
3928 is corrupted due to disk cache incoherency and then replay is
3929 started it is possible to trip an assertion. Avoid putting
3930 committed RPCs into the replay list at all to avoid this issue.
3935 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3936 Details : Large single O_DIRECT read and write calls can fail to allocate
3937 a sufficiently large buffer to process the request. In case of
3938 allocation failure the allocation is retried with a smaller
3939 buffer and broken into smaller requests.
3941 Severity : enhancement
3943 Description: Add -o localflock option to simulate old noflock behaviour.
3944 Details : This will achieve local-only flock/fcntl locks coherentness.
3949 Description: versioning check is incomplete
3950 Details : Checking the version difference of client vs. server, report
3951 error if the gap is too big.
3956 Description: add support PG_writeback bit
3957 Details : add support for PG_writeback bit for Lustre, for more carefull
3958 work with page cache in 2.6 kernel. This also fix some deadlocks
3959 and remove hack for work O_SYNC with 2.6 kernel.
3961 Severity : enhancement
3963 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3964 Details : The uninit_groups feature works in conjunction with the kernel
3965 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3966 up the pass1 processing of e2fsck. This is a read-only feature
3967 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3968 mount filesystems that have had this feature enabled.
3970 Severity : enhancement
3972 Description: Improve multi-block allocation algorithm to avoid fragmentation
3973 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3974 allocation locality and avoid filesystem fragmentation.
3976 ------------------------------------------------------------------------------
3978 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3980 * Support for kernels:
3981 2.6.9-42.0.3.EL (RHEL 4)
3982 2.6.5-7.276 (SLES 9)
3983 2.4.21-47.0.1.EL (RHEL 3)
3984 2.6.12.6 vanilla (kernel.org)
3985 2.6.16.21-0.8 (SLES10)
3986 * Recommended e2fsprogs version: 1.39.cfs2-0
3988 * The backwards-compatible /proc/sys/portals symlink has been removed
3989 in this release. Before upgrading, please ensure that you change
3990 any configuration scripts or /etc/sysctl.conf files that access
3991 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3992 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3993 in advance of the upgrade on any system running Lustre 1.4.6 or
3994 newer, since /proc/sys/lnet was added in that version.
3995 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3999 Frequency : only when quota is used
4001 Description: avoid scanning export list for quota master
4002 Details : Change the algorithms to avoid scanning export list in order
4003 to improve the efficiency.
4006 Frequency : MDS failover only, very rarely
4008 Description: "went back in time" messages on mds failover
4009 Details : The greatest transno may be lost when the current operation
4010 finishes with an error (transno==0) and the client's last_rcvd
4011 record is over-written. Save the greatest transno in the
4012 mds_last_transno for this case.
4015 Frequency : always for specific kernels and striping counts
4017 Description: client may get "Matching packet too big" without ACL support
4018 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4019 when trying to access files in certain configurations. The
4020 clients should in fact be denied when mounting because they do
4021 not understand ACLs.
4024 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4026 Description: many clients connecting with IO in progress causes connect timeouts
4027 Details : Avoid synchronous journal commits to avoid delays caused by many
4028 clients connecting/disconnecting when bulk IO is in progress.
4029 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4030 OST_IO_PORTAL to avoid delays behind potentially many pending
4034 Frequency : occasionally with multiple writers to a single file
4036 Description: shared writes to file may result in wrong size reported by stat()
4037 Details : Allow growing of kms when extent lock is cancelled
4040 Frequency : always with random mmap IO to multi-striped file
4042 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4043 Details : Only if the hole is at the end of OST object so that kms is too
4044 small. Fix is to increase kms accordingly in ll_nopage.
4047 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4049 Description: writes to a missing object would leak memory on the OST
4050 Details : If there is an inconsistency between the MDS and OST filesystems,
4051 such that the MDS references an object that doesn't exist, writes
4052 to that object will leak memory due to incorrect cleanup in the
4053 error handling path, eventually running out of memory on the OST.
4058 Description: Creating too long symlink causes lustre errors
4059 Details : Check symlink and name lengths before sending requests to MDS.
4062 Frequency : only if flock is enabled (not on by default)
4064 Description: posix locks not released on fd closure on 2.6.9+
4065 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4066 kernels, this caused such locks not to be released on fd close and
4067 then assertions on fs unmount about still used locks.
4070 Frequency : MDS failover only, very rarely
4072 Description: clients may get ASSERTION(granted_lock != NULL)
4073 Details : When request was taking a long time, and a client was resending
4074 a getattr by name lock request. The were multiple lock requests
4075 with the same client lock handle and
4076 mds_getattr_name->fixup_handle_for_resent_request found one of the
4077 lock handles but later failed with ASSERTION(granted_lock != NULL).
4082 Description: handle->h_buffer_credits > 0, assertion failure
4083 Details : h_buffer_credits is zero after truncate, causing assertion
4084 failure. This patch extends the transaction or creates a new
4088 Frequency : NFS re-export or patchless client
4089 Bugzilla : 11179, 10796
4090 Description: Crash on NFS re-export node (__d_move)
4091 Details : We do not want to hash the dentry if we don't have a lock.
4092 But if this dentry is later used in d_move, we'd hit uninitialised
4093 list head d_hash, so we just do this to init d_hash field but
4094 leave dentry unhashed.
4097 Frequency : NFS re-export or patchless client
4099 Description: NFS exports has problem with symbolic link
4100 Details : lustre client didn't properly install dentry when re-exported
4101 to NFS or running patchless client.
4104 Frequency : NFS re-export or patchless client
4106 Description: Various nfs/patchless fixes.
4107 Details : fixes reuse disconected alias for lookup process - this fixes
4108 warning "find_exported_dentry: npd != pd",
4109 fix permission error with open files at nfs.
4110 fix apply umask when do revalidate.
4113 Frequency : occasional
4115 Description: Crash on NFS re-export node
4116 Details : calling clear_page() on the wrong pointer triggered oops in
4117 generic_mapping_read().
4120 Frequency : rarely, using O_DIRECT IO
4122 Description: unaligned directio crashes client with LASSERT
4123 Details : check for unaligned buffers before trying any requests.
4126 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4128 Description: stale data returned from RAID cache
4129 Details : If only a small amount of IO is done to the RAID device before
4130 reading it again it is possible to get stale data from the RAID
4131 cache instead of reading it from disk.
4134 Frequency : always for sles10 kernel
4136 Description: sles10 support
4137 Details : ll_follow_link: compile fixes and using of nd_set_link
4138 under newer kernels.
4141 Frequency : depends on arch, kernel and compiler version, always on sles10
4144 Description: recursive or deep enough symlinks cause stack overflow
4145 Details : getting rid of large stack-allocated variable in
4149 Frequency : depends on hardware
4151 Description: lustre write performance loss in the SLES10 kernel
4152 Details : the performance loss is caused by using of write barriers in the
4153 ext3 code. The SLES10 kernel turns barrier support on by
4154 default. The fix is to undo that change for ldiskfs.
4156 ------------------------------------------------------------------------------
4158 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4160 * Support for kernels:
4161 2.6.9-42.0.3EL (RHEL 4)
4162 2.6.5-7.276 (SLES 9)
4163 2.4.21-47.0.1.EL (RHEL 3)
4164 2.6.12.6 vanilla (kernel.org)
4168 Frequency : quota enabled and large files being deleted
4170 Description: releasing more than 4GB of quota at once hangs OST
4171 Details : If a user deletes more than 4GB of files on a single OST it
4172 will cause the OST to spin in an infinite loop. Release
4173 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4178 Description: statfs data retrieved from /proc may be stale or zero
4179 Details : When reading per-device statfs data from /proc, in the
4180 {kbytes,files}_{total,free,avail} files, it may appear
4181 as zero or be out of date.
4184 Frequency : systems with MD RAID1 external journal devices
4186 Description: lconf's call to blkid is confused by RAID1 journal devices
4187 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4188 of returning all block devices that match the journal UUID.
4191 Frequency : always, for aggregate stripe size over 4GB
4193 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4194 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4195 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4198 Frequency : always if "lfs find" used on a local file/directory
4200 Description: "lfs find" segfaults if used on a local file/directory
4201 Details : The case where a directory component was not specified wasn't
4202 handled correctly. Handle this properly.
4205 Frequency : always on ppc64
4207 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4208 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4209 perform endianness conversion.
4212 Frequency : rarely (truncate to non-zero file size after write under load)
4213 Bugzilla : 10730, 10687
4214 Description: Files padded with zeros to next 4K multiple
4215 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4216 it is possible that files that are truncated to a non-zero size
4217 immediately after being written are filled with zero bytes beyond
4218 the truncated size. No file data is lost.
4220 Severity : enhancement
4222 Description: Allow recovery/failover for liblustre clients.
4223 Details : liblustre clients were unaware of failover configurations until
4226 Severity : enhancement
4228 Description: user file locks should fail when not mounting with flock option
4229 Details : Set up an error-returning stub in ll_file_operations.lock field
4230 to prevent incorrect behaviour when client is mounted without
4231 flock option. Also, set up properly f_op->flock field for
4235 Frequency : always on ia64
4237 Description: "lfs df" loops on printing out MDS statfs information
4238 Details : The obd_ioctl_data was not initialized and in some systems
4239 this caused a failure during the ioctl that did not return
4240 an error. Initialize the struct and return an error on failure.
4243 Frequency : SLES 9 only
4245 Description: Error of copying files with lustre special EAs as root
4246 Details : Client side always return success for setxattr call for lustre
4247 special xattr (currently only "trusted.lov").
4250 Frequency : rarely on clusters with both ia64+i386 clients
4252 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4253 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4254 clients are doing concurrent writes to the same file it is
4255 possible that the ia64 clients may LASSERT because the OST
4256 extent locks are not PAGE_SIZE aligned. Ensure that grown
4257 locks are always aligned on the request boundary.
4260 Frequency : specific use, occasional
4262 Description: Overwriting in use executable truncates on-disk binary image
4263 Details : If one node attempts to overwrite an executable in use by
4264 another node, we now correctly return ETXTBSY instead of
4265 truncating the file.
4267 Severity : enhancement
4269 Description: Async OSC create to avoid the blocking unnecessarily.
4270 Details : If an OST has no remaining object, system will block on the creating
4271 when need to create a new object on this OST. Now, ways use
4272 pre-created objects when available, instead of blocking on an
4273 empty osc while others are not empty. If we must block, we block
4274 for the shortest possible period of time.
4279 Description: chmod on Lustre root is propagated to other clients
4280 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4281 invalid by the follow_mount time.
4284 Frequency : liblustre clients only
4286 Description: Race in 'instant cancel' lock handling could lead to such locks
4287 never to be granted in case of SMP MDS
4288 Details : Do not destroy not yet granted but cbpending locks in
4292 Frequency : replay/resend of open
4294 Description: non null lock assetion failure in mds_intent_policy
4295 Details : Trying to replay/resend lockless open requests resulted in
4296 mds_open() returning 0 with no lock. Now it sets a flag if
4297 a lock is going to be returned.
4299 Severity : enhancement
4301 Description: Checksum enhancements
4302 Details : New checksum enhancements allow for resending RPCs that failed
4305 Severity : enhancement
4307 Description: Tunables on number of dirty pages in cacche
4308 Details : Allow to set limit on number of dirty pages cached.
4313 Description: client crash on unmount - lock still has references
4314 Details : In some error handling cases it was possible to leak a lock
4315 reference on a client while accessing a file. This was not
4316 harmful to the client during operation, but would cause the
4317 client to crash when the filesystem is unmounted.
4320 Frequency : specific case, rare
4322 Description: ETXTBSY on mds though file not in use by client
4323 Details : ETXTBSY is no longer incorrectly returned when attempting to
4324 chmod or chown a directory that the user previously tried to
4325 execute or a currently-executing binary.
4328 Frequency : extremely rare except on liblustre-based clients
4330 Description: Lustre space not freed when files are deleted
4331 Details : Clean up open-unlinked files after client eviction. Previously
4332 the unlink was skipped and the files remained as orphans.
4337 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4338 Details : In some cases it was possible to send a blocking callback to
4339 a client doing a glimpse, even though that client didn't get
4340 a lock granted. When the glimpse lock is cancelled on the OST
4341 the freed lock is left on the waiting list and corrupted the list.
4344 Frequency : all core dumps
4346 Description: Broke core dumps to lustre
4347 Details : Negative dentry may be unhashed if parent does not have UPDATE
4348 lock, but some callers, e.g. do_coredump, expect dentry to be
4349 hashed after successful create, hash it in ll_create_it.
4351 ------------------------------------------------------------------------------
4353 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4355 * Support for kernels:
4356 2.6.9-42.0.2.EL (RHEL 4)
4357 2.6.5-7.276 (SLES 9)
4358 2.4.21-40.EL (RHEL 3)
4359 2.6.12.6 vanilla (kernel.org)
4363 Frequency : always on RHEL 3
4365 Description: Number of open files grows over time
4366 Details : The number of open files grows over time, whether or not
4367 Lustre is started. This was due to a filp leak introduced
4368 by one of our kernel patches.
4370 ------------------------------------------------------------------------------
4372 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4374 * Support for kernels:
4375 2.6.9-42.EL (RHEL 4)
4376 2.6.5-7.276 (SLES 9)
4377 2.4.21-40.EL (RHEL 3)
4378 2.6.12.6 vanilla (kernel.org)
4383 Bugzilla : 5719, 9635, 9792, 9684
4384 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4385 Details : If a server is under heavy load and cannot reply to new
4386 connection requests before the client resends the (re)connect,
4387 the connection handling code can behave badly if two service
4388 threads are concurrently handing separate (re)connections from
4389 the same client. Add better locking to the connection handling
4390 code, and ensure that only a single connection will be processed
4391 for a given client UUID, even if the lock is dropped.
4393 Severity : enhancement
4395 Description: add TCP zero-copy support to kernel
4396 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4397 sends if the hardware supports scatter-gather and checksumming.
4398 This allows socklnd to do client-write and server-read more
4399 efficiently and reduce CPU utilization from skbuf copying.
4402 Frequency : only if NFS exporting from client
4404 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4405 Details : If setting attributes on a file created under NFS that had
4406 never been opened it would be possible to oops the client
4407 if the file had no objects.
4410 Frequency : always for liblustre
4412 Description: liblustre client does MDS+OSTs setattr RPC for each write
4413 Details : When doing a write from a liblustre client, the client
4414 incorrectly issued an RPC to the MDS and each OST the file was
4415 striped over in order to update the timestamps. When writing
4416 with small chunks and many clients this could overwhelm the MDS
4417 with RPCs. In all cases it would slow down the write because
4418 these RPCs are unnecessary.
4420 Severity : enhancement
4422 Description: allow number of MDS service threads to be changed at module load
4423 Details : It is now possible to change the number of MDS service threads
4424 running. Adding "options mds mds_num_threads={N}" to the MDS's
4425 /etc/modprobe.conf will set the number of threads for the next
4426 time Lustre is restarted (assuming the "mds" module is also
4427 reloaded at that time). The default number of threads will
4428 stay the same, 32 for most systems.
4433 Description: OST crash if filesystem is unformatted or corrupt
4434 Details : If an OST is started on a device that has never been formatted
4435 or if the filesystem is corrupt and cannot even mount then the
4436 error handling cleanup routines would dereference a NULL pointer.
4441 Description: NULL pointer deref in llap_from_page.
4442 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4443 incorrect) mapping if the page was truncated/reclaimed while it was
4444 searched for. Check for this condition and skip such pages when
4445 doing readahead. Introduce extra check to llap_from_page() to
4446 verify page->mapping->host is non-NULL (so page is not anonymous).
4449 Frequency : Sometimes when using sys_sendfile
4451 Description: "page not covered by a lock" warnings from ll_readpage
4452 Details : sendfile called ll_readpage without right page locks present.
4453 Now we introduced ll_file_sendfile that does necessary locking
4454 around call to generic_file_sendfile() much like we do in
4458 Frequency : with certain MDS communication failures at client mount time
4460 Description: NULL pointer deref after failed client mount
4461 Details : a client connection request may delayed by the network layer
4462 and not be sent until after the PTLRPC layer has timed out the
4463 request. If the client fails the mount immediately it will try
4464 to clean up before the network times out the request. Add a
4465 reference from the request import to the obd device and delay
4466 the cleanup until the network drops the request.
4469 Frequency : occasionally during client (re)connect
4471 Description: assertion failure during client (re)connect
4472 Details : processing a client connection request may be delayed by the
4473 client or server longer than the client connect timeout. This
4474 causes the client to resend the connection request. If the
4475 original connection request is replied in this interval, the
4476 client may trip an assertion failure in ptlrpc_connect_interpret()
4477 which thought it would be the only running connect process.
4480 Frequency : only with obd_echo servers and clients that are rebooted
4482 Description: kernel BUG accessing uninitialized data structure
4483 Details : When running an obd_echo server it did not start the ping_evictor
4484 thread, and when a client was evicted an uninitialized data
4485 structure was accessed. Start the ping_evictor in the RPC
4486 service startup instead of the OBD startup.
4488 Severity : enhancement
4489 Bugzilla : 10193 (patchless)
4490 Description: Remove dependency on various unexported kernel interfaces.
4491 Details : No longer need reparent_to_init, exit_mm, exit_files,
4492 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4495 Frequency : rare (only users of deprecated and unsupported LDAP config)
4497 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4498 Details : LDAP apparently contains 'lustreName' attributes instead of
4499 'name'. A simple remapping of the name is sufficient.
4502 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4504 Description: waiting_locks_callback trips kernel BUG if client is evicted
4505 Details : Running with the dump_on_timeout debug flag turned on makes
4506 it possible that the waiting_locks_callback() can try to dump
4507 the Lustre kernel debug logs from an interrupt handler. Defer
4508 this log dumping to the expired_lock_main() thread.
4510 Severity : enhancement
4512 Description: Support NFS exporting on 2.6 kernels.
4513 Details : Implement non-rawops metadata methods for NFS server to use without
4514 changing NFS server code.
4517 Frequency : very rare (synthetic metadata workload only)
4519 Description: two racing renames might cause an MDS thread to deadlock
4520 Details : Running the "racer" program may cause one MDS thread to rename
4521 a file from being the source of a rename to being the target of
4522 a rename at exactly the same time that another thread is doing
4523 so, and the second thread has already enqueued these locks after
4524 doing a lookup of the target and is trying to relock them in
4525 order. Ensure that we don't try to re-lock the same resource.
4528 Frequency : only very large systems with liblustre clients
4530 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4531 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4532 all OSTs in parallel. This allows the eviction of stale liblustre
4533 clients to proceed much faster than if they were done in series,
4534 and also offers similar improvements for other set_info RPCs.
4539 Description: excessive CPU usage during initial read phase on client
4540 Details : During the initial read phase on a client, it would agressively
4541 retry readahead on the file, consuming too much CPU and impacting
4542 performance (since 1.4.5.8). Improve the readahead algorithm
4543 to avoid this, and also improve some other common cases (read
4544 of small files in particular, where "small" is files smaller than
4545 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4550 Description: MDS crash when receiving packet with unknown intent.
4551 Details : Do not LBUG in unknown intent case, just return -EFAULT
4553 Severity : enhancement
4554 Bugzilla : 9293, 9385
4555 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4556 Details : Do not serialize getattr (non-intent version) and statfs.
4559 Frequency : occasional, when OST network is overloaded/intermittent
4561 Description: client evicted by OST after bulk IO timeout
4562 Details : If a client sends a bulk IO request (read or write) the OST
4563 may evict the client if it is unresposive to its data GET/PUT
4564 request. This is incorrect if the network is overloaded (takes
4565 too long to transfer the RPC data) or dropped the OST GET/PUT
4566 request. There is no need to evict the client at all, since
4567 the pinger and/or lock callbacks will handle this, and the
4568 client can restart the bulk request.
4571 Frequency : Always when mmapping file with no objects
4573 Description: client crashes when mmapping file with no objects
4574 Details : Check that we actually have objects in a file before doing any
4575 operations on objects in ll_vm_open, ll_vm_close and
4581 Description: Request leak when working with deleted CWD
4582 Details : Introduce advanced request refcount tracking for requests
4583 referenced from lustre intent.
4585 Severity : Enhancement
4587 Description: Cache open file handles on client.
4588 Details : MDS now will return special lock along with openhandle, if
4589 requested and client is allowed to hold openhandle, even if unused,
4590 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4591 closing files for every read/write openration.
4593 Severity : Enhancement
4595 Description: Cache open negative dentries on client when possible.
4596 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4597 negative dentries on lock revocation.
4602 Description: Remounting a client read-only wasn't possible with a zconf mount
4603 Details : It wasn't possible to remount a client read-only with llmount.
4605 Severity : enhancement
4606 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4607 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4608 MPICH is included to add Lustre-specific ADIO interfaces.
4609 This is based closely on the UFS ADIO layer and only differs
4610 in file creation, in order to allow the OST striping to be set.
4611 This is user-contributed code and not supported by CFS.
4616 Description: extended inode attributes (immutable, append-only) work improperly
4617 when 2.4 and 2.6 kernels are used on client/server or vice versa
4618 Details : Introduce kernel-independent values for these flags.
4620 Severity : enhancement
4623 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4624 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4625 granularity. Now, Fractional MB granularity is be supported,
4626 this is very useful for low memory system.
4628 Severity : enhancement
4630 Description: Getattr by fid
4631 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4632 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4633 have locks and avoids getattr rpc requests for it on every stat.
4636 Frequency : Always, for filesystems larger than 2TB
4638 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4639 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4640 filesystems larger than 2TB to be mounted with mballoc enabled.
4643 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4645 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4646 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4647 filesystems larger than 2TB is mounted then the kernel will
4648 silently corrupt the start of the filesystem. CONFIG_LBD is
4649 enabled for all CFS-supported kernels, but the possibility of
4650 this happening with a modified kernel config exists.
4652 Severity : enhancement
4654 Description: add client O_DIRECT support for 2.6 kernels
4655 Details : It is now possible to do O_DIRECT reads and writes to files
4656 in the Lustre client mountpoint on 2.6 kernel clients.
4658 Severity : enhancement
4660 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4661 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4662 parallel, not waiting for response from every OST before sending
4663 a rpc to the next OST.
4668 Description: setattr vs write race when updating file timestamps
4669 Details : Client processes that update a file timestamp into the past
4670 right after writing to the file (e.g. tar) it is possible that
4671 the updated file modification time can be reset to the current
4672 time due to a race between processing the setattr and write RPC.
4674 Severity : enhancement
4676 Description: Bring 'lfs find' closer in line with regular Linux find.
4677 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4678 -print0 options and obtains all the needed info through the lustre
4681 Severity : enhancement
4683 Description: support up to 1024 configured devices on one node
4684 Details : change obd_dev array from statically allocated to dynamically
4685 allocated structs as they are first used to reduce memory usage
4690 Description: Flush dirty partially truncated pages during truncate
4691 Details : Immediatelly flush partially truncated pages in filter_setattr,
4692 this way we completely avoid having any pages in page cache on OST
4693 and can retire ugly workarounds during writes to flush such pages.
4698 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4699 Details : protect inode from truncation within vfs_unlink() context
4700 just take a reference before calling vfs_unlink() and release it
4701 when parent's i_sem is free.
4706 Description: last_id value checked outside lock on OST caused LASSERT failure
4707 Details : If there were multiple MDS->OST object precreate requests in
4708 flight, it was possible that the OST's last object id was checked
4709 outside a lock and incorrectly tripped an assertion. Move checks
4710 inside locks, and discard old precreate requests.
4713 Frequency : always, if extents are used on OSTs
4715 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4716 Details : OSTs using the extents format would not zero the high 16 bits of
4717 the index physical block number. This is not a problem for any
4718 OST filesystems smaller than 16TB, and no kernels support ext3
4719 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4720 new/modified files) and can be fixed for existing filesystems
4721 with e2fsprogs-1.39-cfs1.
4726 Description: import connection selection may be incorrect if timer wraps
4727 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4728 import connections to be ignored if the 32-bit jiffies counter
4729 wraps. Use a 64-bit jiffies counter.
4732 Frequency : very large clusters immediately after boot
4734 Description: LNET request buffers exhausted under heavy short-term load
4735 Details : If a large number of client requests are generated on a service
4736 that has previously never seen so many requests it is possible
4737 that the request buffer growth cannot keep up with the spike in
4738 demand. Instead of dropping incoming requests, they are held in
4739 the LND until the RPC service can accept more requests.
4742 Frequency : Sometimes during replay
4744 Description: Assertion failure in ll_local_open after replay.
4745 Details : If replay happened on an open request reply before we were able
4746 to set replay handler, reply will become not swabbed tripping the
4747 assertion in ll_local_open. Now we set the handler right after
4748 recognising of open request
4751 Frequency : very rare
4753 Description: kernel reports "badness in vsnprintf"
4754 Details : Reading from the "recovery_status" /proc file in small chunks
4755 may cause a negative length in lprocfs_obd_rd_recovery_status()
4756 call to vsnprintf() (which is otherwise harmless). Exit early
4757 if there is no more space in the output buffer.
4759 Severity : enhancement
4761 Description: clear OBD RPC statistics by writing to them
4762 Details : It is now possible to clear the OBD RPC statistics by writing
4763 to the "stats" file.
4768 Description: Client mtime is not the same on different clients after utimes
4769 Details : In some cases, the client was using the utimes() syscall on
4770 a file cached on another node. The clients now validate the
4771 ctime from the MDS + OSTs to determine which one is right.
4776 Description: Inability to activate failout mode
4777 Details : lconf script incorrectly assumed that in pythong string's numeric
4778 value is used in comparisons.
4781 Frequency : always with multiple stripes per file
4783 Description: Inefficient object allocation for mutli-stripe files
4784 Details : When selecting which OSTs to stripe files over, for files with
4785 a stripe count that divides evenly into the number of OSTs,
4786 the MDS is always picking the same starting OST for each file.
4787 Return the OST selection heuristic to the original design.
4792 Description: mount failures may take full timeout to return an error
4793 Details : Under some heavy load conditions it is possible that a
4794 failed mount can wait for the full obd_timeout interval,
4795 possibly several minutes, before reporting an error.
4796 Instead return an error as soon as the status is known.
4798 Frequency : quota enabled and large files being deleted
4800 Description: releasing more than 4GB of quota at once hangs OST
4801 Details : If a user deletes more than 4GB of files on a single OST it
4802 will cause the OST to spin in an infinite loop. Release
4803 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4808 Description: statfs data retrieved from /proc may be stale or zero
4809 Details : When reading per-device statfs data from /proc, in the
4810 {kbytes,files}_{total,free,avail} files, it may appear
4811 as zero or be out of date.
4814 Frequency : systems with MD RAID1 external journal devices
4816 Description: lconf's call to blkid is confused by RAID1 journal devices
4817 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4818 of returning all block devices that match the journal UUID.
4821 Frequency : always, for aggregate stripe size over 4GB
4823 Description: assertion fails when trying to use 4GB stripe size
4824 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4825 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4828 Frequency : always on ppc64
4830 Description: the first write on an ext3 filesystem with mballoc got stuck
4831 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4832 perform endianness conversion.
4834 ------------------------------------------------------------------------------
4836 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4838 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4839 INTEROPERATE with older versions automatically. Please read the
4840 user documentation before upgrading any part of a live system.
4841 * WARNING: Lustre networking configuration changes are required with
4842 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4845 * Support for kernels:
4846 2.6.9-22.0.2.EL (RHEL 4)
4847 2.6.5-7.244 (SLES 9)
4848 2.6.12.6 vanilla (kernel.org)
4851 Severity : enhancement
4852 Bugzilla : 7981/8208
4853 Description: Introduced Lustre Networking (LNET)
4854 Details : LNET is new networking infrastructure for Lustre, it includes
4855 a reorganized network configuration mode (see the user
4856 documentation for full details) as well as support for routing
4857 between different network fabrics. Lustre Networking Devices
4858 (LNDs) for the supported network fabrics have also been
4859 created for this new infrastructure.
4861 Severity : enhancement
4862 Description: Introduced Access control lists
4863 Details : clients can set ACLs on files and directories in order to have
4864 more fine-grained permissions than the standard Unix UGO+RWX.
4865 The MDS must be started with the "-o acl" mount option.
4867 Severity : enhancement
4868 Description: Introduced filesystem quotas
4869 Details : Administrators may now establish per-user quotas on the
4872 Severity : enhancement
4874 Description: Configuration change for the XT3
4875 The PTLLND is now used to run Lustre over Portals on the XT3
4876 The configure option(s) --with-cray-portals are no longer used.
4877 Rather --with-portals=<path-to-portals-includes> is used to
4878 enable building on the XT3. In addition to enable XT3 specific
4879 features the option --enable-cray-xt3 must be used.
4884 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4885 Details : A race between allocating a new llcd and re-getting the llcd_lock
4886 allowed another thread to grab newly-allocated llcd.
4888 Severity : enhancement
4890 Description: 2.6 OST async journal commit and locking fix to improve performance
4891 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4892 2.6 kernels are now async as they already were in 2.4 kernels so
4893 that they can commit concurrently with the network bulk transfer.
4894 For block-allocated files the filter allocation semaphore is held
4895 to avoid filesystem fragmentation during allocation. BKL lock
4896 removed for 2.6 xattr operations where it is no longer needed.
4901 Description: lconf incorrectly determined whether two IP networks could talk
4902 Details : In some more complicated routing and multiple-network
4903 configurations, lconf will avoid trying to make a network
4904 connection to a disjoint part of the IP space. It was doing the
4905 math incorrectly for one set of cases.
4910 Description: Fix for potential infinite loop processing records in an llog.
4911 Details : If an llog record is corrupted/zeroed, it is possible to loop
4912 forever in llog_process(). Validate the llog record length
4913 and skip the remainder of the block on error.
4916 Frequency : occasional (liblustre only)
4918 Description: liblustre could not open files whose last component is a symlink
4919 Details : sysio_path_walk() would incorrectly pass the open intent to
4920 intermediate path components.
4923 Frequency : rare (liblustre only with non-standard tuning)
4924 Bugzilla : 7201 (7350)
4925 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4926 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4927 after it has been released, only do this for referenced locks
4929 Severity : enhancement
4931 Description: specifying an (invalid) directory default stripe_size of -1
4932 would reset the directory default striping
4933 Details : stripe_size -1 was used internally to signal directory stripe
4934 removal, now use "all default" to signal dir stripe removal
4935 as a directory striping of "all default" is not useful
4938 Frequency : common for large clusters running liblustre clients
4940 Description: doing an ls when liblustre clients are running is slow
4941 Details : sending a glimpse AST to a liblustre client waits for every AST
4942 to time out, as liblustre clients will not respond. Since they
4943 cannot cache data we refresh the OST lock LVB from disk instead.
4945 Severity : enhancement
4947 Description: doing an ls at the same time as file IO can be slow
4948 Details : enqueue and other "small" requests can be blocked behind many
4949 large IO requests. Create a new OST IO portal for non-IO
4950 requests so they can be processed faster.
4953 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4955 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4956 created file with mode 000
4957 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4958 allow an ftruncate() as a non-root user to a file with mode 000.
4959 Lustre now respects this flag to disable mode checks when
4960 truncating a file owned by the user
4963 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4966 Description: Revoking locks from clients that went dead or catatonic might take
4968 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4969 cancellation of such locks instant on servers without waiting for
4970 any reply from clients. Clients drops these locks when cancel
4971 notification from server is received without replying.
4974 Frequency : liblustre-only, when liblustre client dies or becomes busy
4976 Description: Doing ls on Linux clients can take a long time with active
4978 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4979 granting such locks to it in the first place if possible. Locks
4980 are taken by proxy on the OST during the read or write and
4981 dropped immediately afterward. Add connect flags handling, do
4982 not grant locks to liblustre clients for glimpse ASTs.
4984 Severity : enhancement
4986 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4987 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4988 sequential reads and behaves badly with applications doing random
4989 reads. Improve it by reading ahead at least read region, and
4990 avoiding excessive large RPC for small reads.
4992 Severity : enhancement
4994 Description: Creating more than 1000 files for a single job may cause a load
4995 imbalance on the OSTs if there are also a large number of OSTs.
4996 Details : qos_prep_create() uses an OST index reseed value that is an
4997 even multiple of the number of available OSTs so that if the
4998 reseed happens in the middle of the object allocation it will
4999 still utilize the OSTs as uniformly as possible.
5004 Description: OST or MDS may oops in ping_evictor_main()
5005 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5006 but doesn't restart at beginning of obd_exports_timed list
5009 Severity : enhancement
5011 Description: improve by-nid export eviction on the MDS and OST
5012 Details : allow multiple exports with the same NID to be evicted at one
5013 time without re-searching the exports list.
5016 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5018 Description: MDS may oops in groups_free()
5019 Details : in rare race conditions a newly allocated group_info struct is
5020 freed again, and this can be NULL. The 2.4 compatibility code
5021 for groups_free() checked for a NULL pointer, but 2.6 did not.
5024 Frequency : common for liblustre clients doing little filesystem IO
5025 Bugzilla : 9352, 7313
5026 Description: server may evict liblustre clients accessing contended locks
5027 Details : if a client is granted a lock or receives a completion AST
5028 with a blocking AST already set it would not reply to the AST
5029 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5032 Frequency : lfs setstripe, only systems with more than 160 OSTs
5034 Description: unable to set striping with a starting offset beyond OST 160
5035 Details : llapi_create_file() incorrectly limited the starting stripe
5036 index to the maximum single-file stripe count.
5039 Frequency : LDAP users only
5041 Description: lconf did not handle in-kernel recovery with LDAP properly
5042 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5044 Severity : enhancement
5046 Description: bind OST threads to NUMA nodes to improve performance
5047 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5048 node and do their allocations there to localize memory access
5050 Severity : enhancement
5052 Description: llmount can determine client NID directly from Myrinet (GM)
5053 Details : the client NID code from gmnalnid was moved directly into
5054 llmount, removing the need to use this or specifying the
5055 client NID explicitly when mounting GM clients with zeroconf
5058 Frequency : if client is started with down MDS
5060 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5061 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5062 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5063 ptlrpc_import_delayed_req() because we didn't check if the
5064 request was interrupted, and we also didn't break out of the
5065 event loop if there was no timeout
5070 Description: data loss during non-page-aligned writes to a single file from
5071 both multiple nodes and multiple threads on one node at same time
5072 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5073 inconsistency led to false short-reads, that were cached and later
5074 used by ->prepare_write() to fill in partially written page,
5075 leading to data loss.
5078 Frequency : always, if lconf --abort_recovery used
5080 Description: lconf --abort_recovery fails with 'Operation not supported'
5081 Details : lconf was attempting to abort recovery on the MDT device and not
5084 Severity : enhancement
5086 Description: remove cleanup logs
5087 Details : replace lconf-generated cleanup logs with lustre internal
5088 cleanup routines. Eliminates the need for client-cleanup and
5091 Severity : enhancement
5093 Description: add support for EAs (user and system) on lustre filesystems
5094 Details : it is now possible to store extended attributes in the Lustre
5095 client filesystem, and with the user_xattr mount option it
5096 is possible to allow users to store EAs on their files also
5098 Severity : enhancement
5100 Description: Add possibility (config option) to show minimal available OST free
5102 Details : When compiled with --enable-mindf configure option, statfs(2)
5103 (and so, df) will return least minimal free space available from
5104 all OSTs as amount of free space on FS, instead of summary of
5105 free spaces of all OSTs.
5107 Severity : enhancement
5109 Description: do not expand extent locks acquired on OST-side
5110 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5111 by server: they are not cached anyway.
5114 Frequency : when mmap is used/binaries executed from Lustre
5116 Description: Unmmap pages before throwing them away from read cache.
5117 Details : llap_shrink cache now attempts to unmap pages before discarding
5118 them (if unmapping failed - do not discard). SLES9 kernel has
5119 extra checks that trigger if this unmapping is not done first.
5124 Description: lconf didn't resolve symlinks before checking to see whether a
5125 given mountpoint was already in use
5128 Frequency : when migrating failover services
5129 Bugzilla : 6395, 9514
5130 Description: When migrating a subset of services from a node (e.g. failback
5131 from a failover service node) the remaining services would
5132 time out and evict clients.
5133 Details : lconf --force (implied by --failover) sets the global obd_timeout
5134 to 5 seconds in order to quickly disconnect, but this caused
5135 other RPCs to time out too quickly. Do not change the global
5136 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5138 Severity : enhancement
5139 Frequency : if MDS is started with down OST
5140 Bugzilla : 9439,5706
5141 Description: Allow startup/shutdown of an MDS without depending on the
5142 availability of the OSTs.
5143 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5144 Add appropriate locking and lov-osc refcounts for safe
5145 cleaning. Add osc abort_inflight calls in case the
5146 synchronize never started.
5149 Frequency : occasional (Cray XT3 only)
5151 Description: root not authorized to access files in CRAY_PORTALS environment
5152 Details : The client process capabilities were not honoured on the MDS in
5153 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5154 been accessed by an authorized user then root was able to access
5155 the file on the local client also. The root user capabilities
5156 are now allowed on the MDS, as this environment has secure UID.
5159 Frequency : occasional
5161 Description: ldiskfs "too long searching" message happens too often
5162 Details : A debugging message (otherwise harmless) prints too often on
5163 the OST console. This has been reduced to only happen when
5164 there are fragmentation problems on the filesystem.
5169 Description: Division by zero in statfs when all OSCs are inactive
5170 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5171 lov_statfs divides by value returned by lov_get_stripecnt().
5175 Bugzilla : 9489, 3273
5176 Description: First write from each client to each OST was only 4kB in size,
5177 to initialize client writeback cache, which caused sub-optimal
5178 RPCs and poor layout on disk for the first written file.
5179 Details : Clients now request an initial cache grant at (re)connect time
5180 and so that they can start streaming writes to the cache right
5181 away and always do full-sized RPCs if there is enough data.
5182 If the OST is rebooted the client also re-establishes its grant
5183 so that client cached writes will be honoured under the grant.
5188 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5189 Details : Now I/O RPCs go to different portal number and (presumably) fast
5190 lock requests (and glimses) and other RPCs get their own service
5191 threads pool that should be able to service those RPCs
5194 Severity : enhancement
5196 Description: Ability to exchange lustre version between client and servers and
5197 issue warnings at client side if client is too old. Also for
5198 liblustre clients there is ability to refuse connection of too old
5200 Details : New 'version' field is added to connect data structure that is
5201 filled with version info. That info is later checked by server and
5205 Frequency : rare, liblustre only.
5206 Bugzilla : 9296, 9581
5207 Description: Two simultaneous writes from liblustre at offset within same page
5208 might proceed at the same time overwriting eachother with stale
5210 Details : I/O lock withing llu_file_prwv was released too early, before data
5211 actually was hitting the wire. Extended lock-holding time until
5212 server acknowledges receiving data.
5215 Frequency : extremely rare. Never observed in practice.
5217 Description: avoid generating lustre_handle cookie of 0.
5218 Details : class_handle_hash() generates handle cookies by incrementing
5219 global counter, and can hit 0 occasionaly (this is unlikely, but
5220 not impossible, because initial value of cookie counter is
5221 selected randonly). Value of 0 is used as a sentinel meaning
5222 "unassigned handle" --- avoid it. Also coalesce two critical
5223 sections in this function into one.
5225 Severity : enhancement
5227 Description: allow liblustre clients to delegate truncate locking to OST
5228 Details : To avoid overhead of locking, liblustre client instructs OST to
5229 take extent lock in ost_punch() on client's behalf. New connection
5230 flag is added to handle backward compatibility.
5232 Severity : enhancement
5233 Bugzilla : 4928, 7341, 9758
5234 Description: allow number of OST service threads to be specified
5235 Details : a module parameter allows the number of OST service threads
5236 to be specified via "options ost ost_num_threads={N}" in the
5237 OSS's /etc/modules.conf or /etc/modprobe.conf.
5241 Bugzilla : 6146, 9635, 9895
5242 Description: servers crash with bad pointer in target_handle_connect()
5243 Details : In rare cases when a client is reconnecting it was possible that
5244 the connection request was the last reference for that export.
5245 We would temporarily drop the export reference and get a new
5246 one, but this may have been the last reference and the export
5247 was just destroyed. Get new reference before dropping old one.
5249 Severity : enhancement
5250 Frequency : if client is started with failover MDS
5252 Description: Allow multiple MDS hostnames in the mount command
5253 Details : Try to read the configuration from all specified MDS
5254 hostnames during a client mount in case the "primary"
5257 Severity : enhancement
5259 Description: Stop sending data to evicted clients as soon as possible.
5260 Details : Check if the client we are about to send or are sending data to
5261 was evicted already. (Check is done every second of waiting,
5262 for which l_wait_event interface was extended to allow checking
5263 of exit condition at specified intervals).
5266 Frequency : rare, normally only when NFS exporting is done from client
5268 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5270 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5273 Frequency : always, if srand() is called before liblustre initialization
5275 Description: Liblustre uses system PRNG disturbing its usage by user application
5276 Details : Introduce internal to lustre fast and high-quality PRNG for
5277 lustre usage and make liblustre and some other places in generic
5278 lustre code to use it.
5280 Severity : enhancement
5281 Bugzilla : 9477, 9557, 9870
5282 Description: Verify that the MDS configuration logs are updated when xml is
5283 Details : Check if the .xml configuration logs are newer than the config
5284 logs stored on the MDS and report an error if this is the case.
5285 Request --write-conf, or allow starting with --old_conf.
5287 Severity : enhancement
5289 Description: Handle symlinks in the path when checking if Lustre is mounted.
5290 Details : Resolve intermediate symlinks when checking if a client has
5291 mounted a filesystem to avoid duplicate client mounts.
5296 Description: lconf can hit an error exception but still return success.
5297 Details : The lconf command catches the Command error exception at the top
5298 level script context and will exit with the associated exit
5299 status, but doesn't ensure that this exit status is non-zero.
5304 Description: failure of ptlrpc thread startup can cause oops
5305 Details : Starting a ptlrpc service thread can fail if there are a large
5306 number of threads or the server memory is very fragmented.
5307 Handle this without oopsing.
5310 Frequency : always, only if liblustre and non-default acceptor port was used
5312 Description: liblustre cannot connect to servers with non-default acceptor port
5313 Details : tcpnal_set_default_params() was not called and was therefore
5314 ignoring the environment varaible TCPNAL_PORT, as well as other
5315 TCPNAL_ environment variables
5320 Description: two objects could be created on the same OST for a single file
5321 Details : If an OST is down, in some cases it was possible to create two
5322 objects on a single OST for a single file. No problems other
5323 than potential performance impact and spurious error messages.
5327 Bugzilla : 5681, 9562
5328 Description: Client may oops in ll_unhash_aliases
5329 Details : Client dcache may become inconsistent in race condition.
5330 In some cases "getcwd" can fail if the current directory is
5336 Description: Inode refcounting problems in NFS export code
5337 Details : link_raw functions used to call d_instantiate without obtaining
5338 extra inode reference first.
5342 Bugzilla : 9942, 9903
5343 Description: Referencing freed requests leading to crash, memleaks with NFS.
5344 Details : We used to require that call to ll_revalidate_it was always
5345 followed by ll_lookup_it. Also with revalidate_special() it is
5346 possible to call ll_revalidate_it() twice for the same dentry
5347 even if first occurence returned success. This fix changes semantic
5348 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5349 reference on a request referred from the intent.
5350 ll_intent_release() then releases such a request.
5353 Frequency : rare, normally benchmark loads only
5355 Description: unlinked inodes were kept in memory on the client
5356 Details : If a client is repeatedly creating and unlinking files it
5357 can accumulate a lot of stale inodes in the inode slab cache.
5358 If there is no other client load running this can cause the
5359 client node to run out of memory. Instead flush old inodes
5360 from client cache that have the same inode number as a new inode.
5363 Frequency : SLES9 2.6.5 kernel and long filenames only
5364 Bugzilla : 9969, 10379
5365 Description: utime reports stale NFS file handle
5366 Details : SLES9 uses out-of-dentry names in some cases, which confused
5367 the lustre dentry revalidation. Change it to always use the
5371 Frequency : rare, unless heavy write-truncate concurrency is continuous
5372 Bugzilla : 4180, 6984, 7171, 9963, 9331
5373 Description: OST becomes very slow and/or deadlocked during object unlink
5374 Details : filter_destroy() was holding onto the parent directory lock
5375 while truncating+unlinking objects. For very large objects this
5376 may block other threads for a long time and slow overall OST
5377 responsiveness. It may also be possible to get a lock ordering
5378 deadlock in this case, or run out of journal credits because of
5379 the combined truncate+unlink. Solution is to do object truncate
5380 first in one transaction without parent lock, and then do the
5381 final unlink in a new transaction with the parent lock. This
5382 reduces the lock hold time dramatically.
5385 Frequency : rare, 2.4 kernels only
5387 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5388 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5389 before unmounting their filesystem in order to avoid deadlock.
5390 The kernel_locked() function in 2.4 kernels only checks whether
5391 the kernel lock is held, not whether it is this process that is
5392 holding it as 2.6 kernels do.
5397 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5398 Details : The client ptlrpc code may be trying to reconnect to a down
5399 server before a previous connection attempt has timed out.
5400 Increase the reconnect interval to be longer than the connection
5401 timeout interval to avoid sending duplicate connections to
5405 Frequency : echo_client brw_test command
5407 Description: fix echo_client to work with OST preallocated code
5408 Details : OST preallocation code (5137) didn't take echo_client IO path
5409 into account: echo_client calls filter methods outside of any
5410 OST thread and, hence, there is no per-thread preallocated
5411 pages and buffers to use. Solution: hijack pga pages for IO. As
5412 a byproduct, this avoids unnecessary data copying.
5416 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5417 Description: Client can oops in mdc_commit_close() after open replay
5418 Details : It was possible for the MDS to return an open request with no
5419 transaction number in mds_finish_transno() if the client was
5420 evicted, but without actually returning an error. Clients
5421 would later try to replay that open and may trip an assertion
5422 Simplify the client close codepath, and always return an error
5423 from the MDS in case the open is not successful.
5426 Frequency : rare, 2.6 OSTs only
5428 Description: OST may deadlock under high load on fragmented files
5429 Details : If there was a heavy load and highly-fragmented OST filesystems
5430 it was possible to have all the OST threads deadlock waiting on
5431 allocation of biovecs, because the biovecs were not released
5432 until the entire RPC IO was completed. Instead, release biovecs
5433 as soon as they are complete to ensure forward IO progress.
5435 Severity : enhancement
5437 Description: Support for specifying external journal device at mount
5438 Details : If an OST or MDS device is formatted with an external journal
5439 device, this device major/minor is stored in the ext3 superblock
5440 and may not be valid for failover. Allow detecting and
5441 specifying the external journal at mount time.
5446 Description: Mounting an MDS with pending unlinked files may cause oops
5447 Details : target_finish_recovery() calls mds_postrecov() which returned
5448 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5449 considers this an error and immediately begins cleaning up the
5450 lov, just after starting the mds_lov process
5452 Severity : enhancement
5454 Description: Implement 'lfs df' to report actual free space on per-OST basis
5455 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5456 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5457 usage of objects; '-h' to report in human readable format.
5459 ------------------------------------------------------------------------------
5461 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5468 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5469 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5470 at startup during orphan unlinking. Instead, skip buddy update
5471 before setup, as it will be regenerated after recovery is complete.
5476 Description: If an OST is inactive, its locks might reference stale inodes.
5477 Details : lov_change_cbdata() must iterate over all namespaces, even if
5478 they are inactive to clear inode references from the lock.
5480 Severity : enhancement
5481 Frequency : occasional, if non-standard max_dirty_mb used
5483 Description: Client will block write RPCs if not enough grant
5484 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5485 then the client will block writes while waiting for another RPC
5486 to complete instead of consuming its dirty limit. With change
5487 we get improved performance when max_dirty_mb is small.
5489 Severity : enhancement
5490 Bugzilla : 3389, 6253
5491 Description: Add support for supplementary groups on the MDS.
5492 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5493 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5494 lookups for user supplementary groups into a cache.
5498 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5499 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5500 for liblustre O_EXCL is enforced by MDS.
5504 Description: readdir never returns NULL in liblustre.
5505 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5506 dentry in struct dirent.
5510 Frequency : liblustre only, depends on application IO pattern
5511 Description: liblustre clients evicted if not contacting servers
5512 Details : Don't put liblustre clients into the ping_evictor list, so
5513 they will not be evicted by the pinger ever.
5515 Severity : enhancement
5517 Description: Add ability to evict clients by NID from MDS.
5518 Details : By echoing "nid:$NID" string into
5519 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5520 $NID would be instantly evicted from this MDS and from all active
5521 OSTs connected to it.
5525 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5526 Details : lookup_it_finish() used to query file size from OSTs that was not
5531 Description: service threads change working directory to that of init
5532 Details : Starting lustre service threads may pin the working directory
5533 of the parent thread, making that filesystem busy. Threads
5534 now change to the working directory of init to avoid this.
5538 Frequency : during shutdown only
5539 Description: shutdown with a failed MDS or OST can cause unmount to hang
5540 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5545 Frequency : 2.6 only
5546 Description: chmod/chown may include an extra supplementary group
5547 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5548 supplementary group and if none is specified this is used.
5551 Bugzilla : 5479 (6816)
5553 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5554 Details : If lookup is in progress on a file that is unlinked we might try
5555 to revalidate the inode and fail in revalidate after lookup is
5556 complete and ll_file_open() enqueues the open again but
5557 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5560 Frequency : always, if lconf --abort_recovery used
5562 Description: lconf --abort_recovery fails with 'Operation not supported'
5563 Details : lconf was attempting to abort recovery on the MDT device and not
5566 ------------------------------------------------------------------------------
5568 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5573 Frequency : rare (only unsupported configurations with a node running as an
5575 Bugzilla : 6514, 5137
5576 Description: Mounting a Lustre file system on a node running as an OST could
5578 Details : OSTs now preallocates memory needed to write out data at
5579 startup, instead of when needed, to avoid having to
5580 allocate memory in possibly low memory situations.
5581 Specifically, if the file system is mounted on on OST,
5582 memory pressure could force it to try to write out data,
5583 which it needed to allocate memory to do. Due to the low
5584 memory, it would be unable to do so and the node would
5585 become unresponsive.
5587 Severity : enhancement
5589 Description: Addition of lconf --service command line option
5590 Details : lconf now accepts a '--service <arg>' option, which is
5591 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5593 Severity : enhancement
5595 Description: Failover mode is now the default for OSTs.
5596 Details : By default, OSTs will now run in failover mode. To return to
5597 the old behaviour, add '--failout' to the lmc line for OSTs.
5599 Severity : enhancement
5601 Description: Health checks are now provided for MDS and OSTs
5602 Details : Additional detailed health check information on MSD and OSTs
5603 is now provided through the procfs health_check value.
5606 Frequency : occasional, depends on IO load
5608 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5609 after numerous create/delete cycles
5610 Details : The ext3 inode allocation policy would not allocate new inodes
5611 very well on the OSTs because there are no new directories
5612 being created. Instead we look for groups with free space if
5613 the parent directories are nearly full.
5618 Description: Network or server problems during mount may cause partially
5619 mounted clients instead of returning an error.
5620 Details : The config llog parsing code may overwrite the error return
5621 code during mount error handling, returning success instead
5627 Description: MDS can fail to allocate large reply buffers
5628 Details : After long uptimes the MDS can fail to allocate large reply
5629 buffers (e.g. zconf client mount config records) due to memory
5630 fragmentation or consumption by the buffer cache. Preallocate
5631 some large reply buffers so that these replies can be sent even
5632 under memory pressure.
5636 Frequency : rare (liblustre)
5637 Description: fsx running with liblustre complained that using truncate() to
5638 extend the file doesn't work. This patch corrects that issue.
5639 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5640 ATTR_SIZE and lsm use in llu_setattr_raw.
5644 Frequency : rare, only 2.6 kernels
5645 Description: Unusual file access patterns on the MDS may result in inode
5646 data being lost in very rare circumstances.
5647 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5648 code in the 2.6 kernel caused the inode and/or EA data not to
5649 be read from disk, causing single-file corruption.
5653 Frequency : rare, only 2.6 filesystems using extents
5654 Description: Heavy concurrent write and delete load may cause data corruption.
5655 Details : It was possible under high-load situations to have an extent
5656 metadata block in the block device cache from a just-unlinked
5657 file overwrite a newly-allocated data block. We now unmap any
5658 metadata buffers that alias just-allocated data blocks.
5662 Frequency : filesystems with default stripe_count larger than 77
5663 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5664 Details : lconf specifies an inode size of 4096 bytes when the default
5665 stripe_count is larger than 77. This conflicts with the default
5666 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5667 this case to avoid pinning too much memory for large EAs.
5669 ------------------------------------------------------------------------------
5671 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5676 Frequency : rare (extremely heavy IO load with hundreds of clients)
5678 Description: Client is evicted, gets IO error writing to file
5679 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5680 caused clients to be evicted for AST timeouts. The fixes in
5681 bug 5192 mean we no longer need to have such short AST timeouts
5682 so ldlm_timeout has been increased.
5685 Frequency : occasional during --force or --failover shutdown under load
5686 Bugzilla : 5949, 4834
5687 Description: Server oops/LBUG if stopped with --force or --failover under load
5688 Details : a collection of import/export refcount and cleanup ordering
5689 issues fixed for safer force cleanup
5692 Frequency : only filesystems larger than 120 OSTs
5693 Bugzilla : 5990, 6223
5694 Description: lfs getstripe would oops on a very large filesystem
5695 Details : lov_getconfig used kfree on vmalloc'd memory
5698 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5699 Bugzilla : 6242, 6243
5700 Description: reading from files that had been truncated to a non-zero size
5701 but never opened returned no data
5702 Details : ll_file_read() reads zeros from no-object files to EOF
5707 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5708 Details : There's little harm in aborting MDS/OSS recovery and letting it
5709 try again, so I removed the LASSERT and return an error instead.
5711 Severity : enhancement
5713 Description: New debugging infrastructure for tracking down data corruption
5714 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5715 (b) cover more of the client-side code path, and (c) try to narrow
5716 down where problems occurred
5720 Bugzilla : 3819, 4364, 4397, 6313
5721 Description: Racing close and eviction MDS could cause assertion in mds_close
5722 Details : It was possible to get multiple mfd references during close and
5723 client eviction, leading to one thread referencing a freed mfd.
5725 Severity: : enhancement
5726 Bugzilla : 3262, 6359
5727 Description: Attempts to reconnect to servers are now more aggressive.
5728 Details : This builds on the enhanced upcall-less recovery that was added
5729 in 1.4.2. When trying to reconnect to servers, clients will
5730 now try each server in the failover group every 10 seconds. By
5731 default, clients would previously try one server every 25 seconds.
5736 Description: After recovery, certain operations trigger a failed
5737 assertion on a client.
5738 Details : Failing over an mds, using lconf -d --failover, while a
5739 client was doing a readdir() call would cause the client to
5740 LBUG after recovery completed and the readdir() was resent.
5742 Severity : enhancement
5744 Description: Default groups are now added by lconf
5745 Details : You can now run lconf --group <servicename> without having to
5746 manually add groups with lmc.
5749 Frequency : occasional
5751 Description: Nodes with an elan id of 0 trigger a failed assertion
5754 Frequency : always when accessing e.g. tty/console device nodes
5756 Description: tty and some other devices nodes cannot be used on lustre
5757 Details : file's private_data field is used by device data and lustre
5758 values in there got lost. New field was added to struct file to
5759 store fs-specific private data.
5762 Frequency : when exporting Lustre via NFS
5764 Description: NFSD failed occasionally when looking up a path component
5765 Details : NFSD is looking up ".." which was broken in ext3 directories
5766 that had grown large enough to become hashed.
5769 Frequency : Clusters with multiple interfaces not on the same subnet
5771 Description: Nodes will repeatedly try to reconnect to an interface which it
5772 cannot reach and report an error to the log.
5773 Details : Extra peer list entries will be created by lconf with some peers
5774 unreachable. lconf now validates the peer before adding it.
5777 Frequency : Only if a default stripe is set on the filesystem root.
5779 Description: Setting a default stripe on the filesystem root prevented the
5780 filesystem from being remounted.
5781 Details : The client was sending extra request flags in the root getattr
5782 request and did not allocate a reply buffer for the dir EA.
5785 Frequency : occasional, higher if lots of files are accessed by one client
5786 Bugzilla : 6159, 6097
5787 Description: Client trips assertion regarding lsm mismatch/magic
5788 Details : While revalidating inodes the VFS looks up inodes with ifind()
5789 and in rare cases can find an inode that is being freed.
5790 The ll_test_inode() code will free the lsm during ifind()
5791 when it finds an existing inode and then the VFS later attaches
5792 this free lsm to a new inode.
5796 Bugzilla : 6422, 7030
5797 Description: MDS deadlock between mkdir and client eviction
5798 Details : Creating a new file via mkdir or mknod (starting a transaction
5799 and getting the ns lock) can deadlock with client eviction
5800 (gets ns lock and trying to finish a synchronous transaction).
5803 Frequency : occasional
5804 Description: While starting a server, the fsfilt_ext3 module could not be
5806 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5807 kernels. Previously, lconf would still use the ext3 name
5808 when trying to load modules. Now, it will correctly use
5809 ext3 on 2.4 and ldiskfs on 2.6.
5811 Severity : enhancement
5812 Description: The default stripe count has been changed to 1
5813 Details : The interpretation of the default stripe count (0, to lfs
5814 or lmc) has been changed to mean striping across a single
5815 OST, rather than all available. For general usage we have
5816 found a stripe count of 1 or 2 works best.
5818 Severity : enhancement
5819 Description: Add support for compiling against Cray portals.
5820 Details : Conditional compiling for some areas that are different
5824 Frequency : occasional
5825 Bugzilla : 6409, 6834
5826 Description: Creating files with an explicit stripe count may lead to
5827 a failed assertion on the MDS
5828 Details : If some OSTs are full or unavailable, creating files may
5829 trigger a failed assertion on the MDS. Now, Lustre will
5830 try to use other servers or return an error to the
5834 Frequency : occasional
5836 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5837 caused lock timeout during readahead (since 1.4.2).
5838 Details : Processes doing readahead might match a lock that hasn't been
5839 granted yet if there are overlapping and conflicting lock
5840 requests. The readahead process waits on ungranted lock
5841 (original lock is CBPENDING), while OST waits for that process
5842 to cancel CBPENDING read lock and eventually evicts client.
5844 Severity : enhancement
5846 Description: Initial enabling of flock support for clients
5847 Details : Implements fcntl advisory locking and file status functions.
5848 This feature is provided as an optional mount flag (default
5849 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5850 locking are implemented yet, and those that are are not guaranteed
5851 to be completely correct in production environments.
5852 mount -t lustre -o [flock|noflock] ...
5855 Frequency : occasional
5857 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5858 assertion in the ext3 JBD (journaling) layer.
5859 Details : The b_committed_data struct is protected by the big kernel lock
5860 in 2.4 kernels, serializing journal_commit_transaction() and
5861 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5862 access to this struct. In 2.6 kernels there is finer grained
5863 locking to improve SMP performance of the JBD layer.
5867 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5872 Description: merge and cleanup kernel patches.
5875 -----------------------------------------------------------------------------
5877 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5879 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5880 versions of Lustre. Please update all servers and clients to
5881 version 1.4.2 or later at the same time. You must also run
5882 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5883 to update the configuration logs.
5885 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5886 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5887 - swab llog records of type '0' so we get proper header size/idx (5861)
5888 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5889 - fix rename of one directory over another leaking an inode (5953)
5890 - avoid SetPageDirty on 2.6 (5981)
5891 - don't re-add just-being-destroyed locks to the waiting list (5653)
5892 - when creating new directories, inherit the parent's custom
5893 striping settings if present parent (3048)
5894 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5895 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5896 - don't hold client locks on temporary worklist from l_lru (5666)
5897 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5898 - automatically evict dead clients (5921)
5899 - Update file size properly in create+truncate+fstat case (6196)
5900 - Do not unhash mountpoint dentries, do not allow removal of
5902 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5903 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5904 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5905 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5906 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5907 - Correctly handle reads of files with no objects (6243)
5908 - lctl recover will also mark a device active if deactivate used (5933)
5910 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5911 - allow --write-conf on an MDS with different nettype than client (5619)
5912 - don't write config llogs to MDS for mounts not from that MDS (5617)
5913 - lconf should create multiple TCP connections from a client (5201)
5914 - init scripts are now turned off by default; run chkconfig --on
5915 lustre and chkconfig --on lustrefs to use them
5916 - upcalls are no longer needed for clients to recover to failover
5918 - add --abort-recovery option to lconf to abort recovery on device
5920 - add support for an arbitrary number of OSTs (3026)
5921 - Quota support protocol changes.
5922 - forward compatibility changes to wire structs (6007)
5923 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5924 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5925 - improved "lustre status" script
5926 - initialize blocksize for non-regular files (6062)
5927 - added --disable-server and --disable-client configure options (5782)
5928 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5929 - Vanilla 2.4.29 support
5930 - increase maximum number of obd devices to 520 (6242)
5931 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5932 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5933 SLES 9 2.6.5 kernels
5934 - limit stripes per file to 160 (the maximum EA size) (6093)
5936 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5939 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5940 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5941 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5942 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5943 - small ext3 extents cleanups and fixes (5733)
5944 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5945 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5946 - avoid needless client->OST connect, fix handle mismatch (5317)
5947 - fix DLM error path that led to out-of-sync client, long delays (5779)
5948 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5949 - fix several locking issues related to i_size (5492,5624,5654,5672)
5950 - don't move pending lock onto export if it is already evicted (5683)
5951 - fix kernel oops when creating .foo in unlinked directory (5548)
5952 - fix deadlock in obdfilter statistics vs. object create (5811)
5953 - use time_{before,after} to avoid timer jiffies wrap (5882)
5954 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5955 - Do not leak request if server was not able to process it (5154)
5956 - If mds_open unable to find parent dir, make that negative lookup(5154)
5957 - don't create new directories with extent-mapping (5909, 5936)
5959 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5960 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5961 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5962 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5963 - support for RHEL4 kernel on i686 (5773)
5964 - provide error messages when incompatible logs are encountered (5898)
5966 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5967 * version 1.4.0.10 (1.4.1 release candidate 1)
5969 - don't keep a lock reference when lock is not granted (4238)
5970 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5971 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5972 - drop import inflight refcount on signal_completed_replay error (5255)
5973 - unlock page after async write error during send (3677)
5974 - handle missing objects in filter_preprw_read properly (5265)
5975 - no transno return for symlink open, don't save no-trasno open (3440)
5976 - don't try to complete elan receive that already failed (4012)
5977 - free RPC server reply state on error (5406)
5978 - clean up thread from ptlrpc_start_thread() on error (5160)
5979 - readahead could read extra page into cache that wasn't ejected (5388)
5980 - prevent races in class_attach/setup/cleanup/detach (5260)
5981 - don't dereference de->d_inode after l_dput of de (5458)
5982 - use "int" for stripe value returned from lock_to_stripe (5544)
5983 - mballoc allocation and error-checking fixes in 2.6 (5504)
5984 - block device patches to fix I/O request sizes in 2.6 (5482)
5985 - look up hostnames for IB nals (5602)
5986 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5987 - don't start multiple acceptors for the same port (5277)
5988 - fix incorrect LASSERT in mds_getattr_name (5635)
5989 - export a proc file for general "ping" checking (5628)
5990 - fix "lfs check" to not block when the MDS is down (5628)
5992 - service request history (4965)
5993 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5994 - create an "evict_client" /proc entry on OSTs, like the MDS has
5995 - fix mount usage message, return errors per mount(8) (5168)
5996 - change grep [] to grep "[]" in tests so they work in more UMLs
5997 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5998 - remove ancient LOV_MAGIC_V0 EA support (5047)
5999 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6000 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6001 - allow readahead allocations to fail when low on memory (5383)
6002 - mmap locking landed again, after considerable improvement (2828)
6003 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6005 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6008 - send OST transaction number in read/write reply to free req (4966)
6009 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6010 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6011 - the watchdog thread now runs as interruptible (5246)
6012 - flock/lockf fixes (but it's still disabled, pending 5135)
6013 - don't use EXT3 constants in llite code (5094)
6014 - memory shortage at startup could cause assertion (5176)
6016 - reorganization of lov code
6017 - single portals codebase
6019 - add extents/mballoc support (5025)
6020 - direct I/O reads in the obdfilter (4048)
6021 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6023 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6026 - send OST transaction number in read/write reply to free req (4966)
6027 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6028 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6029 - the watchdog thread now runs as interruptible (5246)
6030 - handle missing objects in filter_preprw_read properly (5265)
6031 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6032 - drop import inflight refcount on signal_completed_replay error (5255)
6033 - unlock page after async write error during send (3677)
6034 - return original error code on reconstructed replies (3761)
6035 - no transno return for symlink open, don't save no-trasno open (3440)
6037 - add pid to ldlm debugging output (4922)
6038 - bump the watchdog timeouts -- we can't handle 30sec yet
6039 - extra debugging for orphan dentry/inode bug (5259)
6041 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6044 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6045 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6046 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6047 - fix NULL dereference in /proc/sys/portals/routes (4827)
6048 - allow failed mdc_close() operations to be interrupted (4561)
6049 - stop precreate on OST before MDS would time out on it (4778)
6050 - don't send partial-page writes before EOF from client (4410)
6051 - discard client grant for sub-page writes on large-page clients (4520)
6052 - don't free dentries not owned by NFS code, check generation (4806)
6053 - fix lsm leak if mds_create_objects() fails (4801)
6054 - limit debug_daemon file size, always print CERROR messages (4789)
6055 - use transno after validating reply (3892)
6056 - process timed out requests if import state changes (3754)
6057 - update mtime on OST during writes, return in glimpse (4829)
6058 - add mkfsoptions to LDAP (4679)
6059 - use ->max_readahead method instead of zapping global ra (5039)
6060 - don't interrupt __l_wait_event() during strace
6062 - add software watchdogs to catch hung threads quickly (4941)
6063 - make lustrefs init script start after nfs is mounted
6064 - fix CWARN/ERROR duplication (4930)
6065 - return async write errors to application if possible (2248)
6066 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6067 - print NAL number in %x format (4645)
6068 - update barely-supported suse-2.4.21-171 series (4842)
6069 - support for sles 9 %post scripts
6070 - support for building 2.6 kernel-source packages
6071 - support for sles km_* packages
6073 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6076 - ignore -ENOENT errors in osc_destroy (3639)
6077 - notify osc create thread that OSC is being cleaned up (4600)
6078 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6079 - reconstruct ost_handle() like mds_handle() (4657)
6080 - create a new thread to do import eviction to avoid deadlock (3969)
6081 - let lconf resolve symlinked-to devices (4629)
6082 - don't unlink "objects" from directory with default EA (4554)
6083 - hold socknal file ref over connect in case target is down (4394)
6084 - allow more than 32000 subdirectories in a single directory (3244)
6085 - fix blocks count for O_DIRECT writes (3751)
6086 - OST returns ENOSPC from object create when no space left (4539)
6087 - don't send truncate RPC if file size isn't changing (4410)
6088 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6089 - bind to privileged port in socknal and tcpnal (3689)
6091 - rate limit CERROR/CWARN console message to avoid overload (4519)
6092 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6093 - basic mmap support (3918)
6094 - kernel patch series update from b1_4 (4711)
6096 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6099 - avoid crash during MDS cleanup with OST shut down (2775)
6100 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6101 - don't use bad inodes on the MDS (3744)
6102 - dynamic object preallocation to improve recovery speed (4236)
6103 - don't hold spinlock over lock dumping or change debug flags (4401)
6104 - don't zero obd_dev when it is force cleaned (3651)
6105 - print grants to console if they go negative (4431)
6106 - "lctl deactivate" will stop automatic recovery attempts (3406)
6107 - look for existing locks in ldlm_handle_enqueue() (3764)
6108 - don't resolve lock handle twice in recovery avoiding race (4401)
6109 - revalidate should check working dir is a directory (4134)
6111 - don't always mark "slow" obdfilter messages as errors (4418)
6113 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6116 - don't close LustreDB during write_conf until it is done (3860)
6117 - fix typo in lconf for_each_profile (3821)
6118 - allow dumping logs from multiple threads at one time (3820)
6119 - don't allow multiple threads in OSC recovery (3812)
6120 - fix debug_size parameters (3864)
6121 - fix mds_postrecov to initialize import for llog ctxt (3121)
6122 - replace config semaphore with spinlock (3306)
6123 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6124 - don't allow enqueue to complete on a destroyed export (3822)
6125 - down write_lock before checking llog header bitmap (3825)
6126 - recover from lock replay timeout (3764)
6127 - up llog sem before sending rpc (3652)
6128 - reduce ns lock hold times when setting kms (3267)
6129 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6130 - fix NULL deref and obd_dev leak on setup error (3312)
6131 - replace some LBUG about llog ops with error handling (3841)
6132 - don't match INVALID dentries from d_lookup and spin (3784)
6133 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6134 - fix invalid assertion in ptlrpc_set_wait (3880)
6136 - add libwrap support for the TCP acceptor (3996)
6137 - add /proc/sys/portals/routes for non-root route listing (3994)
6138 - allow setting MDS UUID in .xml (2580)
6139 - print the stack of a process that LBUGs (4228)
6141 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6144 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6145 - make sure to unset replay flag from failed open requests (3440)
6146 - if default stripe count is 0, use OST count for inode size (3636)
6147 - update parent mtime/ctime on client for create/unlink (2611)
6148 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6149 - free recovery state on server during a forced cleanup (3571)
6150 - unregister_reply for resent reqs (3063)
6151 - loop back devices mounting and status check on 2.6 (3563)
6152 - fix resource-creation race that can provoke i_size == 0 (3513)
6153 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6154 - more debugging for page-accounting assertion (3746)
6155 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6156 - avoid deadlock after precreation failure (3758)
6157 - fix race and lock order deadlock in orphan handling (3450, 3750)
6158 - add validity checks when grabbing inodes from l_ast_data (3599)
6160 - add /proc/.../recovery_status to obdfilter (3428)
6161 - lightweight CDEBUG infrastructure, debug daemon (3668)
6162 - change default OSC RPC parameters to be better on small clusters
6163 - turn off OST read cache for files smaller than 32MB
6164 - install man pages and include them in rpms (3100)
6165 - add new init script for (un)mounting lustre filesystems (2593)
6166 - run chkconfig in %post for init scripts (3701)
6167 - drop scimac NAL (unmaintained)
6169 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6172 - clean kiobufs before and after use (3485)
6173 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6174 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6175 - change init script to fail more clearly if not run as root (1528)
6176 - allow clients to reconnect during replay (1742)
6177 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6178 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6179 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6180 - fix cond_resched (3554)
6181 - only evict unfinished clients after recovery (3515)
6182 - allow bulk resend, prevent data loss (3570)
6183 - dynamic ptlrpc request buffer allocation (2102)
6184 - don't allow unlinking open directory if it isn't empty (2904)
6185 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6186 - remove extraneous obd dereference causing LASSERT failure (3334)
6187 - don't use get_cycles() when creating temp. files on the mds (3156)
6188 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6189 - handle EEXIST for set-stripe, set proper directory name (3336)
6191 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6192 - fix ksocknal_fmb_callback() error messages (2918)
6194 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6197 - don't copy lvb into (possibly NULL) reply on error (2983)
6198 - don't deref dentry after dput, don't free lvb on error (2922)
6199 - use the kms to determine writeback rpc length (2947)
6200 - increment oti_logcookies when osc is inactive (2948)
6201 - update client's i_blocks count via lvb messages (2543)
6202 - handle intent open/close of special files properly (1557)
6203 - mount MDS with errors=remount-ro, like obdfilter (2009)
6204 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6205 - don't use cancelling-locks' kms values (2947)
6206 - use highest lock extent for kms, not last one (2925)
6207 - don't dereference ERR_PTR() dentry in error handling path (3107)
6208 - fix thread race in portals_debug_dumplog() (3122)
6209 - create lprocfs device entries at setup instead of at attach (1519)
6210 - common AST error handler, don't evict client on completion race (3145)
6211 - zero nameidata in detach_mnt in 2.6 (3118)
6212 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6213 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6214 - initialize RPC timeout timer earlier for 2.6 (3219)
6215 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6216 - print nal/nid for unknown nid (3258)
6217 - additional checks for oscc recovery before doing precreate (3284)
6218 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6219 - don't crash in mdc_close for bad permissions on open (3285)
6220 - zero i_rdev for non-device files (3147)
6221 - clear page->private before handing to FS, better assertion (3119)
6222 - tune the read pipeline (3236)
6223 - fix incorrect decref of invalidated dentry (2350)
6224 - provide read-ahead stats and refine rpc in flight stats (3328)
6225 - don't hold journal transaction open across create RPC (3313)
6226 - update atime on MDS at close time (3265)
6227 - close LDAP connection when recovering to avoid server load (3315)
6228 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6229 - don't leak open file on MDS after open resend (3325)
6230 - serialize filter_precreate and filter_destroy_precreated (3329)
6231 - loop device shouldn't call sync_dev() for nul device (3092)
6232 - clear page cache after eviction (2766)
6233 - resynchronize MDS->OST in background (2824)
6234 - refuse to mount the same filesystem twice on same mountpoint (3394)
6235 - allow llmount to create routes for mounting behind routers (3320)
6236 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6237 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6238 - fix rare problem with rename on htree directories (3417)
6240 - allow default OST striping configuration per directory (1414)
6241 - fix compilation for qswnal for 2.6 kernels (3125)
6242 - increase maximum number of MDS request buffers for large systems
6243 - change liblustreapi to be useful for external progs like lfsck (3098)
6244 - increase local configuration timeout for slow disks (3353)
6245 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6247 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6250 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6251 - don't overwrite extent policy data in reply if lock was blocked (2901)
6252 - drop filter export grants atomically with removal from device (2663)
6253 - del obd_self_export from work_list in class_disconnect_exports (2908)
6254 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6255 - swab reply message in mdc_close, other PPC fixes (2464)
6256 - fix destroying of named logs (2325)
6257 - overwrite old logs when running lconf --write_conf (2264)
6258 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6259 - fix race in target_handle_connect (2898)
6260 - mds_reint_create() should take same inode create lock (2926)
6261 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6262 - don't close files for self_export to avoid uninitialized obd (2936)
6263 - allow MDS with the same name as client node (2939)
6264 - hold dentry reference for closed log files for unlink (2325)
6265 - reserve space for all logs during transactions (2059)
6266 - don't evict page beyond end of stripe extent (2925)
6267 - don't oops on a deleted current working directory (2399)
6268 - handle hard links to targets without a parent properly (2517)
6269 - don't dereference NULL lock when racing during eviction (2867)
6270 - don't grow lock extents when lots of conflicting locks (2919)
6272 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6275 - account for cache space usage on clients to avoid data loss (974)
6276 - lfsck support in lustre kernel code (2349)
6277 - reduce journal credits needed for BRW writes (2370)
6278 - orphan handling to avoid losing space on client/server crashes
6279 - ptlrpcd can be blocked, stopping ALL progress (2477)
6280 - use lock value blocks to assist in proper KMS, faster stat (1021)
6281 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6282 - recovery for initial connections (2355)
6283 - fixes for mds_cleanup_orphans (1934)
6284 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6285 - block all file creations until orphan recovery completes (1901)
6286 - client remove rq_connection from request struct (2423)
6287 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6288 - recovery timer race (2670)
6289 - mdc_close recovey bug (2532)
6290 - ptlrpc cleanup bug (2710)
6291 - mds timeout on local locks (2588)
6292 - namespace lock held during RPCs (2431)
6293 - handle interrupted sync write properly (2503)
6294 - don't try to handle a message that hasn't been replied to (2699)
6295 - client assert failure during cleanup after abort recovery (2701)
6296 - leak mdc device after failed mount (2712)
6297 - ptlrpc_check_set allows timedout requests to complete (2714)
6298 - wait for inflight reqs when ptlrpcd finishes (2710)
6299 - make sure unregistered services are removed from the srv_list
6300 - reset bulk XID's when resending them (caught by 1138 test)
6301 - unregister_bulk after timeout
6302 - fix lconf error (2694)
6303 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6304 - readahead locks pages, leaves pending causing memory pressure (2673)
6305 - increase OST request buffers to 4096 on large machines (2729)
6306 - fix up permission of existing directories in simple_mkdir (2661)
6307 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6308 - don't assign transno to errored transactions (2742)
6309 - don't delete objects on OST if given a bogus objid from MDS (2751)
6310 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6311 - if rq_no_resend, then timeout request after recovery (2432)
6312 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6313 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6314 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6315 - MDS recovery shouldn't delete objects that it has given out (2730)
6316 - if enqueue arrives after completion, don't clobber LVB (2819)
6317 - don't unlock pages twice when trigger_group_io returns error (2814)
6318 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6319 - don't write pages to disk if there was an error (1450)
6320 - don't ping imports that have recovery disabled (2676)
6321 - take buffered bytes into account when balancing socknal conn (2817)
6322 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6323 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6324 - remove little-endian swabbing of llog records (1987)
6325 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6326 - retry reposting request buffers if they fail (1191)
6327 - grow extent at grant time to avoid granting a revoked lock (2809)
6328 - lock revoke doesn't evict page if covered by a second lock (2765)
6329 - disable VM readahead to avoid reading outside lock extents (2805)
6331 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6332 - updated kernel patches for hp-2.4.20 kernel (2681)
6334 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6337 - fix truncated write corruption (2366)
6338 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6340 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6341 - ensure len(uuid) < 37 in lmc (1171)
6342 - fix ia64 OOPS in llog_test (2255)
6343 - zero end of page at obdfilter for partial page writes (2648)
6344 - don't leave stale dentries around after renames (bug 2428)
6345 - fix timeouts when evicting a client with a single lock held (2642)
6346 - set deadline for the initial HELLO message to drain (2634)
6347 - print out dotted-quad IP addresses in the socknal (2302)
6349 - additional debugging for MDS client eviction problem (2443)
6350 - fix mkfsoptions support for osts (2603, 2604)
6352 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6355 - add series for the vanilla 2.6.0 kernel
6356 - add series for the vanilla 2.4.24 kernel
6357 - add series for a cray x86/64 UL kernel drop
6358 - fix xattr patches for the vanilla 2.4.19 series
6360 - generate true UUIDs in lmc (1171)
6361 - have portals stack dumping break in UML (2466)
6362 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6363 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6364 - wake callback waiting threads on client eviction (2460)
6365 - Add --ptldebug and --subsystem to lmc (1719)
6366 - update assertion to allow safe interrupt allocation
6367 - set rq_no_resend for cancel requests (2432)
6368 - recalculate ptlrpcd timeout after resend (2494)
6369 - call vfs_rmdir when removing pending directories (2368)
6370 - fix renaming a file to itself (2429)
6371 - lmc creates a default one-stripe lov (2454)
6372 - expand procfs space to handle large clusters (2326)
6373 - increase UML stack to avoid overflow
6374 - update lconf's list of debug and subsystem masks
6375 - fix lfs find --obd (2510)
6376 - /proc tunable for disabling filter read caching (2591)
6377 - stop rpm packages from altering slapd.conf (2301)
6378 - disable nagle in the socknal under 0conf (2578)
6379 - choose mds inode size based on stripe count (2572)
6380 - fix kernel-source rpm problems (2516)
6382 - add --disable-doc to avoid pdf generation (2421)
6383 - update documentation, tests, type-os, comments
6384 - avoid format warnings on ia64
6385 - remove the TOE NAL
6386 - tiny code cleanups by removing unused fields
6388 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6391 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6392 - remove the most blatant lies from BUILDING file (2371)
6393 - change default debug level to reasonable production setting
6394 - reduce client side cache size to reduce cache flush time
6395 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6396 - make TCP zerocopy and pinger support enabled by default (2476)
6397 - sync writes completed after process exits caused crashes (2319)
6398 - maintain correct mount count on the MDS (2356)
6399 - backout 1557, because 2316 wasn't really fixed
6400 - better file I/O statistics gathering in /proc
6401 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6402 - another kernel patch to fix zero-copy TCP function export
6403 - don't take duplicate lock when processing re-sent getattr (2420)
6404 - lctl uses obd_self_export instead of creating new conn (2353)
6405 - MDS/OST recovery case which requires object creation asserted (2425)
6406 - move lfs from /usr/sbin to /usr/bin in packages
6407 - fix race between mds_client_add and mds_client_free (2417)
6408 - use kmalloc instead of slabs in portals (2430)
6409 - don't create duplicate records when a failover MDS is present (2442)
6410 - remove unnecessary mount age check (2332)
6411 - don't remove directory inodes from locks prematurely (2451)
6412 - don't break if MDS service name is the same as hostname (2103)
6413 - fix races in client write RPC generation when cache full (2482)
6415 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6418 - remove now-unused request->rq_obd (278)
6419 - if an allocation fails, print out how much memory we've used (1933)
6420 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6421 - add the "configurable stack size" patch to most series files (1256)
6422 - ability to write large log records, for 100+ OST configs (2306)
6423 - fix NULL deref when filter_prep fails (2314)
6424 - fix operator precedence error in filter_sync
6425 - dynamic allocation of socknal TX descriptors (2315)
6426 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6427 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6428 - ensure that CWARN messages in Portals always get to the syslog
6429 - __init/__exit are not for prototype decls (ldlm_init/exit)
6430 - x86-64 compile warning fixes
6431 - fix gateway LMC keyword conflict (2318)
6432 - fix MDS lock inversions in getattr/reint paths (1844)
6433 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6434 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6435 - DLM race condition prevented some lock evictions (2328)
6436 - ENOMEM detection and retry on socknal sends (2230)
6437 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6438 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6439 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6440 - better error messages when a client is rejected during recovery (1505)
6441 - avoid cancelling locks which were never granted, after failure (2330)
6442 - fix i_sem/journal inversion in mds_client_add (2333)
6443 - fix truncate/getattr lock cycle deadlock (2334)
6444 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6445 - fix two rare exit paths which could leak an l_lock() ref (2321)
6446 - fencepost error in MDS/OST orphan recovery (2226)
6447 - make log record alignment 8 bytes (1988)
6448 - lstripe now fails when requested offset > ost_count (2237)
6449 - ensure that all kernel series have a complete list.h (1607)
6450 - fix crashes in special-file operations (2316)
6451 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6453 - allow configurable automake binary, for testing new versions
6454 - small update to the lfs documentation
6456 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6458 * fix negative export reference count in fsfilt_sync (2312)
6460 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6461 * release candidate 0.9.1
6463 - orphans are moved into the PENDING directory for possible recovery
6464 - replayed opens now open by fid for orphan/rename safety (1042)
6465 - last close of an orphan inode generates a transno (683)
6466 - chdir() and mount() now pin the directory entry (1020)
6467 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6468 - discard very old requests without processing them (1502)
6469 - remove some common, well-understood CERRORs (1505)
6470 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6471 - clear "grant" flags in OST replies until OST grant code lands (1644)
6472 - fix read performance by not clobbering i_blksize on client (1598)
6473 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6474 - make LRU size a /proc tunable, clears locks when reduced (707)
6475 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6476 - server locks take references on exports now (1558)
6477 - build fixes for 2.4.20-rh trees (1663)
6478 - return an error from lov_create if all OSCs are inactive (1751)
6479 - fix import levels when a reconnect happens without a timeout (1597)
6480 - exit early from mds_open if we get a lookup error (1749)
6481 - partial page read at EOF wouldn't wait for disk before sending (1642)
6482 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6483 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6484 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6485 - write/truncate lock inversion (1639)
6486 - Don't auto-load obdclass, portals modules during cleanup (1495)
6487 - fix timestamps from jumping to "now" (1763)
6488 - extra journal assertions (1648)
6489 - add an extra multiunlink test (1771)
6490 - fix read_record/write_record API (1776)
6491 - fix leak of offset_extent, possible incorrect i_size later (1772)
6492 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6493 - Debugging for the kqswnal_get_idle_tx problems (1820)
6494 - Allow recovery to be attempted multiple times (1536)
6495 - Write out MDS last_rcvd file after it is first created (1600)
6496 - Fix tx_descriptor leak in failed transmit situations (1827)
6497 - ext3 journaling fixes for assertion failure after IO error (1871)
6498 - class_export_put() on freed export after completion AST error (1896)
6499 - Fix revalidate looping in VFS (1322)
6500 - Don't access a freed export during MDS_REINT timeout (1521)
6501 - Add open-unlink recovery support on the MDS (1673,1764)
6502 - Return an error if no MDS data was read from last_rcvd (1946)
6503 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6504 - Don't setup a disk device that doesn't match exported UUID (317)
6505 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6506 - avoid committing NULL handle in force close
6507 - local.sh is now a one-stripe LOV configuration
6508 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6509 - fix ext3 htree duplicate directory entry corruption (1516)
6510 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6511 - update to new LOV EA format (2097)
6512 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6514 2003-06-15 Phil Schwan <phil@clusterfs.com>
6517 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6518 - per-import/export recovery handling (958, 931, 959)
6519 - multiple last-rcvd slots, for serving multiple FSes (949)
6520 - connections are again shared between multiple imp/exports (963, 964)
6521 - "umount -f" would hang if any requests needed to be sent (393, 978)
6522 - avoid pinning large req buffer by copying for queued messages (989)
6523 - add "uuid" to "lctl device" command to help upcalls (991)
6524 - "open" RPCs with transnos would confuse recovery counters (1037)
6525 - do proper endian conversion of all wire messages (288, 340, 891)
6526 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6527 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6528 - fix ptlrpc_connection leak in target_handle_connect (1174)
6529 - fix import refcounting bug in OST and MDS cleanup (1134)
6530 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6531 - fix very unlikely obd_types race condition (501)
6532 - remove osc_open hack for echo_client (1187)
6533 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6534 - a failure in read_inode2 leads to deadlock (1139)
6535 - cancel ack-locks as soon as transaction is committed (1072)
6536 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6537 - make sure to commitrw after any preprw to avoid deadlock (1162)
6538 - failing to execute a file in a lustre FS would lock inode (1203)
6539 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6540 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6541 - obd_uuid2tgt would walk off the end of the list (1255)
6542 - on IA64 the osc would give portals incorrect bulk size (1258)
6543 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6544 - fix lock inversion caused by new llite matching code (1282)
6545 - limit the number of dirty pages on a client to 10MB (1286)
6546 - timed out locks were not being corrected cancelled (1289)
6547 - fix O_DIRECT above 4GB on IA-32 (1292)
6548 * major user-visible changes
6549 - fail out/fail over policy now controlled by the upcall (993)
6551 - add OBD_PING to check server availability and failure (954)
6552 - lustre messages are now sent in sending host order (288, 340, 891)
6553 - add eadatalen to MDS getattr reply (340)
6554 - OST read replies may contain second buffer, with per-page status (593)
6556 2003-03-11 Phil Schwan <phil@clusterfs.com>
6559 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6560 - failed open()s could cause deadlock; fixed (867, 869)
6561 - stop cancelling OST locks when files are closed (481)
6562 - overlapping XID spaces caused network corruption (851, 853)
6563 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6564 - setattr_raw added, to do single-RPC, server-side setattrs
6565 - lmc/lconf syntax change for OST UUIDs
6566 - fix crashy race condition between ptlrpc_free_req and osc_close
6567 - don't use request in mdc_enqueue if we hit a timeout (889)
6568 - don't set the inode i_size for regular files from the MDS (896)
6569 - handle out of order completion AST (842)
6570 - don't LBUG if a lock request times out after receiving AST (913)
6571 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6572 - if a bad lock AST arrives, send an error instead of dropping entirely
6573 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6574 - fix leak in bulk IO when only partially completed (899, 900, 926)
6575 - fix O_DIRECT for ia64 (55)
6576 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6577 - C-z after timeout could hang a process forever; fixed (977)
6579 - client-side I/O cache (678, 924, 929, 941, 970)
6581 - READPAGE and SETATTRs which don't take server-side locks get
6584 2003-02-11 Phil Schwan <phil@clusterfs.com>
6587 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6588 - Fix fsfilt_extN_readpage() to read a full page of directory
6589 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6590 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6591 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6592 - Most intent operations were removed, in favour of a new RPC mode
6593 that does a single RPC to the server and bypasses most of the VFS
6594 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6595 - Aggressively cancel local locks on DLM servers
6596 - mds_reint_unlink sends EA to the client if it's the last nlink.
6597 client uses that EA to unlink OST objects.
6598 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6599 - recursive symlinks were fixed (439)
6600 - fixed NULL deref in DEBUG_REQ
6601 - filter_update_lastobjid no longer calls sync, which annoyed extN
6602 - fixed multi-client small-writes to a single file problem (445)
6603 - fixed mtime updates during file writes (607)
6604 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6605 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6606 - fixed Linux OST size reporting problem (444, 656)
6607 - OST now updates object mtime with writes or setattr (607, 619)
6608 - client verifies file size before zeroing page past EOF (445)
6609 - OST now writes last allocated objid to disk with allocation (108)
6610 - LOV on echo now works (409)
6612 - mds_reint_unlink sends a new buffer, with the EA included. this
6613 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6614 set if a regular file was being unlinked, and it was the last link
6615 - use PtlGet from the target for bulk writes (315)
6616 - OST now updates object mtime with writes or setattr (607, 619)
6617 - LDLM now has a grant-time callback to revalidate locked items, if
6619 - Many MDS operations were reorganized to combat race conditions
6621 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6622 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6624 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6627 - Fully reactivate OST imports after reconnection (512, others)
6628 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6629 - More graceful error handling for truncating on dead OST (515)
6630 - Don't error out unless we're actually accessing dead stripes (474)
6631 - Fix garbage sizes when stripes are missing (410)
6632 - LRU counters were broken, causing constant lock purge (433, 432)
6633 - garbage on read from stripes with failed OSTs (441)
6634 - mark OSCs as active before reconnecting during recovery (438)
6635 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6636 - lfind did not preserve OST order in output (443)
6637 - symlinks cause hung clients, incorrect data (439)
6638 - stop dereferencing request after dropping refcount (457)
6639 - don't LASSERT(spin_is_locked) on non-SMP (455)
6640 - fixes for many rename() bugs
6641 - fstat didn't correctly synchronize attributes (399)
6642 - server must handle lock cancellation during blocking AST prep (487)
6643 - bulk descriptors were free()d too soon (511)
6644 - fix paths in lconf, which would load incorrect modules (451, 507)
6645 - fix confusing lconf 'host not found' error message (386)
6646 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6647 - fix race condition in mdc_blocking_ast() for inode access (526)
6648 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6649 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6650 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6651 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6652 - fix lstripe in several ways (kernel side) (527)
6653 - fix request leak in ldlm_cli_enqueue (262)
6654 - incorrect OSC was marked inactive after OST failure
6655 - call mds_fs_cleanup before unmounting filesystem (524)
6656 - fix races between taking ns_lock and ldlm_lock_change_resource
6657 - fix races updating LOV export open file list
6658 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6659 - fix recovery NULL deref in ldlm_cli_cancel_unused
6660 - fix some DLM races by using new hash table for lock handles (419)
6661 - permit the client to specify desired inodes, at replay
6662 - duplicate requests when we queue them for replay reintegration
6663 - fix last_rcvd offset calculation
6664 - sync after each recovered transaction, so we always make progress
6665 - never, not always, ERESTART requests without transnos
6666 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6668 - skip replay if the MDS says that the client is already connected
6669 - don't check for a recovery-enabled export to match lctl's UUID
6670 - don't INC_USE_COUNT for phantom exports
6671 - don't crash when cleaning up phantom exports (567)
6672 - don't double-finish or set replay data for errored mdc_open requests
6673 - abort requests when they time out, so we don't get old replies
6674 - send/receive replies for AST messages again
6675 - if the client says that it doesn't have the lock, cancel it on the
6677 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6678 mark it as destroyed, it will all work out when decref is called
6679 - fix module use counts (22, 581)
6681 - ASTs now expect a reply (server cancels lock on error reply)
6683 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6686 - fix many simultaneous client startup (392)
6687 - fix dentry->d_it clobbering
6688 - credentials weren't being shipped for readdir/getattr operations
6689 - remove invalid assertions triggered during some concurrent MD
6691 - proper Lustre versions added (336, 389)
6692 - fix memory leak for create error case (398)
6693 - fix LOV locking bug that would get cli/srv out of sync
6694 - fix echo client over LOV (409)
6695 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6696 - fix double-O_EXCL intent crash (424)
6697 - avoid sending multiple lock CANCELs (352)
6699 - MDS can do multi-client recovery (modulo bugs in new code)
6701 - many updates, edits, cleanups
6703 2002-11-18 Phil Schwan <phil@clusterfs.com>
6706 - fix null d_it dereference (346)
6707 - fix full OST/dbench hang (333)
6708 - fix permission problem with file removal (286)
6709 - fix removal of OSCs from LOV when they fail
6710 - fix NULL deref during bulk timeout (214)
6711 - fix problems related to multiple filesystems on one MDS (241)
6712 - fixed serious subtle metadata locking bugs
6713 - free locks on clients when inodes are removed due to memory
6715 - fix inode pointer in lock data (285)
6716 - partial support for multiple MDS on a single host (241)
6717 - data locks weren't cancelled at clear_inode time (290, 311)
6718 - intent locks could lead to unbounded lock growth (205)
6719 - added a maximum lock count, an LRU list, and a flusher
6720 - fix multiple rename (365)
6721 - properly abstracted the echo client
6722 - OSC locked 1 byte too many; fixed
6723 - rewrote brw callback code:
6724 - fixed recovery bugs related to LOVs (306)
6725 - fixed too-many-pages-in-one-write crash (191)
6726 - fixed (again) crash in sync_io_timeout (214)
6727 - probably fixed callback-related race (385)
6729 - Add capability to MDS protocol
6730 - LDLM cancellations and callbacks on different portals
6732 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6735 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6736 - defer open object destruction to close time (601981, 138)
6737 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6738 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6739 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6740 - don't call obd_setattr in ll_file_release for destroyed objects
6741 * protocol change to lustre_msg: move |version| and add |flags|
6742 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6743 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6744 * added replay of create, unlink, link and rename operations during
6745 MDS failover; recovery should be much more robust now
6746 * remove failed OSCs from LOVs (only lov_create uses this so far)
6747 * the lustre-HOWTO was brought (more) up to date (582544)
6749 2002-10-23 Phil Schwan <phil@clusterfs.com>
6752 - in-use dentries weren't being reused properly (617851)
6753 - prevent multiple LDLM setup (599178)
6754 - fix LOV size calculations for truncate (617853)
6755 - fix client handling of MDS intent errors (POSIX)
6756 - fix permission bug in lovstripe.c test (624321)
6757 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6758 - truncate past end of file could corrupt data
6759 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6760 - a race in recovery could return ETIMEDOUT to apps (623947)
6761 - building outside the source directory was fixed
6762 * the lustre-HOWTO was brought (more) up to date (582544)
6763 * major progress was made on recovery functionality
6765 2002-10-10 Phil Schwan <phil@clusterfs.com>
6768 - recovery deadlock fix
6769 - rm -rf causes LBUG fix (617817)
6770 - file open by multiple tasks fix (618962)
6771 - directory permissions bugs (602707 and 620007)
6772 - journal_stop fixed with locking (611313)
6773 - O_APPEND failures resolved (618273, perhaps 614459)
6774 - lconf PATH fix (619770)
6775 - IA64 build fix (621450)
6776 - RPC buffer sizes scale with amount of memory
6778 2002-10-01 Phil Schwan <phil@clusterfs.com>
6781 - locks would be cancelled without throwing away data pages,
6782 resulting in inconsistent data (605627)
6783 - inode attributes were not always being refreshed (605627, 612449)
6784 - lconf now continues to cleanup after lctl reports an error
6785 - MDS now enforces user permissions (602707)
6786 - lprocfs cleanup fixed, but not yet enabled (614157)
6787 - fixed infinite server hang, should a client not respond to an AST
6788 - avoid going into recovery if user calls readlink() with a buffer
6789 that's too small (613941)
6790 - AST RPCs no longer require replies (614867) -- this may be changed
6791 - don't crash server if client sends an IOV that's too big (611336)
6792 - fixed lock conversion deadlock (611892)
6793 - fixed the following of symlinks (614622)
6794 * recovery: the server can remove locks from a client that dies, other
6795 clients can make progress
6796 * more extN patch fixes
6797 * compile-time configurable ptlrpc buffer allocations
6799 - collaborative read cache document
6800 - Lustre Lite Performance CDR document-in-progress
6802 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6805 - fix typo in patch-2.4.18
6807 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6810 - clear ptlrpc request each time in handle_incoming_request()
6811 - unlink of files now destroys the object on the OST
6813 2002-09-19 Peter Braam <braam@clusterfs.com>
6815 * add hard link support
6816 * change obdfile creation method
6817 * kernel patch changed
6819 2002-09-19 Peter Braam <braam@clusterfs.com>
6822 - stack overflow bug in extN fixed
6824 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6826 * documentation updates
6827 - add man pages for config tools
6828 - update tests/README to describe testing with new config tools
6829 - finish metadata API descriptions
6830 * bug fixes and cleanups
6831 - statfs workaround for 16TB limit
6832 - LOV stripe allocation improved, can stripe on subset of OSTs
6833 - LOV file size/IO offset was wrong for files > 4GB in size
6834 - object EA data was being dropped, caused files to be unreadable
6835 - memory overflow with non-LOV OST caused memory corruption
6836 - fixed regression tests to work with new config tools, obdfilter
6837 - fixed bug when directory size became larger than 1 block
6838 - fixed bug (for single client case) when PWD was deleted
6839 - invalidate local directory pages when doing intent-based ops
6840 - avoid LDLM oops when lock callback contained bad data
6842 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6844 * documentation updates
6845 * bug fixes and cleanups
6846 - configuration tools
6849 - 64-bit compile warnings
6850 - 64-bit internal statfs data
6852 * test_brw on persistent OST devices
6854 * lprocfs (disabled)
6856 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6858 * documentation updates
6859 * bug fixes and cleanups
6860 * configuration tools
6862 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6865 * many small fixes to 0_5_4
6866 * io/network handling
6867 * thinkos in MDS operations
6869 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6872 * crucial basic fixes to 0.5.3
6873 * IOR, Iozone work over Elan
6876 2002-08-07 Phil Schwan <phil@clusterfs.com>
6877 * version 0_5_3, our first alpha
6878 * we use the new Portals iovs
6879 * documentation updates
6880 * bug fixes and cleanups
6881 * small changes in the DLM wire protocol
6883 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6884 * version 0_5_1 with some initial stability,
6885 * locking on MD and file I/O.
6886 * documentation updates
6887 * several bug fixes since 0.5.0
6888 * small changes in wire protocol
6890 2002-07-18 Phil Schwan <phil@clusterfs.com>
6892 * delivered as Lustre Light Alpha
6893 * fixed a crash after handling invalid MDS requests
6894 * fixed directory pages for architectures with non-4k pages sizes
6896 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6897 * release version v0_4_4
6898 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6900 * Fixes a number of interruption problems with OST operations.
6901 * Update documentation for portals header changes
6902 * Move all wire protocol structs/defines to lustre_idl.h
6903 * Fixes symlink length bug.
6904 * Add tcpdump to repository.
6906 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6907 * release version v0_4_3
6908 * Fixes statfs for inodes on extN.
6909 * Fixes bug in runtests which would delete /etc/hosts.
6910 * Use 64-bit object IDs wherever possible (not into VFS though)
6911 Remove ost_get_info, which is unused by lustre, and out of date.
6913 2002-07-03 Peter Braam <braam@clusterfs.com>
6914 * release version v0_4_2 Fixes a lookup error (type not passed)
6915 * move forward to head of Portals
6916 * move forward to latest Lustre kernel
6918 2002-06-25 Peter Braam <braam@clusterfs.com>
6919 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob