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 5.14.0-284.30.1.el9 (RHEL9.2)
7 5.14.0-162.23.1.el9 (RHEL9.1)
8 4.18.0-513.5.1.el8 (RHEL8.9)
9 4.18.0-477.27.1.el8 (RHEL8.8)
10 * Other server kernels known to build and work at some point (others may also work):
11 3.10.0-862.14.4.el7 (RHEL7.5)
12 3.10.0-957.27.2.el7 (RHEL7.6)
13 3.10.0-1062.18.1.el7 (RHEL7.7)
14 3.10.0-1127.19.1.el7 (RHEL7.8)
15 3.10.0-1160.88.1.el7 (RHEL7.9)
16 4.18.0-80.11.2.el8 (RHEL8.0)
17 4.18.0-147.8.1.el8 (RHEL8.1)
18 4.18.0-193.28.1.el8 (RHEL8.2)
19 4.18.0-240.22.1.el8 (RHEL8.3)
20 4.18.0-305.25.1.el8 (RHEL8.4)
21 4.18.0-348.23.1.el8 (RHEL8.5)
22 4.18.0-372.32.1.el8 (RHEL8.6)
23 4.18.0-425.10.1.el8 (RHEL8.7)
24 4.4.120-92.70 (SLES12 SP2)
25 4.4.180-94.100 (SLES12 SP3)
26 4.4.0-131 (Ubuntu 16.04)
27 4.15.0-32 (Ubuntu 18.04)
28 5.4.0-48 (Ubuntu 20.04)
29 vanilla linux 5.4.0 (ZFS + ldiskfs)
30 vanilla linux 5.4.21 (ZFS + ldiskfs)
31 vanilla linux 5.4.136 (ZFS + ldiskfs)
32 5.10.0-60.94.0.118.oe2203 (openEuler 22.03 LTS)
33 5.10.0-136.32.0.108.oe2203sp1 (openEuler 22.03 LTS SP1)
34 5.10.0-153.19.0.95.oe2203sp2 (openEuler 22.03 LTS SP2)
35 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
36 * Client primary kernels built and tested during release cycle:
37 5.14.0-362.8.1.el9 (RHEL9.3)
38 5.14.0-284.30.1.el9 (RHEL9.2)
39 4.18.0-513.5.1.el8 (RHEL8.9)
40 4.18.0-477.27.1.el8 (RHEL8.8)
41 5.4.0-37 (Ubuntu 20.04)
42 5.14.21-150500.55.36 (SLES15 SP5)
43 * Other clients known to build on these kernels at some point (others may also work):
44 3.10.0-862.14.4.el7 (RHEL7.5)
45 3.10.0-957.27.2.el7 (RHEL7.6)
46 4.14.0-49.13.1.el7a (RHEL7.5)
47 4.14.0-115.2.2.el7a (RHEL7.6)
48 3.10.0-1062.18.1.el7 (RHEL7.7)
49 3.10.0-1127.19.1.el7 (RHEL7.8)
50 3.10.0-1160.88.1.el7 (RHEL7.9)
51 4.18.0-80.11.2.el8 (RHEL8.0)
52 4.18.0-147.8.1.el8 (RHEL8.1)
53 4.18.0-193.28.1.el8 (RHEL8.2)
54 4.18.0-240.22.1.el8 (RHEL8.3)
55 4.18.0-305.25.1.el8 (RHEL8.4)
56 4.18.0-348.23.1.el8 (RHEL8.5)
57 4.18.0-372.32.1.el8 (RHEL8.6)
58 4.18.0-425.10.1.el8 (RHEL8.7)
59 5.14.0-70.30.1.el9 (RHEL9.0)
60 5.14.0-162.23.1.el9 (RHEL9.1)
61 4.4.120-92.70 (SLES12 SP2)
62 4.4.180-94.100 (SLES12 SP3)
63 4.12.14-95.48 (SLES12 SP4)
64 4.12.14-122.133 (SLES12 SP5)
65 4.12.14-197.75 (SLES15 SP1)
66 5.3.18-24.96 (SLES15 SP2)
67 5.3.18-150300.59.93 (SLES15 SP3)
68 5.14.21-150400.24.92 (SLES15 SP4)
69 4.4.0-131 (Ubuntu 16.04)
70 4.15.0-48 (Ubuntu 18.04)
71 5.8.0-53 (Ubuntu 20.04.2 HWE)
72 5.11.0-31 (Ubuntu 20.04.3 HWE)
73 5.11.0 (vanilla kernel.org)
74 5.10.0 (openEuler 22.03 LTS)
75 * Recommended e2fsprogs version: 1.47.0-wc5 or newer
76 * Recommended ZFS version: 2.1.11
77 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
78 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
79 stack overflow. For more information, please refer to bugzilla 17630.
80 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
81 the re-exporting nodes to be mounted with "32bitapi" mount option
83 --------------------------------------------------------------------------------
88 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
89 for currently supported client and server kernel versions.
90 * Server primary kernels built and tested during release cycle:
91 4.18.0-348.2.1.el8 (RHEL8.5)
92 * Other server kernels known to build and work at some point (others may also work):
93 3.10.0-862.14.4.el7 (RHEL7.5)
94 3.10.0-957.27.2.el7 (RHEL7.6)
95 3.10.0-1062.18.1.el7 (RHEL7.7)
96 3.10.0-1127.19.1.el7 (RHEL7.8)
97 3.10.0-1160.49.1.el7 (RHEL7.9)
98 4.18.0-80.11.2.el8 (RHEL8.0)
99 4.18.0-147.8.1.el8 (RHEL8.1)
100 4.18.0-193.28.1.el8 (RHEL8.2)
101 4.18.0-240.22.1.el8 (RHEL8.3)
102 4.18.0-305.25.1.el8 (RHEL8.4)
103 4.4.120-92.70 (SLES12 SP2)
104 4.4.180-94.100 (SLES12 SP3)
105 4.4.0-131 (Ubuntu 16.04)
106 4.15.0-32 (Ubuntu 18.04)
107 5.4.0-48 (Ubuntu 20.04)
108 vanilla linux 5.4.0 (ZFS + ldiskfs)
109 vanilla linux 5.4.21 (ZFS + ldiskfs)
110 vanilla linux 5.4.136 (ZFS + ldiskfs)
111 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
112 * Client primary kernels built and tested during release cycle:
113 4.18.0-348.2.1.el8 (RHEL8.5)
114 5.4.0-37 (Ubuntu 20.04)
115 5.3.18-59.27 (SLES15 SP3)
116 * Other clients known to build on these kernels at some point (others may also work):
117 3.10.0-862.14.4.el7 (RHEL7.5)
118 3.10.0-957.27.2.el7 (RHEL7.6)
119 4.14.0-49.13.1.el7a (RHEL7.5)
120 4.14.0-115.2.2.el7a (RHEL7.6)
121 3.10.0-1062.18.1.el7 (RHEL7.7)
122 3.10.0-1127.19.1.el7 (RHEL7.8)
123 3.10.0-1160.49.1.el7 (RHEL7.9)
124 4.18.0-80.11.2.el8 (RHEL8.0)
125 4.18.0-147.8.1.el8 (RHEL8.1)
126 4.18.0-193.28.1.el8 (RHEL8.2)
127 4.18.0-240.22.1.el8 (RHEL8.3)
128 4.18.0-305.25.1.el8 (RHEL8.4)
129 4.4.120-92.70 (SLES12 SP2)
130 4.4.180-94.100 (SLES12 SP3)
131 4.12.14-95.48 (SLES12 SP4)
132 4.12.14-122.91 (SLES12 SP5)
133 4.12.14-197.75 (SLES15 SP1)
134 5.3.18-24.96 (SLES15 SP2)
135 4.4.0-131 (Ubuntu 16.04)
136 4.15.0-48 (Ubuntu 18.04)
137 5.8.0-53 (Ubuntu 20.04.2 HWE)
138 5.11.0-31 (Ubuntu 20.04.3 HWE)
139 5.11.0 (vanilla kernel.org)
140 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
141 * Recommended ZFS version: 2.0.7
142 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
143 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
144 stack overflow. For more information, please refer to bugzilla 17630.
145 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
146 the re-exporting nodes to be mounted with "32bitapi" mount option
148 --------------------------------------------------------------------------------
152 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
153 for currently supported client and server kernel versions.
154 * Server primary kernels built and tested during release cycle:
155 3.10.0-1062.18.1.el7 (RHEL7.7)
156 3.10.0-1127.19.1.el7 (RHEL7.8)
157 3.10.0-1160.6.1.el7 (RHEL7.9)
158 * Other server kernels known to build and work at some point (others may also work):
159 3.10.0-862.14.4.el7 (RHEL7.5)
160 3.10.0-957.27.2.el7 (RHEL7.6)
161 4.18.0-80.11.2.el8 (RHEL8.0)
162 4.18.0-147.8.1.el8 (RHEL8.1)
163 4.18.0-193.28.1.el8 (RHEL8.2)
164 4.18.0-240.1.1.el8 (RHEL8.3)
165 4.4.120-92.70 (SLES12 SP2)
166 4.4.180-94.100 (SLES12 SP3)
167 4.4.0-131 (Ubuntu 16.04)
168 4.15.0-32 (Ubuntu 18.04)
169 5.4.0-48 (Ubuntu 20.04)
170 vanilla linux 5.4.0 (ZFS + ldiskfs)
171 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
172 * Client primary kernels built and tested during release cycle:
173 3.10.0-1062.18.1.el7 (RHEL7.7)
174 3.10.0-1127.19.1.el7 (RHEL7.8)
175 3.10.0-1160.6.1.el7 (RHEL7.9)
176 4.12.14-95.48 (SLES12 SP4)
177 4.12.14-122.20 (SLES12 SP5)
178 4.12.14-197.75 (SLES15 SP1)
179 5.3.18-24.24 (SLES15 SP2)
180 4.15.0-48 (Ubuntu 18.04)
181 * Other clients known to build on these kernels at some point (others may also work):
182 3.10.0-862.14.4.el7 (RHEL7.5)
183 3.10.0-957.27.2.el7 (RHEL7.6)
184 4.14.0-49.13.1.el7a (RHEL7.5)
185 4.14.0-115.2.2.el7a (RHEL7.6)
186 4.18.0-80.11.2.el8 (RHEL8.0)
187 4.18.0-147.8.1.el8 (RHEL8.1)
188 4.18.0-193.28.1.el8 (RHEL8.2)
189 4.18.0-240.1.1.el8 (RHEL8.3)
190 4.4.120-92.70 (SLES12 SP2)
191 4.4.180-94.100 (SLES12 SP3)
192 4.4.0-131 (Ubuntu 16.04)
193 5.4.0-37 (Ubuntu 20.04)
194 5.4.0 (vanilla kernel.org)
195 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
196 * Recommended ZFS version: 2.0.0
197 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
198 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
199 stack overflow. For more information, please refer to bugzilla 17630.
200 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
201 the re-exporting nodes to be mounted with "32bitapi" mount option
203 --------------------------------------------------------------------------------
207 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
208 for currently supported client and server kernel versions.
209 * Server primary kernels built and tested during release cycle:
210 3.10.0-1062.1.1.el7 (RHEL7.7)
211 * Other server kernels known to build and work at some point (others may also work):
212 3.10.0-862.14.4.el7 (RHEL7.5)
213 3.10.0-957.27.2.el7 (RHEL7.6)
214 4.18.0-80.7.1.el8 (RHEL8.0)
215 4.4.120-92.70 (SLES12 SP2)
216 4.4.180-94.100 (SLES12 SP3)
217 4.4.0-131 (Ubuntu 16.04)
218 4.15.0-32 (Ubuntu 18.04)
219 vanilla linux 4.6.7 (ZFS only)
220 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
221 * Client primary unpatched kernels built and tested during release cycle:
222 3.10.0-1062.1.1.el7 (RHEL7.7)
223 4.12.14-95.29 (SLES12 SP4)
224 4.15.0-48 (Ubuntu 18.04)
225 * Other clients known to build on unpatched kernels at some point (others may also work):
226 3.10.0-862.14.4.el7 (RHEL7.5)
227 3.10.0-957.27.2.el7 (RHEL7.6)
228 4.14.0-49.13.1.el7a (RHEL7.5)
229 4.14.0-115.2.2.el7a (RHEL7.6)
230 4.18.0-80.7.1.el8 (RHEL8.0)
231 4.4.120-92.70 (SLES12 SP2)
232 4.4.180-94.100 (SLES12 SP3)
233 4.4.0-131 (Ubuntu 16.04)
234 vanilla linux 4.15.0 (ZFS only)
235 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
236 * Recommended ZFS / SPL version: 0.7.13
237 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
238 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
239 stack overflow. For more information, please refer to bugzilla 17630.
240 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
241 the re-exporting nodes to be mounted with "32bitapi" mount option
242 * Note for Ubuntu 14.04.5:
243 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
244 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
246 --------------------------------------------------------------------------------
250 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
251 for currently supported client and server kernel versions.
252 * Server known to build on the following kernels (others may also work):
253 2.6.32-696.18.7.el6 (RHEL6.9)
254 3.10.0-862.14.4.el7 (RHEL7.5)
255 3.10.0-957.el7 (RHEL7.6)
256 4.4.120-92.70 (SLES12 SP2)
257 4.4.162-94.69 (SLES12 SP3)
258 4.4.0-131 (Ubuntu 16.04)
259 4.15.0-32 (Ubuntu 18.04)
260 vanilla linux 4.6.7 (ZFS only)
261 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
262 * Client known to build on unpatched kernels (others may also work):
263 2.6.32-696.18.7.el6 (RHEL6.9)
264 2.6.32-754.6.3.el6 (RHEL6.10)
265 3.10.0-862.14.4.el7 (RHEL7.5)
266 3.10.0-957.el7 (RHEL7.6)
267 4.14.0-49.13.1.el7a (RHEL7.5)
268 4.14.0-115.2.2.el7a (RHEL7.6)
269 4.4.120-92.70 (SLES12 SP2)
270 4.4.162-94.69 (SLES12 SP3)
271 4.4.0-131 (Ubuntu 16.04)
272 4.15.0-32 (Ubuntu 18.04)
273 vanilla linux 4.15.0 (ZFS only)
274 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
275 * Recommended ZFS / SPL version: 0.7.9
276 * Tested with ZFS / SPL version: 0.7.9
277 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
278 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
279 stack overflow. For more information, please refer to bugzilla 17630.
280 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
281 the re-exporting nodes to be mounted with "32bitapi" mount option
282 * Note for Ubuntu 14.04.5:
283 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
284 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
286 --------------------------------------------------------------------------------
288 2018-04-03 Intel Corporation
290 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
291 for currently supported client and server kernel versions.
292 * Server known to build on patched kernels:
293 2.6.32-431.29.2.el6 (RHEL6.5)
294 2.6.32-504.30.3.el6 (RHEL6.6)
295 2.6.32-573.26.1.el6 (RHEL6.7)
296 2.6.32-642.15.1.el6 (RHEL6.8)
297 2.6.32-696.18.7.el6 (RHEL6.9)
298 3.10.0-693.21.1.el7 (RHEL7.4)
299 3.0.101-0.47.71 (SLES11 SP3)
300 3.0.101-107 (SLES11 SP4)
301 3.12.74-60.64.40 (SLES12 SP1)
302 4.4.114-92.64 (SLES12 SP2)
303 4.4.114-94.11 (SLES12 SP3)
304 3.13.0-101 (Ubuntu 14.04, ZFS only)
305 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
306 4.4.0-116 (Ubuntu 16.04)
307 vanilla linux 4.6.7 (ZFS only)
308 * Client known to build on unpatched kernels:
309 2.6.32-431.29.2.el6 (RHEL6.5)
310 2.6.32-504.30.3.el6 (RHEL6.6)
311 2.6.32-573.26.1.el6 (RHEL6.7)
312 2.6.32-642.15.1.el6 (RHEL6.8)
313 2.6.32-696.18.7.el6 (RHEL6.9)
314 3.10.0-693.21.1.el7 (RHEL7.4)
315 3.0.101-0.47.71 (SLES11 SP3)
316 3.0.101-107 (SLES11 SP4)
317 3.12.74-60.64.40 (SLES12 SP1)
318 4.4.114-92.64 (SLES12 SP2)
319 4.4.114-94.11 (SLES12 SP3)
320 3.13.0-101 (Ubuntu 14.04)
321 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
322 4.4.0-116 (Ubuntu 16.04)
324 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
325 * Recommended ZFS / SPL version: 0.7.6
326 * Tested with ZFS / SPL version: 0.7.6
327 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
328 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
329 stack overflow. For more information, please refer to bugzilla 17630.
330 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
331 the re-exporting nodes to be mounted with "32bitapi" mount option
332 * Note for Ubuntu 14.04.5:
333 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
334 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
336 --------------------------------------------------------------------------------
338 2017-07-05 Intel Corporation
340 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
341 for currently supported client and server kernel versions.
342 * Server known to build on patched kernels:
343 2.6.32-431.29.2.el6 (RHEL6.5)
344 2.6.32-504.30.3.el6 (RHEL6.6)
345 2.6.32-573.26.1.el6 (RHEL6.7)
346 2.6.32-642.15.1.el6 (RHEL6.8)
347 2.6.32-696.3.1.el6 (RHEL6.9)
348 3.10.0-514.21.1.el7 (RHEL7.3)
349 3.0.101-0.47.71 (SLES11 SP3)
350 3.0.101-100 (SLES11 SP4)
351 3.12.74-60.64.40 (SLES12 SP1)
352 4.4.59-92.17 (SLES12 SP2)
353 vanilla linux 4.6.7 (ZFS only)
354 * Client known to build on unpatched kernels:
355 2.6.32-431.29.2.el6 (RHEL6.5)
356 2.6.32-504.30.3.el6 (RHEL6.6)
357 2.6.32-573.26.1.el6 (RHEL6.7)
358 2.6.32-642.15.1.el6 (RHEL6.8)
359 2.6.32-696.3.1.el6 (RHEL6.9)
360 3.10.0-514.21.1.el7 (RHEL7.3)
361 3.0.101-0.47.71 (SLES11 SP3)
362 3.0.101-100 (SLES11 SP4)
363 3.12.74-60.64.40 (SLES12 SP1)
364 4.4.59-92.17 (SLES12 SP2)
366 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
367 * Recommended ZFS / SPL version: 0.6.5.9
368 * Tested with ZFS / SPL version: 0.6.5.9
369 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
370 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
371 stack overflow. For more information, please refer to bugzilla 17630.
372 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
373 the re-exporting nodes to be mounted with "32bitapi" mount option
375 --------------------------------------------------------------------------------
377 12-07-2016 Intel Corporation
379 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
380 for currently supported client and server kernel versions.
381 * Server known to build on patched kernels:
382 2.6.32-431.29.2.el6 (RHEL6.5)
383 2.6.32-504.30.3.el6 (RHEL6.6)
384 2.6.32-573.26.1.el6 (RHEL6.7)
385 2.6.32-642.11.1.el6 (RHEL6.8)
386 3.10.0-514.el7 (RHEL7.3)
387 3.0.101-0.47.71 (SLES11 SP3)
388 3.0.101-80 (SLES11 SP4)
389 3.12.62-60.64.8 (SLES12 SP1)
390 vanilla linux 4.5.7 (ZFS only)
391 * Client known to build on unpatched kernels:
392 2.6.32-431.29.2.el6 (RHEL6.5)
393 2.6.32-504.30.3.el6 (RHEL6.6)
394 2.6.32-573.26.1.el6 (RHEL6.7)
395 2.6.32-642.11.1.el6 (RHEL6.8)
396 3.10.0-514.el7 (RHEL7.3)
397 3.0.101-0.47.71 (SLES11 SP3)
398 3.0.101-80 (SLES11 SP4)
399 3.12.62-60.64.8 (SLES12 SP1)
401 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
402 * Recommended ZFS / SPL version: 0.6.5.8
403 * Tested with ZFS / SPL version: 0.6.5.8
404 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
405 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
406 stack overflow. For more information, please refer to bugzilla 17630.
407 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
408 the re-exporting nodes to be mounted with "32bitapi" mount option
410 --------------------------------------------------------------------------------
412 02-29-2016 Intel Corporation
414 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
415 for currently supported client and server kernel versions.
416 * Server known to build on patched kernels:
417 2.6.32-431.29.2.el6 (RHEL6.5)
418 2.6.32-504.30.3.el6 (RHEL6.6)
419 2.6.32-573.12.1.el6 (RHEL6.7)
420 3.10.0-327.3.1.el7 (RHEL7.2)
421 3.0.101-0.47.71 (SLES11 SP3)
422 3.0.101-68 (SLES11 SP4)
423 vanilla linux 4.2.1 (ZFS only)
424 * Client known to build on unpatched kernels:
425 2.6.32-431.29.2.el6 (RHEL6.5)
426 2.6.32-504.30.3.el6 (RHEL6.6)
427 2.6.32-573.12.1.el6 (RHEL6.7)
428 3.10.0-327.3.1.el7 (RHEL7.2)
429 3.0.101-0.47.71 (SLES11 SP3)
430 3.0.101-68 (SLES11 SP4)
433 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
434 * Recommended ZFS / SPL version: 0.6.4.2
435 * Tested with ZFS / SPL version: 0.6.4.2
436 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
437 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
438 stack overflow. For more information, please refer to bugzilla 17630.
439 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
440 the re-exporting nodes to be mounted with "32bitapi" mount option
442 --------------------------------------------------------------------------------
444 03-10-2015 Intel Corporation
446 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
447 for currently supported client and server kernel versions.
448 * Server known to build on patched kernels:
449 2.6.32-431.29.2.el6 (RHEL6.5)
450 2.6.32-504.8.1.el6 (RHEL6.6)
451 3.0.101-0.46 (SLES11 SP3)
452 * Client known to build on unpatched kernels:
453 2.6.32-431.29.2.el6 (RHEL6.5)
454 2.6.32-504.8.1.el6 (RHEL6.6)
455 3.10.0-123.20.1.el7 (RHEL7)
456 3.0.101-0.46 (SLES11 SP3)
457 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
458 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
459 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
460 stack overflow. For more information, please refer to bugzilla 17630.
461 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
462 the re-exporting nodes to be mounted with "32bitapi" mount option
464 Severity : enhancement
466 Description: control OST-index in IDIF via ROCOMPAT flag.
467 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
468 last_rcvd file. For new formatted OST device, it will be auto set;
469 for the case of upgrading from old OST device, you can enable it
470 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
471 enabled, for new created OST-object, its IDIF-in-LMA will contain
472 the OST-index; for the existing OST-object, the OSD will convert
473 old format IDIF as new format IDIF with OST-index stored in the
474 LMA EA when accessing such OST-object or via OI scrub. Once such
475 flag is enabled, it cannot be reverted back, so the system cannot
476 be downgraded to the orignal incompatible version.
478 --------------------------------------------------------------------------------
480 07-30-2014 Intel Corporation
482 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
483 for currently supported client and server kernel versions.
484 * Server known to build on patched kernels:
485 2.6.32-431.20.3.el6 (RHEL6.5)
486 3.0.101-0.31 (SLES11 SP3)
487 * Client known to build on unpatched kernels:
488 2.6.32-431.20.3.el6 (RHEL6.5)
489 3.0.101-0.31 (SLES11 SP3)
492 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
493 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
494 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
495 stack overflow. For more information, please refer to bugzilla 17630.
496 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
497 the re-exporting nodes to be mounted with "32bitapi" mount option
499 --------------------------------------------------------------------------------
501 10-31-2013 Intel Corporation
503 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
504 * Server builds for kernels:
505 2.6.32-358.18.1.el6 (RHEL6.4)
506 * Client builds for unpatched kernels:
507 2.6.32-358.18.1.el6 (RHEL6.4)
508 3.0.92-0.8 (SLES11 SP3)
510 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
511 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
512 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
513 stack overflow. For more information, please refer to bugzilla 17630.
514 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
515 the re-exporting nodes to be mounted with "32bitapi" mount option
517 --------------------------------------------------------------------------------
519 05-30-2013 Intel Corporation
521 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
522 * Server builds for kernels:
523 2.6.32-358.6.2.el6 (RHEL6.4)
524 * Client builds for unpatched kernels:
525 2.6.32-358.6.2.el6 (RHEL6.4)
526 3.0.74-0.6.6 (SLES11 SP2)
528 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
529 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
530 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
531 stack overflow. For more information, please refer to bugzilla 17630.
532 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
533 the re-exporting nodes to be mounted with "32bitapi" mount option
535 --------------------------------------------------------------------------------
537 9-30-2012 Whamcloud, Inc.
539 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
540 * Server support for kernels:
541 2.6.32-279.5.1.el6 (RHEL6)
542 * Client support for unpatched kernels:
543 2.6.32-279.5.1.el6 (RHEL6)
545 2.6.18-238.19.1.el5 (RHEL5)
546 2.6.16.60-0.69.1 (SLES10)
547 * Recommended e2fsprogs version: 1.41.90.wc4
548 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
549 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
550 stack overflow. For more information, please refer to bugzilla 17630.
551 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
552 the re-exporting nodes to be mounted with "32bitapi" mount option
554 Severity : enhancement
556 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
558 Severity : enhancement
560 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
562 Severity : enhancement
564 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
566 Severity : enhancement
568 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
570 Severity : enhancement
572 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
574 --------------------------------------------------------------------------------
576 04-30-2012 Whamcloud, Inc.
578 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
579 * Server support for kernels:
580 2.6.32-220.4.2.el6 (RHEL6)
581 * Client support for unpatched kernels:
582 2.6.32-220.4.2.el6 (RHEL6)
584 2.6.18-238.19.1.el5 (RHEL5)
585 2.6.16.60-0.69.1 (SLES10)
586 * Recommended e2fsprogs version: 1.41.90.wc4
587 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
588 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
589 stack overflow. For more information, please refer to bugzilla 17630.
590 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
591 the re-exporting nodes to be mounted with "32bitapi" mount option
593 Severity : enhancement
595 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
597 Severity : enhancement
599 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
601 Severity : enhancement
603 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
605 Severity : enhancement
607 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
609 --------------------------------------------------------------------------------
611 09-30-2011 Whamcloud, Inc.
613 * Server support for kernels:
614 2.6.18-238.19.1.el5 (RHEL5)
615 2.6.32-131.6.1.el6 (RHEL6)
616 * Client support for unpatched kernels:
617 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
618 2.6.18-238.19.1.el5 (RHEL5)
619 2.6.32-131.6.1.el6 (RHEL6)
620 2.6.29.4-167.fc11 (FC11)
621 2.6.16.60-0.69.1 (SLES10)
623 * Recommended e2fsprogs version: 1.41.90.wc3
624 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
625 Lustre filesystem with 4K stack may cause a stack overflow. For more
626 information, please refer to bugzilla 17630.
627 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
628 mounted with "32bitapi" mount option
630 Severity : enhancement
631 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
633 Severity : enhancement
634 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
636 Severity : enhancement
637 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
640 Frequency : only with SLES10
642 Description: Use OFED "KMP" provided by Novell
643 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
644 Lustre is now built against this package. That means you need to
645 install the ofed-kmp package from Novell for the patchless client.
646 Note that the ofed-kmp that Novell ships may not exactly match the
647 kernel version but should still be compatible.
649 Severity : enhancement
651 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
652 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
654 --------------------------------------------------------------------------------
656 2010-07-15 Oracle, Inc.
658 * Server support for kernels:
659 2.6.18-164.11.1.el5 (RHEL 5)
660 2.6.18-164.11.1.0.1.el5 (OEL 5)
661 * Client support for unpatched kernels:
662 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
663 2.6.18-164.11.1.el5 (RHEL 5),
664 2.6.18-164.11.1.0.1.el5 (OEL 5)
665 2.6.16.60-0.42.8 (SLES 10),
667 2.6.29.4-167.fc11 (FC11)
668 * Recommended e2fsprogs version: 1.41.10-sun2
670 Severity : enhancement
672 Description: Distribute lfs_migrate script with lustre
673 Details : The lfs_migrate mentioned in the manual is now included in the
674 lustre tree in lustre/scripts and will distributed in lustre RPMs.
676 Severity : enhancement
678 Description: Failover nid list added to import proc file.
680 Severity : enhancement
682 Description: "som_preview" mount option required to use SOM
683 Details : The clients will need to use som_preview mount option
687 Description: Lustre does not do 1MB IOs to HW RAID
688 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
690 Severity : enhancement
692 Description: lfs quota output is not very convenient for awk/sed-parsing
693 Details : Some positions in lfs quota output table could be empty or
694 non-empty which made it hard to parse it with scripts, now
695 a dash is put instead of space where there is not supposed
700 Description: Update to OFED 1.5.1
701 Details : OFED 1.5.1 will be used in environments that don't use
702 vendor supplied OFED.
706 Description: Remove lnet. prefix from lctl params display
707 Details : Also renamed lnet.memused to lnet_memused to
712 Description: Use conf_param -d to remove permanant settings.
716 Description: Add list_param -R
717 Details : lctl list_param now supports recursive mode.
720 Bugzilla : 21829, 22312
721 Description: Fix llobdstat.
722 Details : 1. Now updating the write_bytes counter
723 2. Port script fixes from 1.8.
727 Description: decrease the usage of memory on clients.
728 Details : 1. On clients, recycle dentries and inodes unused.
729 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
734 Description: llobdstat fix and enhancement
735 Details : add a counter to set a limit to how many samples will be returned
736 fix a wildcard in the path to limit to obdfilter stats only
740 Description: lfs check only allowed for root
741 Details : Code cleanups and added a regresion test of non-root use of
746 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
747 Details : Fix hr->hr_index overload.
751 Description: Fix FIEMAP functionality.
755 Description: FC11 patchless client support added
759 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
760 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
764 Description: New acc-sm script and various test-framework enhancments.
768 Description: Previously landed patch was reverted.
772 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
776 Description: Add ldiskfs maxdirsize mount option.
780 Description: Fix mkfs.lustre support for 16TB LUNs.
784 Description: Print mmp_check_interval
785 Details : print mmp_check_interval and make it possible to abort
786 mount operation in case it takes too long.
790 Description: Add cascading_rw to lustre mpi tests.
794 Description: Store SOM data in EA.
795 Details : Also several SOM related cleanups.
799 Description: Add refcount to md_open_data
803 Description: Use refcount on lov_request_set to prevent memory corruption.
807 Description: Statahead: use parent dir's i_mutex for synchronization.
811 Description: loadgen improvements.
815 Description: Enable peer health detection in ksocklnd and ko2iblnd.
819 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
823 Description: Fix race on obd_devcie:md_stats
827 Description: Fix SLES 11 client builds.
828 Details : Resolved uninitalized pointer gcc warnings.
832 Description: Allow non-root access for "lfs check".
833 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
837 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
839 Severity : enhancement
841 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
842 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
846 Description: Sanity test_27p fails.
847 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
848 Sanity tests 27 were cleaned up and improved.
852 Description: Unify req format on client/servers
853 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
854 buffers and swabbers approach..
859 Description: drop join file stuff.
861 Severity : enhancement
863 Description: Add 2.6.27 and SLES11 (patchless client) support
867 Description: Fix in ptlrpc_expire_one_request() to print the signed time
870 Severity : enhancement
872 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
875 Severity : enhancement
877 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
879 Severity : enhancement
881 Description: correctly handle big reply message.
882 Details : send LNet event if reply is bigger than buffer and adjust this buffer
887 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
889 Severity : enhancement
891 Description: Add support for OEL5.
895 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
897 Severity : enhancement
899 Description: "weak-modules" support
900 Details : Implement "weak-modules" support which enables kernel modules
901 to be used with any kernel that implements the same kABI. In
902 order to achieve this modules are now installed in
903 /lib/modules/$(uname -r)/updates/kernel on all distributions.
907 Description: client could not reconnect to OST because of an active request.
908 Details : The client now retries to reconnect to the same server, if a
909 connect request failed with EBUSY or -EAGAIN. Test case 35b
910 conf-sanity.sh has been created.
914 Description: Don't allow make backward step on assignin osc next id.
915 Details : race between allocation next id and ll_sync thread can be cause
916 of set wrong osc next id and can be kill valid ost objects.
920 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
924 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
928 Description: Error handling in osc_statfs_interpret() has been improved.
929 Details : Check in osc_statfs_interpret() for EBADR.
933 Description: Do not send statfs() requests to OSTs disabled by administrator.
934 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
936 Severity : enhancement
938 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
940 Severity : enhancement
942 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
944 Severity : enhancement
946 Description: Enhancement for lfs(1) command to use numeric uid/gid.
950 Description: Improved cleanup procedure for conf-sanity test_46a.
954 Description: Do not start bulk transfer, if client tried to reconnect.
955 Details : Do not start bulk transfer, if client tried to reconnect,
956 but failed due to active rpcs. Check for exp_abort_active_req
957 has been implemented.
961 Description: Prevent inconsistences between linux and lustre mount structures.
962 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
963 Make the sleep interruptible.
965 Severity : enhancement
967 Description: provide server to client comms path
968 Details : server to client communications path via new LDLM_SET_INFO rpc
972 Description: truncate starts GFP_FS allocation under transaction causing deadlock
973 Details : ldiskfs_truncate calls grab_cache_page which may start page
974 allocation under an open transaction. This may lead to
975 calling prune_icache with consequent lustre reentrance.
979 Description: send correctly lsm on open replay
980 Details : MDS is trust to LSM size on replay open, but client can set wrong size
985 Description: enable client interface failover
986 Details : When a child reconnects from another NID, properly update export
987 nid hash position and ldlm reverse import.
989 Severity : enhancement
991 Description: Communicate OST degraded/readonly state via statfs to MDS
992 Details : Flags in the statfs returned from OSTs indicate whether the
993 OST is in a degraded RAID state, or if the filesystem has
994 turned read-only after a filesystem error is detected.
999 Description: don't panic if EPROTO was hit when reading symlink
1000 Details : correctly handling request reference in error cases.
1002 Severity : enhancement
1004 Description: Add LustreNetLink, a kernel-userspace communcation path.
1006 Severity : enhancement
1008 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1013 Description: access to llog context before init.
1014 Details : move handling CATALOGS file at osc layer and forbid access to llog
1015 context before init.
1018 Frequency : rare, only when client is writing to OST on local node
1020 Description: Avoid deadlock for local client writes
1021 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1022 memory freeing context. This allows OST threads to set the
1023 PF_MEMALLOC flag on task structures in order to allocate memory
1024 from reserved pools and complete IO.
1025 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1026 so that the OST threads generate memory pressure and allow
1027 inactive pages to be reclaimed.
1029 Severity : enhancement
1031 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1036 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1037 Details : move ldlm namespace creation in setup phase to avoid grab
1038 _lprocfs_lock with cli_sem held.
1042 Description: Temporarily disable grant shrink.
1043 Details : Disable the feature for debugging.
1046 Frequency : only when formatting test filesystems
1048 Description: Unable to run several mkfs.lustre on loop devices at the same time
1049 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1050 formatting. The solution is to proper handle the error.
1052 Severity : enhancement
1054 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1056 Severity : enhancement
1058 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1063 Description: resolve race between obd_disconnect and class_disconnect_exports
1064 Details : if obd_disconnect will be called to already disconnected export he
1065 forget release one reference and osc module can't unloaded.
1067 Severity : enhancement
1069 Description: Allow tuning service thread via /proc
1070 Details : For each service a new
1071 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1072 created that can be used to set min/max thread counts, and get the
1073 current number of running threads.
1078 Descriptoin: don't return error if have partially created objects for file.
1079 Details : lov_update_create_set uses set->set_success as index for created
1080 objects, so if some requests will be failed, they will have hole
1081 at end of array and qos_shrink_lsm for allocate correct lsm.
1083 Severity : enhancement
1085 Description: Update OFED support to 1.4.1
1090 Description: Reduce small size read RPC
1091 Details : Set read-ahead limit for every file and only do read-ahead when
1092 available read-ahead pages are bigger than 1M to avoid small size
1095 Severity : enhancement
1097 Description: move AT tunable parameters for more consistent usage
1098 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1100 Severity : enhancement
1102 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1103 Details : allow skip disconnected ost for send statfs request and hide error
1109 Description: fix racy locking of mballoc block bitmaps causing BUG
1110 Details : The locking of the mballoc buddy bitmap and the in-memory
1111 block bitmap was using two different spin locks in some
1112 cases. This made it possible to incorrectly access the
1113 mballoc bitmap while another process was modifying it,
1114 causing a sanity assertion to fail. While no on-disk corruption
1115 was reported, there was some risk of this happening.
1118 Frequency : rare, on llog test 6
1120 Descriptoin: don't allow connect to already connected import
1121 Details : allowing connect to already connected import is hide connecting problem.
1126 Descriptoin: allow kill process which wait statahead result
1127 Details : for some reasons 'ls' can stick in waiting result from statahead,
1128 in this case need way for kill this process.
1130 Severity : enhancement
1132 Description: Add state history info file, enhance import info file
1133 Details : Track import connection state changes in a new osc/mdc proc file;
1134 add overview-type data to the osc/mdc import proc file.
1136 Severity : enhancement
1138 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1140 Severity : enhancement
1142 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1147 Descriptoin: OST grant too much space to client even there are not enough space.
1148 Details : Client will shrink its grant cache to OST if there are no write
1149 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1150 this grant cache if there are already not enough avaible space
1151 (left_space < total_clients * 32M).
1154 Frequency : start MDS on uncleanly shutdowned MDS device
1156 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1157 Details : stay in waiting mds<>ost recovery finished produce random bugs
1158 due race between two ll_sync thread for one lov target. send
1159 ACTIVATE event only if connect realy finished and import have
1163 Frequency : rare, connect and disconnect target at same time
1165 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1166 Details : don't call obd_disconnect under lov_lock. this long time
1167 operation and can block ptlrpcd which answer to connect request.
1172 Descriptoin: don't lose wakeup for imp_recovery_waitq
1173 Details : recover_import_no_retry or invalidate_import and import_close can
1174 both sleep on imp_recovery_waitq, but we was send only one wakeup
1178 Frequency : always with long access acl
1180 Descriptoin: mds can't pack reply with long acl.
1181 Details : mds don't control size of acl but they limited by reint/getattr
1184 Severity : enhancement
1186 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1188 Severity : enhancement
1190 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1193 Frequency : start MDS on uncleanly shutdowned MDS device
1195 Descriptoin: aborting recovery hang on MDS
1196 Details : don't throttle destroy RPCs for the MDT.
1199 Frequency : on remount
1201 Description: external journal device not working after the remount
1202 Details : clear dev_rdonly flag for external journal devices in
1208 Description: shutdown vs evict race
1209 Details : client_disconnect_export vs connect request race.
1210 if client will evicted at this time - we start invalidate
1211 thread without referece to import and import can be freed
1215 Frequency : rare, need acl's on inode.
1217 Description: client can't handle ost additional correctly
1218 Details : if ost was added after client connected to mds client can have
1219 hit lnet_try_match_md ... to big messages to wide striped files.
1220 in this case need teach client to handle config events about add
1221 lov target and update client max ea size at that event.
1223 Severity : enhancement
1225 Description: Changelogs
1226 Details : Changelogs are a lightweight mechanism to track filesystem
1227 metadata and namespace changes. The changelog is recorded
1228 permanently on the MDTs, and is periodically "consumed" / purged
1229 when records are no longer needed.
1231 Severity : enhancement
1233 Description: compact fld format with extents
1234 Details : Store range of seq rather than every seq in FLD. Seq
1235 controller update FLD rather than clients. In Case of CMD, mdt0
1236 has FLD, all other metadata server act as non persistent proxy
1237 for FLD queries and cache fld entries in fld cache.
1242 Description: don't skip ost target if they assigned to file
1243 Details : Drop slow OSCs if we can, but not for requested start idx.
1244 This means "if OSC is slow and it is not the requested
1245 start OST, then it can be skipped, otherwise skip it only
1246 if it is inactive/recovering/out-of-space.
1250 Description: more cleanup in mds_lov
1251 Details : not send LOV EA under replay, we can't know about they size at this
1252 time. Don't allow client connect to mds before any ost connected,
1253 for avoid problems with LOV EA size and returning EIO to client.
1255 Severity : enhancement
1257 Description: Interoperability at server side (Disk interoperability)
1259 Severity : enhancement
1261 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1263 Severity : enhancement
1265 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1267 Severity : enhancement
1269 Description: New client IO stack (CLIO).
1271 Severity : enhancement
1273 Description: Commit on sharing. Eliminate inter-client dependencies between
1274 uncommitted transactions by doing transaction commits.
1275 Thereby clients may recovery independently.
1278 Frequency : Create a symlink file with a very long name
1280 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1281 Details : If there is no extra space in the request for early cancels,
1282 ldlm_req_handles_avail() returns 0 instead of a negative value.
1284 Severity : enhancement
1286 Description: Add /proc entry for import status
1287 Details : The mdc, osc, and mgc import directories now have
1288 an import directory that contains useful import data for debugging
1289 connection problems.
1291 Severity : enhancement
1293 Description: Re-disable certain /proc logging
1294 Details : Enable and disable client's offset_stats, extents_stats and
1295 extents_stats_per_process stats logging on the fly.
1298 Frequency : Only on FC kernels 2.6.22+
1300 Description: oops in statahead
1301 Details : Do not drop reference count for the dentry from VFS when lookup,
1302 VFS will do that by itself.
1304 Severity : enhancement
1306 Description: Generic /proc file permissions
1307 Details : Set /Proc file permissions in a more generic way to enable non-
1308 root users operate on some /proc files.
1312 Description: Hitting mdc_commit_close() ASSERTION
1313 Details : Properly handle request reference release in
1314 ll_release_openhandle().
1318 Description: quota recovery deadlock during mds failover
1319 Details : This patch includes att18982, att18236, att18237 in bz14840.
1321 1. fix osts hang when mds does failover with quotaon
1322 2. prevent watchdog storm when osts threads wait for the
1327 Frequency : only patchless client
1328 Description: add workaround for race between add/remove dentry from hash
1330 Severity : enhancement
1332 Description: Allow OST glimpses to return PW locks
1336 Description: LBUG when llog conf file is full
1337 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1341 Description: Prevent import from entering FULL state when server in recovery
1345 Description: service mount cannot take device name with ":"
1346 Details : Only when device name contains ":/" will mount treat it as
1352 Description: replace ptlrpcd with the statahead thread to interpret the async
1353 statahead RPC callback
1357 Frequency : on recovery
1358 Description: I/O failures after umount during fail back
1359 Details : if client reconnected to restarted server we need join to recovery
1360 instead of find server handler is changed and process self
1361 eviction with cancel all locks.
1363 Severity : enhancement
1365 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1367 Severity : enhancement
1369 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1371 Severity : enhancement
1373 Description: Upcall on Lustre log has been dumped
1374 Details : Allow for a user mode script to be called once a Lustre log has
1375 been dumped. It passes the filename of the dumped log to the
1376 script, the location of the script can be specified via
1377 /proc/sys/lnet/debug_log_upcall.
1382 Description: avoid idr_remove called for id which is not allocated.
1383 Details : Move assigment s_dev for clustered nfs to end of initialization,
1384 to avoid problem with error handling.
1389 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1390 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1391 into CONN_USED_HASH and this prodice warning when put connection
1392 again in unused hash.
1395 Severity : enhancement
1397 Description: File striping can now be set to use an arbitrary pool of OSTs.
1399 Severity : enhancement
1401 Description: Export bytes_read/bytes_write count on OSC/OST.
1405 Description: Early reply size mismatch, MGC loses connection
1406 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1407 the connect flags are properly negotiated.
1412 Description: quotas are not honored with O_DIRECT
1413 Details : all writes with the flag O_DIRECT will use grants which leads to
1414 this problem. Now using OBD_BRW_SYNC to guard this.
1418 Description: add quota statistics
1419 Details : 1. sort out quota proc entries and proc code.
1420 2. add quota statistics
1422 Severity : enhancement
1424 Description: enable quota support for HEAD.
1428 Description: Properly propagate oinfo flags from lov to osc for statfs
1429 Details : restore missing copy oi_flags to lov requests.
1431 Severity : enhancement
1433 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1434 : l_getgroups(8), lst(8), routerstat(8)
1435 Details : included man pages for llobdstat(8), llstat(8),
1436 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1438 Severity : enhancement
1440 Description: configure's --enable-quota should check the
1441 : kernel .config for CONFIG_QUOTA
1442 Details : configure is terminated if --enable-quota is passed but
1443 : no quota support is in kernel
1447 Description: Remove portals compatibility
1448 Details : Remove portals compatibility, not interoperable with releases
1453 Description: Resolve device initialization race
1454 Details : Prevent proc handler from accessing devices added to the
1455 obd_devs array but yet be intialized.
1457 Severity : enhancement
1459 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1461 Severity : enhancement
1463 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1468 Description: Using wrong pointer in osc_brw_prep_request
1469 Details : Access to array[-1] can produce panic if kernel compiled with
1470 CONFIG_PAGE_ALLOC enabled
1474 Description: Client runs out of low memory
1475 Details : Consider only lowmem when counting initial number of llap pages
1479 Description: *optional* service tags registration
1480 Details : if the "service tags" package is installed on a Lustre node
1481 When the filesystem is mounted, a local-node service tag will
1482 be created. See http://inventory.sun.com/ for more information
1483 about the Service Tags asset management system.
1487 Description: Kernel BUG tries to release flock
1488 Details : Lustre does not destroy flock lock before last reference goes
1489 away. So always drop flock locks when client is evicted and
1490 perform unlock regardless of successfulness of speaking to MDS.
1494 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1498 Description: sanity test 65a fails if stripecount of -1 is set
1499 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1504 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1505 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1508 Severity : enhancement
1510 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1514 Description: do not process already freed flock
1515 Details : flock can possibly be freed by another thread before it reaches
1516 to ldlm_flock_completion_ast.
1520 Description: LBUG during stress test
1521 Details : Need properly lock accesses the flock deadlock detection list.
1525 Description: oops in page fault handler
1526 Details : kernel page fault handler can return two special 'pages' in error
1527 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1531 Description: timeout with invalidate import.
1532 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1533 should be handled by ptlrpcd. This produce long age waiting and
1534 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1536 Severity : enhancement
1538 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1542 Description: enable MGS and MDT services start separately
1543 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1544 a co-located MGS without starting the MGS, which is a complement
1545 to 'nosvc' mount option.
1549 Frequency : after recovery
1550 Description: precreate to many object's after del orphan.
1551 Details : del orphan st in oscc last_id == next_id and this triger growing
1552 count of precreated objects. Set flag LOW to skip increase count
1553 of precreated objects.
1557 Frequency : rare, on clear nid stats
1558 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1559 Details : when clean nid stats sometimes try destroy live entry,
1560 and this produce panic in free.
1564 Description: Stack overflow during MDS log replay
1565 ease stack pressure by using a thread dealing llog_process.
1569 Description: wait until IO finished before start new when do lock cancel.
1570 Details : VM protocol want old IO finished before start new, in this case
1571 need wait until PG_writeback is cleared until check dirty flag and
1572 call writepages in lock cancel callback.
1574 Severity : enhancement
1576 Description: using special macro for print time and cleanup in includes.
1580 Description: mds_mfd_close() ASSERTION(rc == 0)
1581 Details : In mds_mfd_close(), we need protect inode's writecount change
1582 within its orphan write semaphore to prevent possible races.
1586 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1590 Frequency : rare, on shutdown ost
1591 Description: don't hit live lock with umount ost.
1592 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1593 use shrink_dcache_sb instead.
1597 Description: don't panic with use echo client
1598 Details : echo client pass NULL as client nid pointer and this produce null
1599 pointer dereference.
1603 Description: fix build on ppc32
1604 Details : compile code with -m64 flag produce wrong object file for ppc32.
1608 Description: add message levels for liblustreapi
1612 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1613 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1614 will result in an -EAGAIN and in the exit call path we call
1615 alloc_rr() which will with increasing aggressiveness attempt to
1616 aquire precreated objects on the minimum number of required OSCs.
1620 Description: Use old size assignment to avoid deadlock
1621 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1622 the scheduling while holding a spinlock. We do not need locking
1623 for size in ll_update_inode() because size is only updated from
1624 the MDS for directories or files without objects, so there is no
1625 other place to do the update, and concurrent access to such inodes
1626 are protected by the inode lock.
1630 Description: resolve "_IOWR redefined" build error on SLES10
1634 Description: dump the memory debugging after all modules are unloaded to
1635 suppress false negative in conf_sanity test 39
1637 Severity : enhancement
1639 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1644 Description: fix SLES kernel versioning
1645 Details : the kernel version for our SLES 10 kernel did not include a "-"
1646 before the "smp" at the end. while this was not a problem in
1647 general, it did mean that software trying to use the kernel
1648 version to try to detect a vendor specific kernel would fail.
1649 this was most evident by the OFED build scripts.
1653 Description: Don't update lov_desc members until making sure they are valid
1654 Details : When updating lov_desc members via proc fs, need fix their
1655 validities before doing the real update.
1659 Description: don't put request into delay list while invalidate in flight.
1660 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1661 import in flight. this produce timeout for invalidate and sometimes
1662 can cause stale data.
1666 Frequency : on ppc only
1667 Description: not convert ost objects for directory because it's not exist.
1668 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1672 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1674 Severity : enhancement
1676 Description: Update to support 2.6.22.14 vanilla kernel.
1680 Frequency : rare, on recovery
1681 Description: read procfs can produce deadlock in some situation
1682 Details : Holding lprocfs lock which send rpc can produce block for destroy
1683 obd objects and this also block reconnect with -EALREADY. This
1684 isn't fix all lprocfs bugs - but make it rare.
1686 Severity : enhancement
1688 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1691 Frequency : frequent on X2 node
1693 Description: mdc_set_open_replay_data LBUG
1694 Details : Set replay data for requests that are eligible for replay.
1698 Description: lustre_mgs: operation 101 on unconnected MGS
1699 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1700 MGC, and late on MGC cannot successfully connect to MGS and a lot
1701 of the error messages complaining that MGS is not connected.
1705 Frequency : on network error
1706 Description: panic with double free request if network error
1707 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1708 but it's true only for synchronus enqueue, for async enqueue
1709 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1712 Severity : enhancement
1714 Description: client-side metadata stat-ahead during readdir(directory readahead)
1715 Details : perform client-side metadata stat-ahead when the client detects
1716 readdir and sequential stat of dir entries therein
1719 Frequency : on start mds
1721 Description: Implement get_info(last_id) in obdfilter.
1724 Frequency : occasional
1726 Description: Correctly check stale fid, not start epoch if ost not support SOM
1727 Details : open with flag O_CREATE need set old fid in op_fid3 because
1728 op_fid2 was overwritten with new generated fid, but mds can answer
1729 with one of these two fids and both is not stale. Setattr
1730 incorrectly started an epoch and assume will be called
1731 done_writing, but without SOM done_writing ever being called.
1734 Frequency : rare, depends on device drivers and load
1736 Description: MDS or OSS nodes crash due to stack overflow
1737 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1738 In some cases, in conjunction with device drivers that use a lot
1739 of stack the MDS (or possibly OSS) service threads could overflow
1740 the stack. One change which was identified to consume additional
1741 stack has been reworked to avoid the extra stack usage.
1744 Frequency : occasional
1746 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1747 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1748 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1749 oscc_has_objects and see OSC has no objests and call
1750 oscc_internal_create to resend create request.
1752 Severity : enhancement
1754 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1756 Severity : enhancement
1758 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1761 Frequency : very rare
1763 Description: Fix replay if there is an un-replied request and open
1764 Details : In some cases, older replay request will revert the
1765 mcd->mcd_last_xid on MDS which is used to record the client's
1766 latest sent request.
1768 Severity : enhancement
1770 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1772 Severity : enhancement
1774 Description: Add rhel5 support to HEAD.
1776 Serverity : enhancement
1778 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1783 Description: Don't allow skipping OSTs if index has been specified.
1784 Details : Don't allow skipping OSTs if index has been specified, make
1785 locking in internal create lots better.
1789 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1790 Details : access to bitfield in structure is always rounded to long
1791 and this produce problem with not atomic change any bit.
1795 Description: Lustre make rpms failed.
1796 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1797 builds Lustre rpms from tarball.
1800 Frequency : rare on shutdown ost
1802 Description: If llog cancel was not send before clean_exports phase, this can
1803 produce deadlock in llog code.
1804 Details : If llog thread has last reference to obd and call class_import_put
1805 this produce deadlock because llog_cleanup_commit_master wait when
1806 last llog_commit_thread exited, but this never success because was
1807 called from llog_commit_thread.
1811 Description: allow userland application know is lost one of stripes.
1812 Details : fill lvb_blocks with error code on ost and return it to
1813 application if error flag found.
1817 Description: NULL lov_tgts causing MDS oops
1818 Details : more safe checks for NULL lov_tgts for avoid oops.
1820 Severity : enhancement
1822 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1826 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1827 Details : If an OST is down the MDS will hang indefinitely in
1828 obd_statfs_rqset() waiting for the statfs data. While for
1829 MDS QOS usage of statfs, it should not stuck in waiting.
1831 Severity : enhancement
1833 Description: remote_acl support
1834 Details : Support ACL-based permission check for remote user.
1835 Support setfacl/getfacl for remote user with the utils
1836 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1837 the system "{s,g}etfacl" utils.
1839 Severity : enhancement
1841 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1843 Severity : enhancement
1845 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1849 Description: make mgs_setparam() handle fsname containing dash
1850 Details : fsname containing a dash does not work with lctl conf_param
1852 Severity : enhancement
1854 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1856 Severity : enhancement
1858 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1862 Frequency : rare, at shutdown
1863 Description: access already free / zero obd_namespace.
1864 Details : if client_disconnect_export was called without force flag set,
1865 and exist connect request in flight, this can produce access to
1866 NULL pointer (or already free pointer) when connect_interpret
1867 store ocd flags in obd_namespace.
1871 Frequency : only at startup
1872 Description: not alloc memory with spinlock held.
1873 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1874 if any spinlock held.
1876 Severity : enhancement
1878 Description: make lustre randomly fail allocating memory
1879 Details : Make lustre randomly failed allocating memory for testing purpose.
1881 Severity : enhancement
1883 Description: lost problems with lov objid file
1884 Details : Fixes some scability and access to not inited memory problems
1885 in work with lov objdid file.
1890 Description: lfs find does not continue on file error
1891 Details : Continue other files processing when a file/dir is absent.
1895 Description: Inconsistent usage of lustre_pack_reply()
1896 Details : Standardize the usage of lustre_pack_reply() such that it
1897 always generate a CERROR on failure.
1900 Frequency : occasional
1902 Description: MDS hang or stay in waiting lock
1903 Details : If client receive lock with CBPENDING flag ldlm need send lock
1904 cancel as separate rpc, to avoid situation when cancel request
1905 can't processed due all i/o threads stay in wait lock.
1909 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1913 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1917 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1921 Description: Update to RHEL4 latest kernel.
1923 Severity : enhancement
1925 Description: Build SLES10 patchless client fails
1926 Details : The configure was broken by run ./configure with
1927 --with-linux-obj=.... argument for patchless client. When the
1928 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1929 can't search header adequately. Use absolute path such as
1930 -I($LINUX)/include instead.
1934 Description: interrupt oig_wait produce painc on resend.
1935 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1936 private set, and this produce situation when rq_ptlrpcd_data not
1937 copyed to new allocated request and triggered LBUG on assert
1938 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1939 ptlrpcd set if request is changed and can be safety changed to use
1944 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1945 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1946 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1947 these macros and replace them with their difinitions instead.
1951 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1952 Details : Update which_patch & target file for SLES10 latest kernel.
1955 Description: Remove client patches from SLES 10 kernel.
1956 Details : This causes SLES 10 clients to behave as patchless clients
1957 even on a Lustre-patched (server) kernel.
1959 Severity : enhancement
1961 Description: self-adjustable client's lru lists
1962 Details : use adaptive algorithm for managing client cached locks lru
1963 lists according to current server load, other client's work
1964 pattern, memory activities, etc. Both, server and client
1965 side namespaces provide number of proc tunables for controlling
1968 Severity : enhancement
1970 Description: light-weight GSS support
1971 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1972 and reduce performance overhead.
1974 Severity : enhancement
1976 Description: Linux keyring support
1977 Details : Support using service of Linux keyring for Lustre GSS internal
1978 context refresh/cache mechanism.
1982 Description: Fix errors in lfs documentation
1983 Details : Fixes man pages
1987 Description: don't use GFP_* in generic Lustre code.
1988 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1992 Description: obdclass is limited by single OBD_ALLOC(idarray)
1993 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1997 Description: warnings with build patchless client with vanila 2.6.19 and up
1998 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1999 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2003 Frequency : Only for SLES
2004 Description: sanity_quota fail test_1
2005 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2006 file, which makes TSTID[2] inunique.
2010 Frequency : Only for SLES9
2011 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2012 Details : Update target/ChangeLog/which_patch .
2016 Frequency : Only for test_10 in sanity-quota.sh
2017 Description: a bug in quota test code
2018 Details : checking a test flag in a wrong place
2022 Frequency : Only for test_13 in sanity-quota.sh
2023 Description: a bug in quota test script
2024 Details : shouldn't check overfull allocated quota
2028 Description: jbd statistics
2029 Details : Port older jbd statistics patch for sles10
2031 Severity : enhancement
2033 Description: Add printing inode info into message about error in writepage.
2038 Description: reply_lock_interpret crash due to race with it and lock cancel.
2039 Details : Do not replay locks that are being cancelled. Do not reference
2040 locks by their address during replay, just by their handle.
2044 Frequency : When flocks are used.
2045 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2046 Details : Flock locks might destroy just granted lock if it could be merged
2047 with another existing flock, this is done in completion handler,
2048 so teach ldlm_cli_enquque_fini that this is a valid case for
2054 Description: Oops in read and write path when failing to allocate lock.
2055 Details : Check if lock allocation failed and return error back.
2059 Description: lstripe command fails for valid OST index
2060 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2061 instead of lov->desc.ld_active_tgt_count.
2065 Description: sanity.sh failed test 103
2066 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2071 Description: df doesn't work properly if diskfs blocksize != 4K
2072 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2076 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2077 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2079 Severity : enhancement
2081 Description: Adaptive timeouts
2082 Details : RPC timeouts adapt to changing server load and network
2083 conditions to reduce resend attempts and improve recovery time.
2087 Description: llapi_file_create() does not allow some changes
2088 Details : add llapi_file_open() that allows specifying the mode and
2089 open flags, and also returns an open file handle.
2093 Description: merge and cleanup kernel patches.
2094 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2098 Description: Add journal checksum support.(Kernel part)
2099 Details : The journal checksum feature adds two new flags i.e
2100 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2101 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2102 indicates that the commit block contains the checksum for
2103 the blocks described by the descriptor blocks. Now commit
2104 record can be sent to disk without waiting for descriptor
2105 blocks to be written to disk. This behavior is controlled
2106 using JBD2_FEATURE_ASYNC_COMMIT flag.
2110 Description: OSS needs mutliple precreate threads
2111 Details : Add ability to start more than one create thread per OSS.
2115 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2116 Details : execute lfs setstripe on client
2120 Description: mds_obd_create error creating tmp object
2121 Details : When the user sets quota on root, llog will be affected and can't
2122 create files and write files.
2125 Frequency : Always on ia64 patchless client, and possibly others.
2127 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2128 Details : This allows the patchless client to be loaded on architectures
2129 without this export.
2133 Description: RedHat Update kernel for RHEL5
2134 Details : Add the kernel config file for RHEL5.
2138 Description: RedHat Update kernel for RHEL5
2139 Details : Modify the kernel config file more closer RHEL5.
2143 Description: Build failure against Centos5 (RHEL5)
2144 Details : Define PAGE_SIZE when it isn't present.
2148 Description: test_7 on sanity-quota.sh can't be used on separate machines
2149 Details : add facet to handle it
2153 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2154 Details : with NFS, the anon dentry's parent was set to itself in
2155 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2156 corresponding dentry other than use rec->ur_name.
2158 Severity : enhancement
2160 Description: lfs setstripe enhancement
2161 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2165 Description: enable data checksumming by default
2166 Details : enable checksum by default, allow --disable-checksum
2167 configure option and "-o nochecksum" mount option. Checksums
2168 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2172 Description: lustre support for RHEL5
2173 Details : Add support for RHEL5.
2177 Description: Client eviction due to failover config
2178 Details : after a connection loss, the lustre client should attempt to
2179 reconnect to the last active server first before trying the
2180 other potential connections.
2184 Description: when mds and osts use different quota unit(32bit and 64bit),
2185 quota will be released repeatly.
2186 Details : void sending multiple quota reqs to mds, which will keep the
2187 status between the reqs.
2191 Description: osts not allocated evenly to files
2192 Details : change the condition to increase offset_idx
2195 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2197 Description: Data corruption for OSTs that are formatted larger than 2TB
2199 Details : When generating the bio request for lustre file writes the
2200 sector number would overflow a temporary variable before being
2201 used for the IO. The data reads correctly from Lustre (which
2202 will overflow in a similar manner) but other file data or
2203 filesystem metadata may be corrupted in some cases.
2207 Description: Tune the kernel for good SCSI performance.
2208 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2209 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2214 Description: rewrite ext2-derived code in obdclass/uuid.c
2215 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2216 from scratch preserving functionality.
2220 Description: "lfs find -obd UUID" prints directories
2221 Details : "lfs find -obd UUID" will return all directory names instead
2222 of just file names. It is incorrect because the directories
2223 do not reside on the OSTs.
2227 Description: use i_size_read and i_size_write in 2.6 port
2228 Details : replace inode->i_size access with i_size_read/write()
2231 Frequency : when removing large files
2233 Description: scheduling issue during removal of large Lustre files
2234 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2235 It causes scheduling issues when removing large files (17TB in the
2239 Frequency : only with liblustre clients on XT3
2241 Description: evictions taking too long
2242 Details : allow llrd to evict clients directly on OSTs
2245 Frequency : only on ppc
2247 Description: /proc/fs/lustre/devices broken on ppc
2248 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2249 We should make sure the type of 'index' is loff_t and then cast
2250 explicitly as needed below. Do not assign an explicitly cast
2254 Frequency : only for rhel5
2256 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2257 Details : Modify the target file & which_kernel.
2259 Severity : enhancement
2261 Description: omit set fsid for export NFS
2262 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2263 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2267 Frequency : Always, for kernels after 2.6.16
2268 Description: Fix warning idr_remove called for id=.. which is not allocated.
2269 Details : Last kernels save old s_dev before kill super and not allow
2270 to restore from callback - restore it before call kill_anon_super.
2274 Description: buffer overruns could theoretically occur
2275 Details : llapi_semantic_traverse() modifies the "path" argument by
2276 appending values to the end of the origin string, and a
2277 overrun may occur. Adding buffer overrun check in liblustreapi.
2281 Description: Fix error on 'ls .' at the top of the Lustre mount.
2282 Details : Don't revalidate dentry if it is a root dentry.
2286 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2287 Details : Modify vm-tunables-rhel4.patch.
2291 Description: Kernel config for 2.6.18-vanilla.
2292 Details : Modify targets/2.6-vanilla.target.in.
2293 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2294 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2295 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2296 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2300 Description: improve handling recoverable errors
2301 Details : if request processig with error which can be recoverable on server
2302 request should be resend, otherwise page released from cache and
2307 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2308 Details : Modify target file & which_patch.
2309 A flaw was found in the IA32 system call emulation provided
2310 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2311 value could be stored in the %RAX register, which could trigger an
2312 out-of-bounds system call table access. An untrusted local user
2313 could exploit this flaw to run code in the kernel
2314 (ie a root privilege escalation). (CVE-2007-4573).
2318 Description: O_DIRECT bypasses client statistics.
2319 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2320 read_ahead_stats were updated. copied the stats section from
2321 osc_send_oap_rpc() into async_internal().
2325 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2329 Description: change order of libsysio includes
2330 Details : '#include sysio.h' should always come before '#include xtio.h'
2334 Description: handle "serious error: objid * already exists" more gracefully
2335 Details : If LAST_ID value on disk is smaller than the objects existing in
2336 the O/0/d* directories, it indicates disk corruption and causes an
2337 LBUG(). If the object is 0-length, then we should use the existing
2338 object. This will help to avoid a full fsck in most cases.
2340 Severity : enhancement
2342 Description: adapt the lustre_config script to support the upgrade case
2343 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2344 server targets to 1.6 in parallel.
2348 Description: To avoid grant space > avaible space when the disk is almost
2349 full. Without this patch you might see the error "grant XXXX >
2350 available" or some LBUG about grant, when the disk is almost
2352 Details : In filter_check_grant, for non_grant cache write, we should
2353 check the left space by if (*left > ungranted + bytes), instead
2354 of (*left > ungranted), because only we are sure the left space
2355 is enough for another "bytes", then the ungrant space should be
2356 increase. In client, we should update cl_avail_grant only there
2357 is OBD_MD_FLGRANT in the reply.
2361 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2362 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2363 described in CVE-2007-4573. This problem could allow untrusted
2364 local users to gain root access.
2367 Frequency : when using O_DIRECT and quotas
2369 Description: Incorrect file ownership on O_DIRECT output files
2370 Details : block usage reported by 'lfs quota' does not take into account
2371 files that have been written with O_DIRECT.
2376 Description: touch file failed when fs is not full
2377 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2378 otherwise we can get ENOSP while fs is not full.
2382 Description: parallel lock callbacks
2383 Details : Instead of sending blocking and completion callbacks as separated
2384 requests, adding them to a set and sending in parallel.
2387 Frequency : only for Cray XT3
2388 Bugzilla : 12829/13455
2389 Description: Changing primary group doesn't change the group lustre assigns to
2391 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2392 overridden with the primary group provided by the group upcall,
2393 whereas the supplied fsgid can be trusted if it is in the list of
2394 supplementary groups returned by the group upcall.
2396 Severity : enhancement
2398 Description: Allow masking D_WARNING, D_ERROR messages from console
2399 Details : Console messages can now be disabled via lnet.printk.
2403 Description: User code with malformed file open parameter crashes client node
2404 Details : Before packing join_file req, all the related reference should be
2405 checked carefully in case some malformed flags cause fake
2406 join_file req on client.
2410 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2411 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2412 mode is equal to the granted mode. However, it should also check
2413 that LDLM_FL_CP_REQD is not set.
2417 Description: Heavy nfs access might result in deadlocks
2418 Details : After ELC code landed, it is now improper to enqueue any mds
2419 locks under och_sem, because enqueue might want to decide to
2420 cancel open locks for same inode we are holding och_sem for.
2424 Description: Client eviction while running blogbench
2425 Details : A lot of unlink operations with concurrent I/O can lead to a
2426 deadlock causing evictions. To address the problem, the number of
2427 oustanding OST_DESTROY requests is now throttled to
2428 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2429 callbacks are processed in priority.
2433 Description: enable ACLs on MDS by default
2434 Details : ACLs must be enabled on MDS by default.
2437 Frequency : PPC/PPC64 only
2439 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2440 Details : fix duplicated definitions between asm-ppc64/types.h and
2441 lustre_types.h on PPC.
2444 Frequency : PPC/PPC64 only
2446 Description: asm-ppc/segment.h does not exist
2447 Details : fix compile issue on PPC.
2451 Description: better handle error messages in extents code
2454 Frequency : RHEL4 only
2456 Description: mkfs is very slow on IA64/RHEL4
2457 Details : A performance regression has been discovered in the MPT Fusion
2458 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2459 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2460 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2462 Severity : enhancement
2464 Description: SNMP support enhancement
2465 Details : Adding total number of sampled request for an MDS node in snmp
2468 Severity : enhancement
2470 Description: Optimize ldlm waiting list processing for PR extent locks
2471 Details : When processing waiting list for read extent lock and meeting read
2472 lock that is same or wider to it that is not contended, skip
2473 processing rest of the list and immediatelly return current
2474 status of conflictness, since we are guaranteed there are no
2475 conflicting locks in the rest of the list.
2479 Description: Time out and refuse to reconnect
2480 Details : When the failover node is the primary node, it is possible
2481 to have two identical connections in imp_conn_list. We must
2482 compare not conn's pointers but NIDs, otherwise we can defeat
2483 connection throttling.
2487 Description: port llog fixes from b1_6 into HEAD
2488 Details : Port llog reference couting and some llog cleanups from b1_6
2489 (bug 10800) into HEAD, for protect from panic and access to already
2490 free llog structures.
2494 Description: Detect stride IO mode in read-ahead
2495 Details : When a client does stride read, read-ahead should detect that and
2496 read-ahead pages according to the detected stride pattern.
2500 Description: data checksumming impacts single node performance
2501 Details : add support for several checksum algorithm. Currently, only CRC32
2502 and Adler-32 are supported. The checksum type can be changed on
2503 the fly via /proc/fs/lustre/osc/*/checksum_type.
2507 Description: use adler32 for page checksums
2508 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2513 Description: build for x2 fails
2514 Details : fix compile issue on Cray systems.
2518 Description: Properly match for duplicate locks
2519 Details : Due to different lock order from skiplists code, we need to
2520 traverse entire list for now
2523 Frequency : only on PPC/SLES10
2525 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2526 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2527 be defined. Add a hack in mkfs_lustre.c to work around this compile
2532 Description: LASSERT on MDS when client holding flock lock dies
2533 Details : ldlm pool logic depends on number of granted locks equal to
2534 number of released locks which is not true for flock locks, so
2535 just exclude such locks from consideration.
2539 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2540 Details : Use fsfilt_sync() for both the whole filesystem sync and
2541 individual file sync to eliminate dangerous inode locking
2542 with I_LOCK that can lead to a deadlock.
2546 Description: performance in 1.6.3
2547 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2548 sure that 1MB requests are not fragmented by the block layer.
2550 Severity : enhancement
2552 Description: organize the server-side client stats on per-nid basis
2553 Details : Change the structure of stats under obdfilter and mds to
2561 The "uuid"s file would list the uuids of _active_ exports.
2562 And the clear entry is to clear all stats and stale nids.
2564 Severity : enhancement
2566 Description: eliminate client locks in face of contention
2567 Details : file contention detection and lockless i/o implementation
2568 for contended files.
2572 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2575 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2577 Description: MDS may lose file striping (and hence file data) in some cases
2578 Details : If there are additional extended attributes stored on the MDS,
2579 in particular ACLs, SELinux, or user attributes (if user_xattr
2580 is specified for the client mount options) then there is a risk
2581 of attribute loss. Additionally, the Lustre file striping
2582 needs to be larger than default (e.g. striped over all OSTs),
2583 and an additional attribute must be stored initially in the
2584 inode and then increase in size enough to be moved to the
2585 external attribute block (e.g. ACL growing in size) for file
2590 Description: skiplist implementation simplification
2591 Details : skiplists are used to group compatible locks on granted list
2592 that was implemented as tracking first and last lock of each
2593 lock group the patch changes that to using doubly linked lists
2597 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2598 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2599 because LASSERT on both the data supplied by a client, and the
2600 data on disk is dangerous and incorrect.
2602 Severity : enhancement
2604 Description: Slow truncate/writes to huge files at high offsets.
2605 Details : Directly associate cached pages to lock that protect those pages,
2606 this allows us to quickly find what pages to write and remove
2607 once lock callback is received.
2611 Description: more ldlm soft lockups
2612 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2613 starve other threads from the resource lock for a long time in
2614 case of long waiting queue, so change the debug level from
2615 D_OTHER to the less frequently used D_INFO.
2617 Severity : enhancement
2619 Description: add -gid, -group, -uid, -user options to lfs find
2623 Description: Hung threads in invalidate_inode_pages2_range
2624 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2625 once one is completed. As a result, some RPCs are stuck in the
2626 queue and are never sent.
2630 Description: filter threads hungs on waiting journal commit
2631 Details : Cleanup filter group llog code, then only filter group llog will
2632 be only created in the MDS/OST syncing process.
2636 Description: Procfs and llog threads access destoryed import sometimes.
2637 Details : Sync the import destoryed process with procfs and llog threads by
2638 the import refcount and semaphore.
2640 Severity : enhancement
2642 Description: openlock cache of b1_6 port to HEAD
2647 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2648 Details : If insertion of an extent fails, then discard the inode
2649 preallocation and free data blocks else it can lead to duplicate
2654 Description: don't always update ctime in ext3_xattr_set_handle()
2655 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2656 In some cases the ctime should not be updated, for example for
2657 2.0->1.8 compatibility it is necessary to delete an xattr and it
2658 should not update the ctime.
2662 Bugzilla : 15713/16362
2663 Description: Assertion in iopen_connect_dentry in 1.6.3
2664 Details : looking up an inode via iopen with the wrong generation number can
2665 populate the dcache with a disconneced dentry while the inode
2666 number is in the process of being reallocated. This causes an
2667 assertion failure in iopen since the inode's dentry list contains
2668 both a connected and disconnected dentry.
2672 Description: assertion failure in ldlm_handle2lock()
2673 Details : fix a race between class_handle_unhash() and class_handle2object()
2674 introduced in lustre 1.6.5 by bug 13622.
2679 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2680 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2685 Description: oops in sd_iostats_seq_show()
2686 Details : unloading/reloading the scsi low level driver triggers a kernel
2687 bug when trying to access the sd iostat file.
2692 Description: Kernel panics during QLogic driver reload
2693 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2694 patch, causing memory corruption.
2699 Description: journal_dev option does not work in b1_6
2700 Details : pass mount option during pre-mount.
2702 Severity : enhancement
2704 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2705 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2706 extent information of a file. It can be used to map logical blocks
2707 in a file to physical blocks in the block device.
2711 Description: LDLM soft lockups - improvement
2712 Details : It is be possible to send the lock handle along with each read
2713 or write request because the client is already doing a lock match
2714 itself so there isn't any reason the OST should have to re-do that
2718 Frequency : only on Cray X2
2720 Description: X2 build failures
2721 Details : fix build failures on Cray X2.
2725 Description: xid & resent requests
2726 Details : Initialize RPC XID from clock at startup (randomly if clock is
2729 Severity : enhancement
2731 Description: Add lustre_start utility to start or stop multiple Lustre servers
2736 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2737 Details : In case of memory pressure, list_del() can be called twice on
2738 req->rq_history_list, causing a kernel oops.
2742 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2743 Details : fix stack overflow in the distributed lock manager by defering
2744 export eviction after a failed AST to the elt thread instead of
2745 handling it in the dlm interpret routine.
2749 Description: Convert some comments to new format.
2750 Details : Update documenting comments to match doxygen conventions.
2754 Description: Grammar fixes.
2755 Details : A couple of trivial spelling fixes.
2759 Description: OSD_COUNTERS-mandatory
2760 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2761 to simplify the code.
2765 Description: simplify cmm_device freeing logic.
2766 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2770 Description: Add lockdep support to dt_object_operations locking interface.
2771 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2772 parameter indicating lock ordering. Update mdd code to use new
2777 Description: Introduce failloc constants for lockless IO tests.
2778 Details : Add two new failloc constants to test lockless IO. Only one of
2779 them in implemented---another is checked in yet to be landed
2784 Description: Add lockdep support for inode mutex.
2785 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2786 in the situations where Lustre has to lock more than one inode
2791 Description: Add optional invariants checking support.
2792 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2793 configure switch. This macro is to be used for consistency and
2794 sanity checks that are too expensive to be left in `production'
2799 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2800 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2805 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2806 Details : Use spin_lock_nested() in (the only) situation where more than
2807 one ldlm_lock is locked simultaneously. Also, fix possible
2808 dead-lock in ldlm_lock_change_resource() by enforcing particular
2813 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2814 Details : Instead of specifying each ldlm_lock call-back through separate
2815 parameter, wrap them into struct ldlm_callback_suite.
2819 Description: Kill join_lru obd method and its callers.
2820 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2821 in memory---a requirement implemented through ->o_join_lru() obd
2822 method in HEAD. Remove this method and its users.
2826 Description: Add asynchronous ldlm ENQUEUE completion handler.
2827 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2828 case that stock ldlm_completion_ast() cannot handle as it waits
2829 until lock is granted. Introduce new ldlm_completion_ast_async()
2830 for this. Also comment ldlm_completion_ast().
2834 Description: ldlm_error <-> errno conversion.
2835 Details : Add functions to map (rather arbitrary) between LDLM error codes
2836 and standard errno values. CLIO needs this to prevent LDLM specific
2837 constants from escaping ldlm and osc.
2841 Description: Kill unused ldlm_handle2lock_ns() function.
2842 Details : Kill unused ldlm_handle2lock_ns() function.
2846 Description: Add lu_ref support to ldlm_lock
2847 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2848 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2849 and ldlm_resource. LDLM interface has to be changed, because code that
2850 releases a reference on a lock, has to "know" what reference this is.
2851 In the most frequent case
2853 lock = ldlm_handle2lock(handle);
2855 LDLM_LOCK_PUT(lock);
2857 no changes are required. When any other reference (received _not_ from
2858 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2859 instead of LDLM_LOCK_PUT().
2861 Arguably, changes are pervasive, and interface requires some discipline
2862 for proper use. On the other hand, it was very instrumental in finding
2863 a few leaked lock references.
2867 Description: Add ldlm_lock_addref_try().
2868 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2869 attempts to addref a lock that might be being canceled
2874 Description: Add ldlm_weigh_callback().
2875 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2876 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2877 hard-coded `number of pages' logic.
2881 Description: Add lockdep annotations to llog code.
2882 Details : Use appropriately tagged _nested() locking calls in the places
2883 where llog takes more than one ->lgh_lock lock.
2887 Description: Add loi_kms_set().
2888 Details : Wrap kms updates into a helper function.
2892 Description: Constify instances of struct lsm_operations.
2893 Details : Constify instances of struct lsm_operations.
2897 Description: lu_conf support.
2898 Details : On a server, a file system object is uniquely identified
2899 by a fid, which is sufficient to locate and load all object
2900 state (inode). On a client, on the other hand, more data are
2901 necessary instantiate an object. Change lu_object_find() and
2902 friends to take additional `lu_conf' argument describing object.
2903 Typically this includes layout information.
2907 Description: lu_context fixes.
2908 Details : Introduce new lu_context functions that are needed on the client
2909 side, where some system threads (ptlrpcd) are shared by multiple
2910 modules, and so cannot be stopped during module shutdown.
2914 Description: Add start and stop methods to lu_device_type_operations.
2915 Details : Introduce two new methods in lu_device_type_operations, that are
2916 invoked when first instance of a given type is created and last one
2917 is destroyed respectively. This is need by CLIO.
2921 Description: Add lu_ref support to struct lu_device.
2922 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2923 track leaked references.
2927 Description: Introduce lu_kmem_descr.
2928 Details : lu_kmem_descr and its companion interface allow to create
2929 and destroy a number of kmem caches at once.
2933 Description: Fix lu_object finalization race.
2934 Details : Fix a race between lu_object_find() finding an object and its
2935 concurrent finalization. This race is (most likely) not possible
2936 on the server, but might happen on the client.
2940 Description: Introduce lu_ref interface.
2941 Details : lu_ref is a debugging module allowing to track references to
2942 a given object. It is quite cpu expensive, and has to be
2943 explicitly enabled with --enable-lu_ref. See usage description
2948 Description: Factor lu_site procfs stats into a separate function.
2949 Details : Separate lu_site stats printing code into a separate function
2950 that can be reused on a client.
2954 Description: Constify instances of struct {lu,dt,md}_device_operations.
2955 Details : Constify instances of struct {lu,dt,md}_device_operations.
2959 Description: Introduce struct md_site and move meta-data specific parts of
2960 struct lu_site here.
2961 Details : Move md-specific fields out of struct lu_site into special struct
2962 md_site, so that lu_site can be used on a client.
2966 Description: Kill mdd_lov_destroy().
2967 Details : Remove unused mdd code.
2971 Description: Add st_block checking to multistat.c.
2972 Details : Add st_block checking to multistat.c.
2976 Description: Add lu_ref support to struct obd_device.
2977 Details : Add lu_ref tracking to obd_device.
2981 Description: Kill obd_set_fail_loc().
2982 Details : Remove unused code.
2986 Description: Add special type for ptlrpc_request interpret functions.
2987 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2988 there. Actual usage will be in CLIO.
2992 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2993 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2994 doesn't work with lockdep.
2998 Description: Add rwv.c test program.
2999 Details : New testing program exercising readv(2) and writev(2) (Qian).
3003 Description: Add sendfile.c test program.
3004 Details : New testing program exercising sendfile(2) (Jay).
3008 Description: Ratelimit a message that can be very frequent.
3009 Details : Ratelimit a memory allocation failure message that can
3014 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3015 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3019 Description: Liblustre build fixes.
3020 Details : Liblustre build fixes.
3024 Description: libcfs: add cfs_{need,cond}_resched() interface.
3025 Details : libcfs: add cfs_{need,cond}_resched() definition and
3026 implementations for Linux, NT, and liblustre.
3028 Severity : enhancement
3030 Description: More exported tunables for mballoc
3031 Details : Add support for tunable preallocation window and new tunables for
3032 large/small requests
3036 Description: Detect corruption of block bitmap and checking for preallocations
3037 Details : Checks validity of on-disk block bitmap. Also it does better
3038 checking of number of applied preallocations. When corruption is
3039 found, it turns filesystem readonly to prevent further corruptions.
3043 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3044 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3045 window to 0. If it does hit cache miss, and read-pattern is still
3046 stride-io mode, does not reset the stride window, but also does not
3047 increase the stride window length in this case.
3051 Frequency : only for big-endian servers
3052 Description: Check if system is big-endian while mounting fs with extents feature
3053 Details : Mounting a filesystem with extents feature will fail on big-endian
3054 systems since ext3-based ldiskfs is not supported on big-endian
3055 systems. This can be over-riden with "bigendian_extents" mount option.
3057 Severity : enhancement
3059 Description: The root squash functionality
3060 Details : A security feature, which is to prevent users from being able
3061 to mount lustre on their desktop, run as root, and delete
3062 all of the files in the filesystem. The goal is accomplished by
3063 remapping user id (UID) and group id (GID) of the root user to
3064 a UID and GID specified by the system administartor via Lustre
3065 configuration management server (MGS). The functionality also
3066 allows to specify sets of clients for which the remapping does
3071 Description: Excessive recovery window
3072 Details : With AT enabled, the recovery window can be excessively long (6000+
3073 seconds). To address this problem, we no longer use
3074 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3075 timeout no longer depends on the service time, it is set to
3076 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3077 time via pb_service_time.
3081 Description: Watchdog triggered on MDS failover
3082 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3083 the OSTs know that the MDS "UUID" can be reused for the same export
3084 from a different NID, so we do not need to wait for the export to be
3088 Frequency : rare, only if using MMP with Linux RAID
3090 Description: MMP doesn't work with Linux RAID
3091 Details : While using HA for Lustre servers with Linux RAID, it is possible
3092 that MMP will not detect multiple mounts. To make this work we
3093 need to unplug the device queue in RAID when the MMP block is being
3094 written. Also while reading the MMP block, we should read it from
3095 disk and not the cached one.
3097 Severity : enhancement
3099 Description: open file using fid
3100 Details : A file can be opened using just its fid, like
3101 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3104 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3107 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3108 Details : add proper locking for creation of jbd_4k slab cache
3112 Description: MMP check in ext3_remount() fails without displaying any error
3113 Details : When multiple mount protection fails during remount, proper error
3116 Severity : enhancement
3118 Description: Allow stripe size to be up to 4G-64k
3119 Details : Fix math logic to allow large stripe sizes.
3123 Description: add check for >8TB ldiskfs filesystems
3124 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3125 Don't allow >8TB ldiskfs filesystems to be mounted without
3126 force_over_8tb mount option
3131 Descriptoin: OSS complains "too many blocks" and crashes
3132 Details : In filter_commitrw_write():
3133 1. lquota_chkquota();
3134 2. obdo_from_inode();
3135 3. lquota_pending_commit();
3136 "2." has changed uid and gid in this case, then caused
3137 lquota_chkquota() and lquota_pending_commit() to use
3138 different uid and gid. That is the root of the bug.
3142 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3143 Details : Detect and print out the 'stripe count = -1' case properly.
3147 Description: Out or order replies might be lost on replay
3148 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3149 request with tid smaller than one currently being added, add it
3150 to the start, not end of the list.
3152 --------------------------------------------------------------------------------
3154 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3156 * Support for kernels:
3157 2.6.5-7.283 (SLES 9),
3158 2.6.9-55.EL (RHEL 4),
3159 2.6.16.46-0.14 (SLES 10),
3160 2.6.18.8 vanilla (kernel.org)
3161 * Client support for unpatched kernels:
3162 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3163 2.6.16 - 2.6.22 vanilla (kernel.org)
3164 * Due to recently discovered recovery problems, we do not recommend
3165 using patchless RHEL 4 clients with this or any earlier release.
3166 * Recommended e2fsprogs version: 1.39.cfs8
3167 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3168 * Starting with this release, the ldiskfs backing filesystem required
3169 by Lustre is now in its own package, lustre-ldiskfs. This package
3170 should be installed. It is versioned separately from Lustre and
3171 may be released separately in future.
3175 Description: block reactivating mgc import until all deactivates complete
3176 Details : Fix race when failing back MDT/MGS to itself (testing)
3179 Frequency : at statup only
3181 Description: mds_lov_synchronize race leads to various problems
3182 Details : simultaneous MDT->OST connections at startup can cause the
3183 sync to abort, leaving the OSC in a bad state.
3185 Severity : enhancement
3187 Description: add optional extra BUILD_VERSION info
3188 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3189 to override the lustre version.
3192 Frequency : 2.6.18 servers only
3194 Description: ll_kern_mount() doesn't release the module reference
3195 Details : The ldiskfs module reference count never drops down to 0
3196 because ll_kern_mount() doesn't release the module reference.
3201 Description: server LBUG when using old ost_num_threads parameter
3202 Details : Accept the old ost_num_threads parameter but warn that it
3203 is deprecated, and fix an off-by-one error that caused an LBUG.
3208 Description: Transient SCSI error results in persistent IO issue
3209 Details : iobuf->dr_error is not reinitialized to 0 between two
3213 Frequency : sometimes when underlying device returns I/O errors
3215 Description: OSTs not going read-only during write failures
3216 Details : OSTs are not remounted read-only when the journal commit threads
3217 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3218 instead of the ext3 wrappers.
3222 Description: poor connect scaling with increasing client count
3223 Details : Don't run filter_grant_sanity_check for more than 100 exports
3224 to improve scaling for large numbers of clients.
3227 Frequency : SLES10 only
3229 Description: sanity-quota.sh quotacheck failed: rc = -22
3230 Details : Quotas cannot be enabled on SLES10.
3233 Frequency : liblustre clients only
3235 Description: getdirentries does not give error when run on compute nodes
3236 Details : getdirentries does not fail when the size specified as an argument
3237 is too small to contain at least one entry
3239 Severity : enhancement
3241 Description: Add LNET router traceability for debug purposes
3242 Details : If a checksum failure occurs with a router as part of the
3243 IO path, the NID of the last router that forwarded the bulk data
3244 is printed so it can be identified.
3249 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3250 Details : Due to a race condition, liblustre clients were occasionally
3251 evicted incorrectly.
3253 Severity : enhancement
3255 Description: lfs setstripe use optional parameters instead of postional
3258 Severity : enhancement
3260 Description: Nanosecond timestamp support for ldiskfs
3261 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3262 resolution timestamps. There is not yet support for this at
3263 the Lustre filesystem level.
3266 Frequency : during server recovery
3268 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3269 Details : request with rq_no_resend flag not awake l_wait_event if they get
3273 Frequency : nfs export on patchless client
3275 Description: connectathon hang when test nfs export over patchless client
3276 Details : Disconnected dentry cannot be found with lookup, so we do not need
3277 to unhash it or make it invalid
3280 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3281 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3282 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3287 Description: open req refcounting wrong on reconnect
3288 Details : If reconnect happened between getting open reply from server and
3289 call to mdc_set_replay_data in ll_file_open, we will schedule
3290 replay for unreferenced request that we are about to free.
3291 Subsequent close will crash in variety of ways.
3292 Check that request is still eligible for replay in
3293 mdc_set_replay_data().
3298 Description: disable writes to filesystem when reading health_check file
3299 Details : the default for reading the health_check proc file has changed
3300 to NOT do a journal transaction and write to disk, because this
3301 can cause reads of the /proc file to hang and block HA state
3302 checking on a healthy but otherwise heavily loaded system. It
3303 is possible to return to the previous behaviour during configure
3304 with --enable-health-write.
3306 Severity : enhancement
3308 Description: 64-bit inode version
3309 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3310 to the inode. This will be required for version-based recovery.
3315 Description: MDS fails to start if a duplicate client export is detected
3316 Details : in some rare cases it was possible for a client to connect to
3317 an MDS multiple times. Upon recovery the MDS would detect this
3318 and fail during startup. Handle this more gracefully.
3320 Severity : enhancement
3322 Description: Add -o localflock option to simulate old noflock
3324 Details : This will achieve local-only flock/fcntl locks
3330 Description: log_commit_thread vs filter_destroy race leads to crash
3331 Details : Take import reference before releasing llog record semaphore
3336 Description: Wrong request locking in request set processing
3337 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3338 imp_delayed_list, in this place should be used imp_lock.
3341 Frequency : when reconnection
3343 Description: Grant Leak when osc reconnect to OST
3344 Details : When osc reconnect ost, OST(filter) should check whether it
3345 should grant more space to client by comparing fed_grant and
3346 cl_avail_grant, and return the granted space to client instead
3347 of "new granted" space, because client will call osc_init_grant
3348 to update the client grant space info.
3351 Frequency : when client reconnect to OST
3353 Description: Grant Leak when osc do resend and replay bulk write
3354 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3355 bulk write request, because the grant info while be sync between
3356 OSC and OST when reconnect, and we should ignore the grant info
3357 these of resend/replay write req.
3362 Description: Grant space more than available left space sometimes.
3363 Details : When then OST is about to be full, if two bulk writing from
3364 different clients came to OST. Accord the avaliable space of the
3365 OST, the first req should be permitted, and the second one
3366 should be denied by ENOSPC. But if the seconde arrived before
3367 the first one is committed. The OST might wrongly permit second
3368 writing, which will cause grant space > available space.
3371 Frequency : when client is evicted
3373 Description: Grant might be wrongly erased when osc is evicted by OST
3374 Details : when the import is evicted by server, it will fork another
3375 thread ptlrpc_invalidate_import_thread to invalidate the
3376 import, where the grant will be set to 0. While the original
3377 thread will update the grant it got when connecting. So if
3378 the former happened latter, the grant will be wrongly errased
3379 because of this race.
3384 Description: Checking Stale with correct fid
3385 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3386 to check whether it is stale, because sometimes, we want the
3387 enqueue happened anyway, and op_data.fid2 will not be initialized.
3389 Severity : enhancement
3391 Description: update patchless client
3392 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3395 Frequency : only with 2.4 kernel
3397 Description: random memory corruption
3398 Details : size of struct ll_inode_info is to big for union inode.u and this
3399 can be cause of random memory corruption.
3404 Description: Memory leak in recovery
3405 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3406 It should also check obd_fail before fsfilt_start, otherwise if
3407 fsfilt_start return -EROFS,(failover mds during mds recovery).
3408 then the req will return with repmsg->transno = 0 and rc = EROFS.
3409 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3410 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3411 be freed no matter whether fsfilt_commit success or not.
3414 Frequency : only with huge count clients
3416 Description: Prevents from taking the superblock lock in llap_from_page for
3418 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3419 for llap_from_page call in ll_removepage prevents from taking the
3420 superblock lock for a soon died page.
3425 Description: Not check open intent error before release open handle
3426 Details : in some rare cases, the open intent error is not checked before
3427 release open handle, which may cause
3428 ASSERTION(open_req->rq_transno != 0), because it tries to release
3429 the failed open handle.
3434 Description: Set cat log bitmap only after create log success.
3435 Details : in some rare cases, the cat log bitmap is set too early. and it
3436 should be set only after create log success.
3440 Description: Accessing a block bevice can re-enable I/O when Lustre is
3441 tearing down a device.
3442 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3447 Description: service threads may hog cpus when there are a lot of requests
3448 Details : Insert cond_resched to give other threads a chance to use some CPU
3453 Description: the cat log was not initialized in recovery
3454 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3455 unlink log on mds will not be initialized until mds post
3456 recovery. And also in mds post recovery, the unlink log will
3457 initialization will be done asynchronausly, so there will be race
3458 between add unlink log and unlink log initialization.
3462 Description: brw_stats were being printed incorrectly
3463 Details : brw_stats were being printed as log2 but all of them were not
3464 recorded as log2. Also remove some code duplication arising from
3465 filter_tally_{read,write}.
3469 Frequency : rare, only in recovery.
3470 Description: ASSERTION(req->rq_type != LI_POISON) failed
3471 Details : imp_lock should be held while iterating over imp_sending_list for
3472 prevent destroy request after get timeout in ptlrpc_queue_wait.
3476 Description: replay-single.sh test 52 fails
3477 Details : A lock's skiplist need to be cleanup when it being unlinked
3478 from its resource list.
3482 Description: Short directio read returns full requested size rather than
3484 Details : Direct I/O operations should return actual amount of bytes
3485 transferred rather than requested size.
3487 Severity : enhancement
3489 Description: metadata RPC reduction (e.g. for rm performance)
3490 Details : decrease the amount of synchronous RPC between clients and servers
3491 by canceling conflicing lock before the operation on the client
3492 and packing thier handles into the main operation RPC to server.
3494 Severity : enhancement
3496 Description: Async OSC create to avoid the blocking unnecessarily.
3497 Details : If an OST has no remaining object, system will block on the
3498 creation when it needs to create a new object on this OST. Now,
3499 ways use pre-created objects when available, instead of blocking on
3500 an empty osc while others are not empty. If we must block, we
3501 block for the shortest possible period of time.
3505 Description: improve handling recoverable errors
3506 Details : if request processig with error which can be recoverable on server
3507 request should be resend, otherwise page released from cache and
3510 Severity : enhancement
3512 Description: refine locking for avoid write wrong info into lov_objid file
3513 Details : fix possible races with add new target and write/update data in
3516 --------------------------------------------------------------------------------
3518 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3523 Frequency : on some architectures
3525 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3526 Details : Uninitialized flags sometimes cause configuration commands to
3530 Frequency : patchless clients only
3532 Description: missing __iget() symbol export
3533 Details : The __iget() symbol export is missing. To avoid the need for
3534 this on patchless clients the deathrow inode reaper is turned
3535 off, and we depend on the VM to clean up old inodes. This
3536 dependency was during via the fix for bug 12181.
3538 --------------------------------------------------------------------------------
3540 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3542 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3543 INTEROPERATE with older versions automatically. In many cases a
3544 special upgrade step is needed. Please read the
3545 user documentation before upgrading any part of a live system.
3546 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3547 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3548 * WARNING: Lustre configuration and startup changes are required with
3549 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3553 Severity : enhancement
3555 Description: MountConf
3556 Details : Lustre configuration is now managed via mkfs and mount
3557 commands instead of lmc and lconf. New obd types (MGS, MGC)
3558 are added for dynamic configuration management. See
3559 https://mail.clusterfs.com/wikis/lustre/MountConf for
3562 Severity : enhancement
3564 Description: dynamic OST addition
3565 Details : OSTs can now be added to a live filesystem
3567 Severity : enhancement
3569 Description: startup order invariance
3570 Details : MDTs and OSTs can be started in any order. Clients only
3571 require the MDT to complete startup.
3573 Severity : enhancement
3575 Description: parallel, asynchronous orphan cleanup
3576 Details : orphan cleanup is now performed in separate threads for each
3577 OST, allowing parallel non-blocking operation.
3579 Severity : enhancement
3581 Description: optimized stripe assignment
3582 Details : stripe assignments are now made based on ost space available,
3583 ost previous usage, and OSS previous usage, in order to try
3584 to optimize storage space and networking resources.
3586 Severity : enhancement
3588 Description: Permanently set tunables
3589 Details : All writable /proc/fs/lustre tunables can now be permanently
3590 set on a per-server basis, at mkfs time or on a live system.
3592 Severity : enhancement
3594 Description: Lustre message v2
3595 Details : Add lustre message format v2.
3597 Severity : enhancement
3599 Description: client OST exclusion list
3600 Details : Clients can be started with a list of OSTs that should be
3601 declared "inactive" for known non-responsive OSTs.
3605 Description: SPEC SFS validation failure on NFS v2 over lustre.
3606 Details : Changes the blocksize for regular files to be 2x RPC size,
3607 and not depend on stripe size.
3609 Severity : enhancement
3611 Description: Multiple MD RPCs in flight.
3612 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3613 To avoid overly-overloading MDS, introduce a limit on number of
3614 MDS RPCs in flight for a single client and add /proc controls
3615 to adjust this limit.
3617 Severity : enhancement
3619 Description: client read/write statistics
3620 Details : Add client read/write call usage stats for performance
3621 analysis of user processes.
3622 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3623 file access. extents_stats shows chunk size distribution.
3624 extents_stats_per_process show chunk size distribution per
3627 Severity : enhancement
3629 Description: mds statistics
3630 Details : Add detailed mds operations statistics in
3631 /proc/fs/lustre/mds/*/stats.
3635 Description: Failure of copying files with lustre special EAs.
3636 Details : Client side always return success for setxattr call for lustre
3637 special xattr (currently only "trusted.lov").
3642 Description: Refcount LNET uuids
3643 Details : The global LNET uuid list grew linearly with every startup;
3644 refcount repeated list entries instead of always adding to
3647 Severity : enhancement
3649 Description: Dynamic service threads
3650 Details : Within a small range, start extra service threads
3651 automatically when the request queue builds up.
3654 Frequency : mixed-endian client/server environments
3656 Description: mixed-endian crashes
3657 Details : The new msg_v2 system had some failures in mixed-endian
3660 Severity : enhancement
3662 Description: Easy OST removal
3663 Details : OSTs can be permanently deactivated with e.g. 'lctl
3664 conf_param lustre-OST0001.osc.active=0'
3666 Severity : enhancement
3668 Description: MGS proc entries
3669 Details : Added basic proc entries for the MGS showing what filesystems
3672 Severity : enhancement
3674 Description: provide MGS failover
3675 Details : Added config lock reacquisition after MGS server failover.
3677 Severity : enhancement
3679 Description: add Linux 2.4 support
3680 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3684 Description: a large application tries to do I/O to the same resource and dies
3685 in the middle of it.
3686 Details : Check the req->rq_arrival time after the call to
3687 ost_brw_lock_get(), but before we do anything about
3688 processing it & sending the BULK transfer request. This
3689 should help move old stale pending locks off the queue as
3690 quickly as obd_timeout.
3693 Frequency : when an incorrect nid is specified during startup
3695 Description: ptlrpc connect to non-existant node causes kernel crash
3696 Details : LNET can't be re-entered from an event callback, which
3697 happened when we expire a message after the export has been
3698 cleaned up. Instead, hand the zombie cleanup off to another
3701 Severity : enhancement
3703 Description: plain/inodebits lock performance improvement
3704 Details : Grouping plain/inodebits in granted list by their request modes
3705 and bits policy, thus improving the performance of search through
3709 Frequency : only if OST filesystem is corrupted
3711 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3712 Details : for a short time RPCs with bulk IO are in the replay list,
3713 but replay of bulk IOs is unimplemented. If the OST filesystem
3714 is corrupted due to disk cache incoherency and then replay is
3715 started it is possible to trip an assertion. Avoid putting
3716 committed RPCs into the replay list at all to avoid this issue.
3719 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3721 Description: System hang on startup
3722 Details : This bug allowed the liblustre (e.g. catamount) client to
3723 return to the app before handling all startup RPCs. This
3724 could leave the node unresponsive to lustre network traffic
3725 and manifested as a server ptllnd timeout.
3727 Severity : enhancement
3729 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3730 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3731 boolean to control whether to print peer debug info when a
3732 client's RPC times out.
3735 Frequency : only for kernels with patches from Lustre below 1.4.3
3737 Description: Remove old rdonly API
3738 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3741 Frequency : only for devices with external journals
3743 Description: Set external device read-only also
3744 Details : During a commanded failover stop, we set the disk device
3745 read-only while the server shuts down. We now also set any
3746 external journal device read-only at the same time.
3749 Frequency : when upgrading from 1.4 while trying to change parameters
3751 Description: The wrong (new) MDC name was used when setting parameters for
3752 upgraded MDT's. Also allows changing of OSC (and MDC)
3753 parameters if --writeconf is specified at tunefs upgrade time.
3756 Frequency : when setting specific ost indicies
3758 Description: QOS code breaks on skipped indicies
3759 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3760 created with --index need not be sequential.
3762 Severity : enhancement
3764 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3765 Details : The uninit_groups feature works in conjunction with the kernel
3766 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3767 up the pass1 processing of e2fsck. This is a read-only feature
3768 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3769 mount filesystems that have had this feature enabled.
3771 Severity : enhancement
3773 Description: Improve multi-block allocation algorithm to avoid fragmentation
3774 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3775 allocation locality and avoid filesystem fragmentation.
3777 ------------------------------------------------------------------------------
3779 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3781 * Support for kernels:
3782 2.4.21-47.0.1.EL (RHEL 3)
3783 2.6.5-7.283 (SLES 9)
3784 2.6.9-42.0.10.EL (RHEL 4)
3785 2.6.12.6 vanilla (kernel.org)
3786 2.6.16.27-0.9 (SLES 10)
3787 * Recommended e2fsprogs version: 1.39.cfs5
3789 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3793 Frequency : occasional, depends on client load and configuration
3794 Bugzilla : 12181, 12203
3795 Description: data loss for recently-modified files
3797 Details : In some cases it is possible that recently written or created
3798 files may not be written to disk in a timely manner (this should
3799 normally be within 30s unless client IO load is very high).
3800 The problem appears as zero-length files or files that are a
3801 multiple of 1MB in size after a client crash or client eviction
3802 that are missing data at the end of the file.
3804 This problem is more likely to be hit on clients where files are
3805 repeatedly created and unlinked in the same directory, clients
3806 have a large amount of RAM, have many CPUs, the filesystem has
3807 many OSTs, the clients are rebooted frequently, and/or the files
3808 are not accessed by other nodes after being written.
3810 The presence of the problem can be detected by looking at
3811 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3812 smaller than the second (nr_unused) then dirty files will not
3813 be flushed automatically to disk. "sync; sleep 10" should be
3814 run several times on the node before unmounting it to update
3815 Lustre (this is also safe to run on nodes without this problem).
3817 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3818 that can cause this same problem, so customers using that kernel
3819 also need to update the kernel in addition to Lustre. In order
3820 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3822 It is normal that files written just before a client crash (less
3823 than 30s) may not yet have been flushed to disk, even for local
3827 Frequency : frequent on thin XT3 nodes
3829 Description: UUID collision on thin XT3 Linux nodes
3830 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3831 randomly, since we relied on an insufficiently-seeded PRNG.
3836 Description: OSS hangs after "All ost request buffers busy"
3837 Details : A deadlock between quota and journal operations caused OSS
3838 hangs after printing "All ost request buffers busy."
3841 Frequency : always on liblustre builds
3843 Description: Cleanup compiler warnings on liblustre
3846 Frequency : always on liblustre builds on XT3
3848 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3852 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3853 > 32000 subdirectories
3854 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3855 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3856 subdirectory count crosses 32000. This will aid e2fsck to
3857 correctly handle more than 32000 subdirectories.
3860 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3862 Description: System hang on startup
3863 Details : This bug allowed the liblustre (e.g. catamount) client to
3864 return to the app before handling all startup RPCs. This
3865 could leave the node unresponsive to lustre network traffic
3866 and manifested as a server ptllnd timeout.
3868 Severity : enhancement
3870 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3871 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3872 boolean to control whether to print peer debug info when a
3873 client's RPC times out.
3878 Description: make O_SYNC working on 2.6 kernels
3879 Details : 2.6 kernels use different method for mark pages for write,
3880 so need add a code to lustre for O_SYNC work.
3885 Description: Failure to close file and release space on NFS
3886 Details : Put inode details into lock acquired in ll_intent_file_open.
3887 Use mdc_intent_lock in ll_intent_open to properly
3888 detect all kind of errors unhandled by mdc_enqueue.
3893 Description: proc file read during shutdown sometimes raced obd removal,
3895 Details : Add lock to prevent obd access after proc file removal.
3898 Frequency : Only for files larger than 4GB on 32-bit clients.
3900 Description: improperly doing page alignment of locks
3901 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3902 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3907 Description: RPCs being resent when they shouldn't be
3908 Details : Some RPCs that should not be resent are being resent. This
3909 can cause inconsistencies in the RPC state machine. Do not
3910 resend such requests.
3913 Frequency : rare, only with NFS export
3915 Description: Crash on NFS re-export node
3916 Details : under very unusual load conditions an assertion is hit in
3917 ll_intent_file_open()
3920 Frequency : only if OST filesystem is corrupted
3922 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3923 Details : for a short time RPCs with bulk IO are in the replay list,
3924 but replay of bulk IOs is unimplemented. If the OST filesystem
3925 is corrupted due to disk cache incoherency and then replay is
3926 started it is possible to trip an assertion. Avoid putting
3927 committed RPCs into the replay list at all to avoid this issue.
3932 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3933 Details : Large single O_DIRECT read and write calls can fail to allocate
3934 a sufficiently large buffer to process the request. In case of
3935 allocation failure the allocation is retried with a smaller
3936 buffer and broken into smaller requests.
3938 Severity : enhancement
3940 Description: Add -o localflock option to simulate old noflock behaviour.
3941 Details : This will achieve local-only flock/fcntl locks coherentness.
3946 Description: versioning check is incomplete
3947 Details : Checking the version difference of client vs. server, report
3948 error if the gap is too big.
3953 Description: add support PG_writeback bit
3954 Details : add support for PG_writeback bit for Lustre, for more carefull
3955 work with page cache in 2.6 kernel. This also fix some deadlocks
3956 and remove hack for work O_SYNC with 2.6 kernel.
3958 Severity : enhancement
3960 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3961 Details : The uninit_groups feature works in conjunction with the kernel
3962 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3963 up the pass1 processing of e2fsck. This is a read-only feature
3964 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3965 mount filesystems that have had this feature enabled.
3967 Severity : enhancement
3969 Description: Improve multi-block allocation algorithm to avoid fragmentation
3970 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3971 allocation locality and avoid filesystem fragmentation.
3973 ------------------------------------------------------------------------------
3975 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3977 * Support for kernels:
3978 2.6.9-42.0.3.EL (RHEL 4)
3979 2.6.5-7.276 (SLES 9)
3980 2.4.21-47.0.1.EL (RHEL 3)
3981 2.6.12.6 vanilla (kernel.org)
3982 2.6.16.21-0.8 (SLES10)
3983 * Recommended e2fsprogs version: 1.39.cfs2-0
3985 * The backwards-compatible /proc/sys/portals symlink has been removed
3986 in this release. Before upgrading, please ensure that you change
3987 any configuration scripts or /etc/sysctl.conf files that access
3988 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3989 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3990 in advance of the upgrade on any system running Lustre 1.4.6 or
3991 newer, since /proc/sys/lnet was added in that version.
3992 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3996 Frequency : only when quota is used
3998 Description: avoid scanning export list for quota master
3999 Details : Change the algorithms to avoid scanning export list in order
4000 to improve the efficiency.
4003 Frequency : MDS failover only, very rarely
4005 Description: "went back in time" messages on mds failover
4006 Details : The greatest transno may be lost when the current operation
4007 finishes with an error (transno==0) and the client's last_rcvd
4008 record is over-written. Save the greatest transno in the
4009 mds_last_transno for this case.
4012 Frequency : always for specific kernels and striping counts
4014 Description: client may get "Matching packet too big" without ACL support
4015 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4016 when trying to access files in certain configurations. The
4017 clients should in fact be denied when mounting because they do
4018 not understand ACLs.
4021 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4023 Description: many clients connecting with IO in progress causes connect timeouts
4024 Details : Avoid synchronous journal commits to avoid delays caused by many
4025 clients connecting/disconnecting when bulk IO is in progress.
4026 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4027 OST_IO_PORTAL to avoid delays behind potentially many pending
4031 Frequency : occasionally with multiple writers to a single file
4033 Description: shared writes to file may result in wrong size reported by stat()
4034 Details : Allow growing of kms when extent lock is cancelled
4037 Frequency : always with random mmap IO to multi-striped file
4039 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4040 Details : Only if the hole is at the end of OST object so that kms is too
4041 small. Fix is to increase kms accordingly in ll_nopage.
4044 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4046 Description: writes to a missing object would leak memory on the OST
4047 Details : If there is an inconsistency between the MDS and OST filesystems,
4048 such that the MDS references an object that doesn't exist, writes
4049 to that object will leak memory due to incorrect cleanup in the
4050 error handling path, eventually running out of memory on the OST.
4055 Description: Creating too long symlink causes lustre errors
4056 Details : Check symlink and name lengths before sending requests to MDS.
4059 Frequency : only if flock is enabled (not on by default)
4061 Description: posix locks not released on fd closure on 2.6.9+
4062 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4063 kernels, this caused such locks not to be released on fd close and
4064 then assertions on fs unmount about still used locks.
4067 Frequency : MDS failover only, very rarely
4069 Description: clients may get ASSERTION(granted_lock != NULL)
4070 Details : When request was taking a long time, and a client was resending
4071 a getattr by name lock request. The were multiple lock requests
4072 with the same client lock handle and
4073 mds_getattr_name->fixup_handle_for_resent_request found one of the
4074 lock handles but later failed with ASSERTION(granted_lock != NULL).
4079 Description: handle->h_buffer_credits > 0, assertion failure
4080 Details : h_buffer_credits is zero after truncate, causing assertion
4081 failure. This patch extends the transaction or creates a new
4085 Frequency : NFS re-export or patchless client
4086 Bugzilla : 11179, 10796
4087 Description: Crash on NFS re-export node (__d_move)
4088 Details : We do not want to hash the dentry if we don't have a lock.
4089 But if this dentry is later used in d_move, we'd hit uninitialised
4090 list head d_hash, so we just do this to init d_hash field but
4091 leave dentry unhashed.
4094 Frequency : NFS re-export or patchless client
4096 Description: NFS exports has problem with symbolic link
4097 Details : lustre client didn't properly install dentry when re-exported
4098 to NFS or running patchless client.
4101 Frequency : NFS re-export or patchless client
4103 Description: Various nfs/patchless fixes.
4104 Details : fixes reuse disconected alias for lookup process - this fixes
4105 warning "find_exported_dentry: npd != pd",
4106 fix permission error with open files at nfs.
4107 fix apply umask when do revalidate.
4110 Frequency : occasional
4112 Description: Crash on NFS re-export node
4113 Details : calling clear_page() on the wrong pointer triggered oops in
4114 generic_mapping_read().
4117 Frequency : rarely, using O_DIRECT IO
4119 Description: unaligned directio crashes client with LASSERT
4120 Details : check for unaligned buffers before trying any requests.
4123 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4125 Description: stale data returned from RAID cache
4126 Details : If only a small amount of IO is done to the RAID device before
4127 reading it again it is possible to get stale data from the RAID
4128 cache instead of reading it from disk.
4131 Frequency : always for sles10 kernel
4133 Description: sles10 support
4134 Details : ll_follow_link: compile fixes and using of nd_set_link
4135 under newer kernels.
4138 Frequency : depends on arch, kernel and compiler version, always on sles10
4141 Description: recursive or deep enough symlinks cause stack overflow
4142 Details : getting rid of large stack-allocated variable in
4146 Frequency : depends on hardware
4148 Description: lustre write performance loss in the SLES10 kernel
4149 Details : the performance loss is caused by using of write barriers in the
4150 ext3 code. The SLES10 kernel turns barrier support on by
4151 default. The fix is to undo that change for ldiskfs.
4153 ------------------------------------------------------------------------------
4155 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4157 * Support for kernels:
4158 2.6.9-42.0.3EL (RHEL 4)
4159 2.6.5-7.276 (SLES 9)
4160 2.4.21-47.0.1.EL (RHEL 3)
4161 2.6.12.6 vanilla (kernel.org)
4165 Frequency : quota enabled and large files being deleted
4167 Description: releasing more than 4GB of quota at once hangs OST
4168 Details : If a user deletes more than 4GB of files on a single OST it
4169 will cause the OST to spin in an infinite loop. Release
4170 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4175 Description: statfs data retrieved from /proc may be stale or zero
4176 Details : When reading per-device statfs data from /proc, in the
4177 {kbytes,files}_{total,free,avail} files, it may appear
4178 as zero or be out of date.
4181 Frequency : systems with MD RAID1 external journal devices
4183 Description: lconf's call to blkid is confused by RAID1 journal devices
4184 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4185 of returning all block devices that match the journal UUID.
4188 Frequency : always, for aggregate stripe size over 4GB
4190 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4191 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4192 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4195 Frequency : always if "lfs find" used on a local file/directory
4197 Description: "lfs find" segfaults if used on a local file/directory
4198 Details : The case where a directory component was not specified wasn't
4199 handled correctly. Handle this properly.
4202 Frequency : always on ppc64
4204 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4205 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4206 perform endianness conversion.
4209 Frequency : rarely (truncate to non-zero file size after write under load)
4210 Bugzilla : 10730, 10687
4211 Description: Files padded with zeros to next 4K multiple
4212 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4213 it is possible that files that are truncated to a non-zero size
4214 immediately after being written are filled with zero bytes beyond
4215 the truncated size. No file data is lost.
4217 Severity : enhancement
4219 Description: Allow recovery/failover for liblustre clients.
4220 Details : liblustre clients were unaware of failover configurations until
4223 Severity : enhancement
4225 Description: user file locks should fail when not mounting with flock option
4226 Details : Set up an error-returning stub in ll_file_operations.lock field
4227 to prevent incorrect behaviour when client is mounted without
4228 flock option. Also, set up properly f_op->flock field for
4232 Frequency : always on ia64
4234 Description: "lfs df" loops on printing out MDS statfs information
4235 Details : The obd_ioctl_data was not initialized and in some systems
4236 this caused a failure during the ioctl that did not return
4237 an error. Initialize the struct and return an error on failure.
4240 Frequency : SLES 9 only
4242 Description: Error of copying files with lustre special EAs as root
4243 Details : Client side always return success for setxattr call for lustre
4244 special xattr (currently only "trusted.lov").
4247 Frequency : rarely on clusters with both ia64+i386 clients
4249 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4250 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4251 clients are doing concurrent writes to the same file it is
4252 possible that the ia64 clients may LASSERT because the OST
4253 extent locks are not PAGE_SIZE aligned. Ensure that grown
4254 locks are always aligned on the request boundary.
4257 Frequency : specific use, occasional
4259 Description: Overwriting in use executable truncates on-disk binary image
4260 Details : If one node attempts to overwrite an executable in use by
4261 another node, we now correctly return ETXTBSY instead of
4262 truncating the file.
4264 Severity : enhancement
4266 Description: Async OSC create to avoid the blocking unnecessarily.
4267 Details : If an OST has no remaining object, system will block on the creating
4268 when need to create a new object on this OST. Now, ways use
4269 pre-created objects when available, instead of blocking on an
4270 empty osc while others are not empty. If we must block, we block
4271 for the shortest possible period of time.
4276 Description: chmod on Lustre root is propagated to other clients
4277 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4278 invalid by the follow_mount time.
4281 Frequency : liblustre clients only
4283 Description: Race in 'instant cancel' lock handling could lead to such locks
4284 never to be granted in case of SMP MDS
4285 Details : Do not destroy not yet granted but cbpending locks in
4289 Frequency : replay/resend of open
4291 Description: non null lock assetion failure in mds_intent_policy
4292 Details : Trying to replay/resend lockless open requests resulted in
4293 mds_open() returning 0 with no lock. Now it sets a flag if
4294 a lock is going to be returned.
4296 Severity : enhancement
4298 Description: Checksum enhancements
4299 Details : New checksum enhancements allow for resending RPCs that failed
4302 Severity : enhancement
4304 Description: Tunables on number of dirty pages in cacche
4305 Details : Allow to set limit on number of dirty pages cached.
4310 Description: client crash on unmount - lock still has references
4311 Details : In some error handling cases it was possible to leak a lock
4312 reference on a client while accessing a file. This was not
4313 harmful to the client during operation, but would cause the
4314 client to crash when the filesystem is unmounted.
4317 Frequency : specific case, rare
4319 Description: ETXTBSY on mds though file not in use by client
4320 Details : ETXTBSY is no longer incorrectly returned when attempting to
4321 chmod or chown a directory that the user previously tried to
4322 execute or a currently-executing binary.
4325 Frequency : extremely rare except on liblustre-based clients
4327 Description: Lustre space not freed when files are deleted
4328 Details : Clean up open-unlinked files after client eviction. Previously
4329 the unlink was skipped and the files remained as orphans.
4334 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4335 Details : In some cases it was possible to send a blocking callback to
4336 a client doing a glimpse, even though that client didn't get
4337 a lock granted. When the glimpse lock is cancelled on the OST
4338 the freed lock is left on the waiting list and corrupted the list.
4341 Frequency : all core dumps
4343 Description: Broke core dumps to lustre
4344 Details : Negative dentry may be unhashed if parent does not have UPDATE
4345 lock, but some callers, e.g. do_coredump, expect dentry to be
4346 hashed after successful create, hash it in ll_create_it.
4348 ------------------------------------------------------------------------------
4350 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4352 * Support for kernels:
4353 2.6.9-42.0.2.EL (RHEL 4)
4354 2.6.5-7.276 (SLES 9)
4355 2.4.21-40.EL (RHEL 3)
4356 2.6.12.6 vanilla (kernel.org)
4360 Frequency : always on RHEL 3
4362 Description: Number of open files grows over time
4363 Details : The number of open files grows over time, whether or not
4364 Lustre is started. This was due to a filp leak introduced
4365 by one of our kernel patches.
4367 ------------------------------------------------------------------------------
4369 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4371 * Support for kernels:
4372 2.6.9-42.EL (RHEL 4)
4373 2.6.5-7.276 (SLES 9)
4374 2.4.21-40.EL (RHEL 3)
4375 2.6.12.6 vanilla (kernel.org)
4380 Bugzilla : 5719, 9635, 9792, 9684
4381 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4382 Details : If a server is under heavy load and cannot reply to new
4383 connection requests before the client resends the (re)connect,
4384 the connection handling code can behave badly if two service
4385 threads are concurrently handing separate (re)connections from
4386 the same client. Add better locking to the connection handling
4387 code, and ensure that only a single connection will be processed
4388 for a given client UUID, even if the lock is dropped.
4390 Severity : enhancement
4392 Description: add TCP zero-copy support to kernel
4393 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4394 sends if the hardware supports scatter-gather and checksumming.
4395 This allows socklnd to do client-write and server-read more
4396 efficiently and reduce CPU utilization from skbuf copying.
4399 Frequency : only if NFS exporting from client
4401 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4402 Details : If setting attributes on a file created under NFS that had
4403 never been opened it would be possible to oops the client
4404 if the file had no objects.
4407 Frequency : always for liblustre
4409 Description: liblustre client does MDS+OSTs setattr RPC for each write
4410 Details : When doing a write from a liblustre client, the client
4411 incorrectly issued an RPC to the MDS and each OST the file was
4412 striped over in order to update the timestamps. When writing
4413 with small chunks and many clients this could overwhelm the MDS
4414 with RPCs. In all cases it would slow down the write because
4415 these RPCs are unnecessary.
4417 Severity : enhancement
4419 Description: allow number of MDS service threads to be changed at module load
4420 Details : It is now possible to change the number of MDS service threads
4421 running. Adding "options mds mds_num_threads={N}" to the MDS's
4422 /etc/modprobe.conf will set the number of threads for the next
4423 time Lustre is restarted (assuming the "mds" module is also
4424 reloaded at that time). The default number of threads will
4425 stay the same, 32 for most systems.
4430 Description: OST crash if filesystem is unformatted or corrupt
4431 Details : If an OST is started on a device that has never been formatted
4432 or if the filesystem is corrupt and cannot even mount then the
4433 error handling cleanup routines would dereference a NULL pointer.
4438 Description: NULL pointer deref in llap_from_page.
4439 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4440 incorrect) mapping if the page was truncated/reclaimed while it was
4441 searched for. Check for this condition and skip such pages when
4442 doing readahead. Introduce extra check to llap_from_page() to
4443 verify page->mapping->host is non-NULL (so page is not anonymous).
4446 Frequency : Sometimes when using sys_sendfile
4448 Description: "page not covered by a lock" warnings from ll_readpage
4449 Details : sendfile called ll_readpage without right page locks present.
4450 Now we introduced ll_file_sendfile that does necessary locking
4451 around call to generic_file_sendfile() much like we do in
4455 Frequency : with certain MDS communication failures at client mount time
4457 Description: NULL pointer deref after failed client mount
4458 Details : a client connection request may delayed by the network layer
4459 and not be sent until after the PTLRPC layer has timed out the
4460 request. If the client fails the mount immediately it will try
4461 to clean up before the network times out the request. Add a
4462 reference from the request import to the obd device and delay
4463 the cleanup until the network drops the request.
4466 Frequency : occasionally during client (re)connect
4468 Description: assertion failure during client (re)connect
4469 Details : processing a client connection request may be delayed by the
4470 client or server longer than the client connect timeout. This
4471 causes the client to resend the connection request. If the
4472 original connection request is replied in this interval, the
4473 client may trip an assertion failure in ptlrpc_connect_interpret()
4474 which thought it would be the only running connect process.
4477 Frequency : only with obd_echo servers and clients that are rebooted
4479 Description: kernel BUG accessing uninitialized data structure
4480 Details : When running an obd_echo server it did not start the ping_evictor
4481 thread, and when a client was evicted an uninitialized data
4482 structure was accessed. Start the ping_evictor in the RPC
4483 service startup instead of the OBD startup.
4485 Severity : enhancement
4486 Bugzilla : 10193 (patchless)
4487 Description: Remove dependency on various unexported kernel interfaces.
4488 Details : No longer need reparent_to_init, exit_mm, exit_files,
4489 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4492 Frequency : rare (only users of deprecated and unsupported LDAP config)
4494 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4495 Details : LDAP apparently contains 'lustreName' attributes instead of
4496 'name'. A simple remapping of the name is sufficient.
4499 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4501 Description: waiting_locks_callback trips kernel BUG if client is evicted
4502 Details : Running with the dump_on_timeout debug flag turned on makes
4503 it possible that the waiting_locks_callback() can try to dump
4504 the Lustre kernel debug logs from an interrupt handler. Defer
4505 this log dumping to the expired_lock_main() thread.
4507 Severity : enhancement
4509 Description: Support NFS exporting on 2.6 kernels.
4510 Details : Implement non-rawops metadata methods for NFS server to use without
4511 changing NFS server code.
4514 Frequency : very rare (synthetic metadata workload only)
4516 Description: two racing renames might cause an MDS thread to deadlock
4517 Details : Running the "racer" program may cause one MDS thread to rename
4518 a file from being the source of a rename to being the target of
4519 a rename at exactly the same time that another thread is doing
4520 so, and the second thread has already enqueued these locks after
4521 doing a lookup of the target and is trying to relock them in
4522 order. Ensure that we don't try to re-lock the same resource.
4525 Frequency : only very large systems with liblustre clients
4527 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4528 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4529 all OSTs in parallel. This allows the eviction of stale liblustre
4530 clients to proceed much faster than if they were done in series,
4531 and also offers similar improvements for other set_info RPCs.
4536 Description: excessive CPU usage during initial read phase on client
4537 Details : During the initial read phase on a client, it would agressively
4538 retry readahead on the file, consuming too much CPU and impacting
4539 performance (since 1.4.5.8). Improve the readahead algorithm
4540 to avoid this, and also improve some other common cases (read
4541 of small files in particular, where "small" is files smaller than
4542 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4547 Description: MDS crash when receiving packet with unknown intent.
4548 Details : Do not LBUG in unknown intent case, just return -EFAULT
4550 Severity : enhancement
4551 Bugzilla : 9293, 9385
4552 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4553 Details : Do not serialize getattr (non-intent version) and statfs.
4556 Frequency : occasional, when OST network is overloaded/intermittent
4558 Description: client evicted by OST after bulk IO timeout
4559 Details : If a client sends a bulk IO request (read or write) the OST
4560 may evict the client if it is unresposive to its data GET/PUT
4561 request. This is incorrect if the network is overloaded (takes
4562 too long to transfer the RPC data) or dropped the OST GET/PUT
4563 request. There is no need to evict the client at all, since
4564 the pinger and/or lock callbacks will handle this, and the
4565 client can restart the bulk request.
4568 Frequency : Always when mmapping file with no objects
4570 Description: client crashes when mmapping file with no objects
4571 Details : Check that we actually have objects in a file before doing any
4572 operations on objects in ll_vm_open, ll_vm_close and
4578 Description: Request leak when working with deleted CWD
4579 Details : Introduce advanced request refcount tracking for requests
4580 referenced from lustre intent.
4582 Severity : Enhancement
4584 Description: Cache open file handles on client.
4585 Details : MDS now will return special lock along with openhandle, if
4586 requested and client is allowed to hold openhandle, even if unused,
4587 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4588 closing files for every read/write openration.
4590 Severity : Enhancement
4592 Description: Cache open negative dentries on client when possible.
4593 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4594 negative dentries on lock revocation.
4599 Description: Remounting a client read-only wasn't possible with a zconf mount
4600 Details : It wasn't possible to remount a client read-only with llmount.
4602 Severity : enhancement
4603 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4604 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4605 MPICH is included to add Lustre-specific ADIO interfaces.
4606 This is based closely on the UFS ADIO layer and only differs
4607 in file creation, in order to allow the OST striping to be set.
4608 This is user-contributed code and not supported by CFS.
4613 Description: extended inode attributes (immutable, append-only) work improperly
4614 when 2.4 and 2.6 kernels are used on client/server or vice versa
4615 Details : Introduce kernel-independent values for these flags.
4617 Severity : enhancement
4620 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4621 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4622 granularity. Now, Fractional MB granularity is be supported,
4623 this is very useful for low memory system.
4625 Severity : enhancement
4627 Description: Getattr by fid
4628 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4629 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4630 have locks and avoids getattr rpc requests for it on every stat.
4633 Frequency : Always, for filesystems larger than 2TB
4635 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4636 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4637 filesystems larger than 2TB to be mounted with mballoc enabled.
4640 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4642 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4643 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4644 filesystems larger than 2TB is mounted then the kernel will
4645 silently corrupt the start of the filesystem. CONFIG_LBD is
4646 enabled for all CFS-supported kernels, but the possibility of
4647 this happening with a modified kernel config exists.
4649 Severity : enhancement
4651 Description: add client O_DIRECT support for 2.6 kernels
4652 Details : It is now possible to do O_DIRECT reads and writes to files
4653 in the Lustre client mountpoint on 2.6 kernel clients.
4655 Severity : enhancement
4657 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4658 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4659 parallel, not waiting for response from every OST before sending
4660 a rpc to the next OST.
4665 Description: setattr vs write race when updating file timestamps
4666 Details : Client processes that update a file timestamp into the past
4667 right after writing to the file (e.g. tar) it is possible that
4668 the updated file modification time can be reset to the current
4669 time due to a race between processing the setattr and write RPC.
4671 Severity : enhancement
4673 Description: Bring 'lfs find' closer in line with regular Linux find.
4674 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4675 -print0 options and obtains all the needed info through the lustre
4678 Severity : enhancement
4680 Description: support up to 1024 configured devices on one node
4681 Details : change obd_dev array from statically allocated to dynamically
4682 allocated structs as they are first used to reduce memory usage
4687 Description: Flush dirty partially truncated pages during truncate
4688 Details : Immediatelly flush partially truncated pages in filter_setattr,
4689 this way we completely avoid having any pages in page cache on OST
4690 and can retire ugly workarounds during writes to flush such pages.
4695 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4696 Details : protect inode from truncation within vfs_unlink() context
4697 just take a reference before calling vfs_unlink() and release it
4698 when parent's i_sem is free.
4703 Description: last_id value checked outside lock on OST caused LASSERT failure
4704 Details : If there were multiple MDS->OST object precreate requests in
4705 flight, it was possible that the OST's last object id was checked
4706 outside a lock and incorrectly tripped an assertion. Move checks
4707 inside locks, and discard old precreate requests.
4710 Frequency : always, if extents are used on OSTs
4712 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4713 Details : OSTs using the extents format would not zero the high 16 bits of
4714 the index physical block number. This is not a problem for any
4715 OST filesystems smaller than 16TB, and no kernels support ext3
4716 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4717 new/modified files) and can be fixed for existing filesystems
4718 with e2fsprogs-1.39-cfs1.
4723 Description: import connection selection may be incorrect if timer wraps
4724 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4725 import connections to be ignored if the 32-bit jiffies counter
4726 wraps. Use a 64-bit jiffies counter.
4729 Frequency : very large clusters immediately after boot
4731 Description: LNET request buffers exhausted under heavy short-term load
4732 Details : If a large number of client requests are generated on a service
4733 that has previously never seen so many requests it is possible
4734 that the request buffer growth cannot keep up with the spike in
4735 demand. Instead of dropping incoming requests, they are held in
4736 the LND until the RPC service can accept more requests.
4739 Frequency : Sometimes during replay
4741 Description: Assertion failure in ll_local_open after replay.
4742 Details : If replay happened on an open request reply before we were able
4743 to set replay handler, reply will become not swabbed tripping the
4744 assertion in ll_local_open. Now we set the handler right after
4745 recognising of open request
4748 Frequency : very rare
4750 Description: kernel reports "badness in vsnprintf"
4751 Details : Reading from the "recovery_status" /proc file in small chunks
4752 may cause a negative length in lprocfs_obd_rd_recovery_status()
4753 call to vsnprintf() (which is otherwise harmless). Exit early
4754 if there is no more space in the output buffer.
4756 Severity : enhancement
4758 Description: clear OBD RPC statistics by writing to them
4759 Details : It is now possible to clear the OBD RPC statistics by writing
4760 to the "stats" file.
4765 Description: Client mtime is not the same on different clients after utimes
4766 Details : In some cases, the client was using the utimes() syscall on
4767 a file cached on another node. The clients now validate the
4768 ctime from the MDS + OSTs to determine which one is right.
4773 Description: Inability to activate failout mode
4774 Details : lconf script incorrectly assumed that in pythong string's numeric
4775 value is used in comparisons.
4778 Frequency : always with multiple stripes per file
4780 Description: Inefficient object allocation for mutli-stripe files
4781 Details : When selecting which OSTs to stripe files over, for files with
4782 a stripe count that divides evenly into the number of OSTs,
4783 the MDS is always picking the same starting OST for each file.
4784 Return the OST selection heuristic to the original design.
4789 Description: mount failures may take full timeout to return an error
4790 Details : Under some heavy load conditions it is possible that a
4791 failed mount can wait for the full obd_timeout interval,
4792 possibly several minutes, before reporting an error.
4793 Instead return an error as soon as the status is known.
4795 Frequency : quota enabled and large files being deleted
4797 Description: releasing more than 4GB of quota at once hangs OST
4798 Details : If a user deletes more than 4GB of files on a single OST it
4799 will cause the OST to spin in an infinite loop. Release
4800 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4805 Description: statfs data retrieved from /proc may be stale or zero
4806 Details : When reading per-device statfs data from /proc, in the
4807 {kbytes,files}_{total,free,avail} files, it may appear
4808 as zero or be out of date.
4811 Frequency : systems with MD RAID1 external journal devices
4813 Description: lconf's call to blkid is confused by RAID1 journal devices
4814 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4815 of returning all block devices that match the journal UUID.
4818 Frequency : always, for aggregate stripe size over 4GB
4820 Description: assertion fails when trying to use 4GB stripe size
4821 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4822 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4825 Frequency : always on ppc64
4827 Description: the first write on an ext3 filesystem with mballoc got stuck
4828 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4829 perform endianness conversion.
4831 ------------------------------------------------------------------------------
4833 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4835 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4836 INTEROPERATE with older versions automatically. Please read the
4837 user documentation before upgrading any part of a live system.
4838 * WARNING: Lustre networking configuration changes are required with
4839 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4842 * Support for kernels:
4843 2.6.9-22.0.2.EL (RHEL 4)
4844 2.6.5-7.244 (SLES 9)
4845 2.6.12.6 vanilla (kernel.org)
4848 Severity : enhancement
4849 Bugzilla : 7981/8208
4850 Description: Introduced Lustre Networking (LNET)
4851 Details : LNET is new networking infrastructure for Lustre, it includes
4852 a reorganized network configuration mode (see the user
4853 documentation for full details) as well as support for routing
4854 between different network fabrics. Lustre Networking Devices
4855 (LNDs) for the supported network fabrics have also been
4856 created for this new infrastructure.
4858 Severity : enhancement
4859 Description: Introduced Access control lists
4860 Details : clients can set ACLs on files and directories in order to have
4861 more fine-grained permissions than the standard Unix UGO+RWX.
4862 The MDS must be started with the "-o acl" mount option.
4864 Severity : enhancement
4865 Description: Introduced filesystem quotas
4866 Details : Administrators may now establish per-user quotas on the
4869 Severity : enhancement
4871 Description: Configuration change for the XT3
4872 The PTLLND is now used to run Lustre over Portals on the XT3
4873 The configure option(s) --with-cray-portals are no longer used.
4874 Rather --with-portals=<path-to-portals-includes> is used to
4875 enable building on the XT3. In addition to enable XT3 specific
4876 features the option --enable-cray-xt3 must be used.
4881 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4882 Details : A race between allocating a new llcd and re-getting the llcd_lock
4883 allowed another thread to grab newly-allocated llcd.
4885 Severity : enhancement
4887 Description: 2.6 OST async journal commit and locking fix to improve performance
4888 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4889 2.6 kernels are now async as they already were in 2.4 kernels so
4890 that they can commit concurrently with the network bulk transfer.
4891 For block-allocated files the filter allocation semaphore is held
4892 to avoid filesystem fragmentation during allocation. BKL lock
4893 removed for 2.6 xattr operations where it is no longer needed.
4898 Description: lconf incorrectly determined whether two IP networks could talk
4899 Details : In some more complicated routing and multiple-network
4900 configurations, lconf will avoid trying to make a network
4901 connection to a disjoint part of the IP space. It was doing the
4902 math incorrectly for one set of cases.
4907 Description: Fix for potential infinite loop processing records in an llog.
4908 Details : If an llog record is corrupted/zeroed, it is possible to loop
4909 forever in llog_process(). Validate the llog record length
4910 and skip the remainder of the block on error.
4913 Frequency : occasional (liblustre only)
4915 Description: liblustre could not open files whose last component is a symlink
4916 Details : sysio_path_walk() would incorrectly pass the open intent to
4917 intermediate path components.
4920 Frequency : rare (liblustre only with non-standard tuning)
4921 Bugzilla : 7201 (7350)
4922 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4923 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4924 after it has been released, only do this for referenced locks
4926 Severity : enhancement
4928 Description: specifying an (invalid) directory default stripe_size of -1
4929 would reset the directory default striping
4930 Details : stripe_size -1 was used internally to signal directory stripe
4931 removal, now use "all default" to signal dir stripe removal
4932 as a directory striping of "all default" is not useful
4935 Frequency : common for large clusters running liblustre clients
4937 Description: doing an ls when liblustre clients are running is slow
4938 Details : sending a glimpse AST to a liblustre client waits for every AST
4939 to time out, as liblustre clients will not respond. Since they
4940 cannot cache data we refresh the OST lock LVB from disk instead.
4942 Severity : enhancement
4944 Description: doing an ls at the same time as file IO can be slow
4945 Details : enqueue and other "small" requests can be blocked behind many
4946 large IO requests. Create a new OST IO portal for non-IO
4947 requests so they can be processed faster.
4950 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4952 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4953 created file with mode 000
4954 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4955 allow an ftruncate() as a non-root user to a file with mode 000.
4956 Lustre now respects this flag to disable mode checks when
4957 truncating a file owned by the user
4960 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4963 Description: Revoking locks from clients that went dead or catatonic might take
4965 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4966 cancellation of such locks instant on servers without waiting for
4967 any reply from clients. Clients drops these locks when cancel
4968 notification from server is received without replying.
4971 Frequency : liblustre-only, when liblustre client dies or becomes busy
4973 Description: Doing ls on Linux clients can take a long time with active
4975 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4976 granting such locks to it in the first place if possible. Locks
4977 are taken by proxy on the OST during the read or write and
4978 dropped immediately afterward. Add connect flags handling, do
4979 not grant locks to liblustre clients for glimpse ASTs.
4981 Severity : enhancement
4983 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4984 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4985 sequential reads and behaves badly with applications doing random
4986 reads. Improve it by reading ahead at least read region, and
4987 avoiding excessive large RPC for small reads.
4989 Severity : enhancement
4991 Description: Creating more than 1000 files for a single job may cause a load
4992 imbalance on the OSTs if there are also a large number of OSTs.
4993 Details : qos_prep_create() uses an OST index reseed value that is an
4994 even multiple of the number of available OSTs so that if the
4995 reseed happens in the middle of the object allocation it will
4996 still utilize the OSTs as uniformly as possible.
5001 Description: OST or MDS may oops in ping_evictor_main()
5002 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5003 but doesn't restart at beginning of obd_exports_timed list
5006 Severity : enhancement
5008 Description: improve by-nid export eviction on the MDS and OST
5009 Details : allow multiple exports with the same NID to be evicted at one
5010 time without re-searching the exports list.
5013 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5015 Description: MDS may oops in groups_free()
5016 Details : in rare race conditions a newly allocated group_info struct is
5017 freed again, and this can be NULL. The 2.4 compatibility code
5018 for groups_free() checked for a NULL pointer, but 2.6 did not.
5021 Frequency : common for liblustre clients doing little filesystem IO
5022 Bugzilla : 9352, 7313
5023 Description: server may evict liblustre clients accessing contended locks
5024 Details : if a client is granted a lock or receives a completion AST
5025 with a blocking AST already set it would not reply to the AST
5026 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5029 Frequency : lfs setstripe, only systems with more than 160 OSTs
5031 Description: unable to set striping with a starting offset beyond OST 160
5032 Details : llapi_create_file() incorrectly limited the starting stripe
5033 index to the maximum single-file stripe count.
5036 Frequency : LDAP users only
5038 Description: lconf did not handle in-kernel recovery with LDAP properly
5039 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5041 Severity : enhancement
5043 Description: bind OST threads to NUMA nodes to improve performance
5044 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5045 node and do their allocations there to localize memory access
5047 Severity : enhancement
5049 Description: llmount can determine client NID directly from Myrinet (GM)
5050 Details : the client NID code from gmnalnid was moved directly into
5051 llmount, removing the need to use this or specifying the
5052 client NID explicitly when mounting GM clients with zeroconf
5055 Frequency : if client is started with down MDS
5057 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5058 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5059 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5060 ptlrpc_import_delayed_req() because we didn't check if the
5061 request was interrupted, and we also didn't break out of the
5062 event loop if there was no timeout
5067 Description: data loss during non-page-aligned writes to a single file from
5068 both multiple nodes and multiple threads on one node at same time
5069 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5070 inconsistency led to false short-reads, that were cached and later
5071 used by ->prepare_write() to fill in partially written page,
5072 leading to data loss.
5075 Frequency : always, if lconf --abort_recovery used
5077 Description: lconf --abort_recovery fails with 'Operation not supported'
5078 Details : lconf was attempting to abort recovery on the MDT device and not
5081 Severity : enhancement
5083 Description: remove cleanup logs
5084 Details : replace lconf-generated cleanup logs with lustre internal
5085 cleanup routines. Eliminates the need for client-cleanup and
5088 Severity : enhancement
5090 Description: add support for EAs (user and system) on lustre filesystems
5091 Details : it is now possible to store extended attributes in the Lustre
5092 client filesystem, and with the user_xattr mount option it
5093 is possible to allow users to store EAs on their files also
5095 Severity : enhancement
5097 Description: Add possibility (config option) to show minimal available OST free
5099 Details : When compiled with --enable-mindf configure option, statfs(2)
5100 (and so, df) will return least minimal free space available from
5101 all OSTs as amount of free space on FS, instead of summary of
5102 free spaces of all OSTs.
5104 Severity : enhancement
5106 Description: do not expand extent locks acquired on OST-side
5107 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5108 by server: they are not cached anyway.
5111 Frequency : when mmap is used/binaries executed from Lustre
5113 Description: Unmmap pages before throwing them away from read cache.
5114 Details : llap_shrink cache now attempts to unmap pages before discarding
5115 them (if unmapping failed - do not discard). SLES9 kernel has
5116 extra checks that trigger if this unmapping is not done first.
5121 Description: lconf didn't resolve symlinks before checking to see whether a
5122 given mountpoint was already in use
5125 Frequency : when migrating failover services
5126 Bugzilla : 6395, 9514
5127 Description: When migrating a subset of services from a node (e.g. failback
5128 from a failover service node) the remaining services would
5129 time out and evict clients.
5130 Details : lconf --force (implied by --failover) sets the global obd_timeout
5131 to 5 seconds in order to quickly disconnect, but this caused
5132 other RPCs to time out too quickly. Do not change the global
5133 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5135 Severity : enhancement
5136 Frequency : if MDS is started with down OST
5137 Bugzilla : 9439,5706
5138 Description: Allow startup/shutdown of an MDS without depending on the
5139 availability of the OSTs.
5140 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5141 Add appropriate locking and lov-osc refcounts for safe
5142 cleaning. Add osc abort_inflight calls in case the
5143 synchronize never started.
5146 Frequency : occasional (Cray XT3 only)
5148 Description: root not authorized to access files in CRAY_PORTALS environment
5149 Details : The client process capabilities were not honoured on the MDS in
5150 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5151 been accessed by an authorized user then root was able to access
5152 the file on the local client also. The root user capabilities
5153 are now allowed on the MDS, as this environment has secure UID.
5156 Frequency : occasional
5158 Description: ldiskfs "too long searching" message happens too often
5159 Details : A debugging message (otherwise harmless) prints too often on
5160 the OST console. This has been reduced to only happen when
5161 there are fragmentation problems on the filesystem.
5166 Description: Division by zero in statfs when all OSCs are inactive
5167 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5168 lov_statfs divides by value returned by lov_get_stripecnt().
5172 Bugzilla : 9489, 3273
5173 Description: First write from each client to each OST was only 4kB in size,
5174 to initialize client writeback cache, which caused sub-optimal
5175 RPCs and poor layout on disk for the first written file.
5176 Details : Clients now request an initial cache grant at (re)connect time
5177 and so that they can start streaming writes to the cache right
5178 away and always do full-sized RPCs if there is enough data.
5179 If the OST is rebooted the client also re-establishes its grant
5180 so that client cached writes will be honoured under the grant.
5185 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5186 Details : Now I/O RPCs go to different portal number and (presumably) fast
5187 lock requests (and glimses) and other RPCs get their own service
5188 threads pool that should be able to service those RPCs
5191 Severity : enhancement
5193 Description: Ability to exchange lustre version between client and servers and
5194 issue warnings at client side if client is too old. Also for
5195 liblustre clients there is ability to refuse connection of too old
5197 Details : New 'version' field is added to connect data structure that is
5198 filled with version info. That info is later checked by server and
5202 Frequency : rare, liblustre only.
5203 Bugzilla : 9296, 9581
5204 Description: Two simultaneous writes from liblustre at offset within same page
5205 might proceed at the same time overwriting eachother with stale
5207 Details : I/O lock withing llu_file_prwv was released too early, before data
5208 actually was hitting the wire. Extended lock-holding time until
5209 server acknowledges receiving data.
5212 Frequency : extremely rare. Never observed in practice.
5214 Description: avoid generating lustre_handle cookie of 0.
5215 Details : class_handle_hash() generates handle cookies by incrementing
5216 global counter, and can hit 0 occasionaly (this is unlikely, but
5217 not impossible, because initial value of cookie counter is
5218 selected randonly). Value of 0 is used as a sentinel meaning
5219 "unassigned handle" --- avoid it. Also coalesce two critical
5220 sections in this function into one.
5222 Severity : enhancement
5224 Description: allow liblustre clients to delegate truncate locking to OST
5225 Details : To avoid overhead of locking, liblustre client instructs OST to
5226 take extent lock in ost_punch() on client's behalf. New connection
5227 flag is added to handle backward compatibility.
5229 Severity : enhancement
5230 Bugzilla : 4928, 7341, 9758
5231 Description: allow number of OST service threads to be specified
5232 Details : a module parameter allows the number of OST service threads
5233 to be specified via "options ost ost_num_threads={N}" in the
5234 OSS's /etc/modules.conf or /etc/modprobe.conf.
5238 Bugzilla : 6146, 9635, 9895
5239 Description: servers crash with bad pointer in target_handle_connect()
5240 Details : In rare cases when a client is reconnecting it was possible that
5241 the connection request was the last reference for that export.
5242 We would temporarily drop the export reference and get a new
5243 one, but this may have been the last reference and the export
5244 was just destroyed. Get new reference before dropping old one.
5246 Severity : enhancement
5247 Frequency : if client is started with failover MDS
5249 Description: Allow multiple MDS hostnames in the mount command
5250 Details : Try to read the configuration from all specified MDS
5251 hostnames during a client mount in case the "primary"
5254 Severity : enhancement
5256 Description: Stop sending data to evicted clients as soon as possible.
5257 Details : Check if the client we are about to send or are sending data to
5258 was evicted already. (Check is done every second of waiting,
5259 for which l_wait_event interface was extended to allow checking
5260 of exit condition at specified intervals).
5263 Frequency : rare, normally only when NFS exporting is done from client
5265 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5267 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5270 Frequency : always, if srand() is called before liblustre initialization
5272 Description: Liblustre uses system PRNG disturbing its usage by user application
5273 Details : Introduce internal to lustre fast and high-quality PRNG for
5274 lustre usage and make liblustre and some other places in generic
5275 lustre code to use it.
5277 Severity : enhancement
5278 Bugzilla : 9477, 9557, 9870
5279 Description: Verify that the MDS configuration logs are updated when xml is
5280 Details : Check if the .xml configuration logs are newer than the config
5281 logs stored on the MDS and report an error if this is the case.
5282 Request --write-conf, or allow starting with --old_conf.
5284 Severity : enhancement
5286 Description: Handle symlinks in the path when checking if Lustre is mounted.
5287 Details : Resolve intermediate symlinks when checking if a client has
5288 mounted a filesystem to avoid duplicate client mounts.
5293 Description: lconf can hit an error exception but still return success.
5294 Details : The lconf command catches the Command error exception at the top
5295 level script context and will exit with the associated exit
5296 status, but doesn't ensure that this exit status is non-zero.
5301 Description: failure of ptlrpc thread startup can cause oops
5302 Details : Starting a ptlrpc service thread can fail if there are a large
5303 number of threads or the server memory is very fragmented.
5304 Handle this without oopsing.
5307 Frequency : always, only if liblustre and non-default acceptor port was used
5309 Description: liblustre cannot connect to servers with non-default acceptor port
5310 Details : tcpnal_set_default_params() was not called and was therefore
5311 ignoring the environment varaible TCPNAL_PORT, as well as other
5312 TCPNAL_ environment variables
5317 Description: two objects could be created on the same OST for a single file
5318 Details : If an OST is down, in some cases it was possible to create two
5319 objects on a single OST for a single file. No problems other
5320 than potential performance impact and spurious error messages.
5324 Bugzilla : 5681, 9562
5325 Description: Client may oops in ll_unhash_aliases
5326 Details : Client dcache may become inconsistent in race condition.
5327 In some cases "getcwd" can fail if the current directory is
5333 Description: Inode refcounting problems in NFS export code
5334 Details : link_raw functions used to call d_instantiate without obtaining
5335 extra inode reference first.
5339 Bugzilla : 9942, 9903
5340 Description: Referencing freed requests leading to crash, memleaks with NFS.
5341 Details : We used to require that call to ll_revalidate_it was always
5342 followed by ll_lookup_it. Also with revalidate_special() it is
5343 possible to call ll_revalidate_it() twice for the same dentry
5344 even if first occurence returned success. This fix changes semantic
5345 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5346 reference on a request referred from the intent.
5347 ll_intent_release() then releases such a request.
5350 Frequency : rare, normally benchmark loads only
5352 Description: unlinked inodes were kept in memory on the client
5353 Details : If a client is repeatedly creating and unlinking files it
5354 can accumulate a lot of stale inodes in the inode slab cache.
5355 If there is no other client load running this can cause the
5356 client node to run out of memory. Instead flush old inodes
5357 from client cache that have the same inode number as a new inode.
5360 Frequency : SLES9 2.6.5 kernel and long filenames only
5361 Bugzilla : 9969, 10379
5362 Description: utime reports stale NFS file handle
5363 Details : SLES9 uses out-of-dentry names in some cases, which confused
5364 the lustre dentry revalidation. Change it to always use the
5368 Frequency : rare, unless heavy write-truncate concurrency is continuous
5369 Bugzilla : 4180, 6984, 7171, 9963, 9331
5370 Description: OST becomes very slow and/or deadlocked during object unlink
5371 Details : filter_destroy() was holding onto the parent directory lock
5372 while truncating+unlinking objects. For very large objects this
5373 may block other threads for a long time and slow overall OST
5374 responsiveness. It may also be possible to get a lock ordering
5375 deadlock in this case, or run out of journal credits because of
5376 the combined truncate+unlink. Solution is to do object truncate
5377 first in one transaction without parent lock, and then do the
5378 final unlink in a new transaction with the parent lock. This
5379 reduces the lock hold time dramatically.
5382 Frequency : rare, 2.4 kernels only
5384 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5385 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5386 before unmounting their filesystem in order to avoid deadlock.
5387 The kernel_locked() function in 2.4 kernels only checks whether
5388 the kernel lock is held, not whether it is this process that is
5389 holding it as 2.6 kernels do.
5394 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5395 Details : The client ptlrpc code may be trying to reconnect to a down
5396 server before a previous connection attempt has timed out.
5397 Increase the reconnect interval to be longer than the connection
5398 timeout interval to avoid sending duplicate connections to
5402 Frequency : echo_client brw_test command
5404 Description: fix echo_client to work with OST preallocated code
5405 Details : OST preallocation code (5137) didn't take echo_client IO path
5406 into account: echo_client calls filter methods outside of any
5407 OST thread and, hence, there is no per-thread preallocated
5408 pages and buffers to use. Solution: hijack pga pages for IO. As
5409 a byproduct, this avoids unnecessary data copying.
5413 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5414 Description: Client can oops in mdc_commit_close() after open replay
5415 Details : It was possible for the MDS to return an open request with no
5416 transaction number in mds_finish_transno() if the client was
5417 evicted, but without actually returning an error. Clients
5418 would later try to replay that open and may trip an assertion
5419 Simplify the client close codepath, and always return an error
5420 from the MDS in case the open is not successful.
5423 Frequency : rare, 2.6 OSTs only
5425 Description: OST may deadlock under high load on fragmented files
5426 Details : If there was a heavy load and highly-fragmented OST filesystems
5427 it was possible to have all the OST threads deadlock waiting on
5428 allocation of biovecs, because the biovecs were not released
5429 until the entire RPC IO was completed. Instead, release biovecs
5430 as soon as they are complete to ensure forward IO progress.
5432 Severity : enhancement
5434 Description: Support for specifying external journal device at mount
5435 Details : If an OST or MDS device is formatted with an external journal
5436 device, this device major/minor is stored in the ext3 superblock
5437 and may not be valid for failover. Allow detecting and
5438 specifying the external journal at mount time.
5443 Description: Mounting an MDS with pending unlinked files may cause oops
5444 Details : target_finish_recovery() calls mds_postrecov() which returned
5445 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5446 considers this an error and immediately begins cleaning up the
5447 lov, just after starting the mds_lov process
5449 Severity : enhancement
5451 Description: Implement 'lfs df' to report actual free space on per-OST basis
5452 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5453 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5454 usage of objects; '-h' to report in human readable format.
5456 ------------------------------------------------------------------------------
5458 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5465 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5466 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5467 at startup during orphan unlinking. Instead, skip buddy update
5468 before setup, as it will be regenerated after recovery is complete.
5473 Description: If an OST is inactive, its locks might reference stale inodes.
5474 Details : lov_change_cbdata() must iterate over all namespaces, even if
5475 they are inactive to clear inode references from the lock.
5477 Severity : enhancement
5478 Frequency : occasional, if non-standard max_dirty_mb used
5480 Description: Client will block write RPCs if not enough grant
5481 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5482 then the client will block writes while waiting for another RPC
5483 to complete instead of consuming its dirty limit. With change
5484 we get improved performance when max_dirty_mb is small.
5486 Severity : enhancement
5487 Bugzilla : 3389, 6253
5488 Description: Add support for supplementary groups on the MDS.
5489 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5490 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5491 lookups for user supplementary groups into a cache.
5495 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5496 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5497 for liblustre O_EXCL is enforced by MDS.
5501 Description: readdir never returns NULL in liblustre.
5502 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5503 dentry in struct dirent.
5507 Frequency : liblustre only, depends on application IO pattern
5508 Description: liblustre clients evicted if not contacting servers
5509 Details : Don't put liblustre clients into the ping_evictor list, so
5510 they will not be evicted by the pinger ever.
5512 Severity : enhancement
5514 Description: Add ability to evict clients by NID from MDS.
5515 Details : By echoing "nid:$NID" string into
5516 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5517 $NID would be instantly evicted from this MDS and from all active
5518 OSTs connected to it.
5522 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5523 Details : lookup_it_finish() used to query file size from OSTs that was not
5528 Description: service threads change working directory to that of init
5529 Details : Starting lustre service threads may pin the working directory
5530 of the parent thread, making that filesystem busy. Threads
5531 now change to the working directory of init to avoid this.
5535 Frequency : during shutdown only
5536 Description: shutdown with a failed MDS or OST can cause unmount to hang
5537 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5542 Frequency : 2.6 only
5543 Description: chmod/chown may include an extra supplementary group
5544 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5545 supplementary group and if none is specified this is used.
5548 Bugzilla : 5479 (6816)
5550 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5551 Details : If lookup is in progress on a file that is unlinked we might try
5552 to revalidate the inode and fail in revalidate after lookup is
5553 complete and ll_file_open() enqueues the open again but
5554 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5557 Frequency : always, if lconf --abort_recovery used
5559 Description: lconf --abort_recovery fails with 'Operation not supported'
5560 Details : lconf was attempting to abort recovery on the MDT device and not
5563 ------------------------------------------------------------------------------
5565 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5570 Frequency : rare (only unsupported configurations with a node running as an
5572 Bugzilla : 6514, 5137
5573 Description: Mounting a Lustre file system on a node running as an OST could
5575 Details : OSTs now preallocates memory needed to write out data at
5576 startup, instead of when needed, to avoid having to
5577 allocate memory in possibly low memory situations.
5578 Specifically, if the file system is mounted on on OST,
5579 memory pressure could force it to try to write out data,
5580 which it needed to allocate memory to do. Due to the low
5581 memory, it would be unable to do so and the node would
5582 become unresponsive.
5584 Severity : enhancement
5586 Description: Addition of lconf --service command line option
5587 Details : lconf now accepts a '--service <arg>' option, which is
5588 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5590 Severity : enhancement
5592 Description: Failover mode is now the default for OSTs.
5593 Details : By default, OSTs will now run in failover mode. To return to
5594 the old behaviour, add '--failout' to the lmc line for OSTs.
5596 Severity : enhancement
5598 Description: Health checks are now provided for MDS and OSTs
5599 Details : Additional detailed health check information on MSD and OSTs
5600 is now provided through the procfs health_check value.
5603 Frequency : occasional, depends on IO load
5605 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5606 after numerous create/delete cycles
5607 Details : The ext3 inode allocation policy would not allocate new inodes
5608 very well on the OSTs because there are no new directories
5609 being created. Instead we look for groups with free space if
5610 the parent directories are nearly full.
5615 Description: Network or server problems during mount may cause partially
5616 mounted clients instead of returning an error.
5617 Details : The config llog parsing code may overwrite the error return
5618 code during mount error handling, returning success instead
5624 Description: MDS can fail to allocate large reply buffers
5625 Details : After long uptimes the MDS can fail to allocate large reply
5626 buffers (e.g. zconf client mount config records) due to memory
5627 fragmentation or consumption by the buffer cache. Preallocate
5628 some large reply buffers so that these replies can be sent even
5629 under memory pressure.
5633 Frequency : rare (liblustre)
5634 Description: fsx running with liblustre complained that using truncate() to
5635 extend the file doesn't work. This patch corrects that issue.
5636 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5637 ATTR_SIZE and lsm use in llu_setattr_raw.
5641 Frequency : rare, only 2.6 kernels
5642 Description: Unusual file access patterns on the MDS may result in inode
5643 data being lost in very rare circumstances.
5644 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5645 code in the 2.6 kernel caused the inode and/or EA data not to
5646 be read from disk, causing single-file corruption.
5650 Frequency : rare, only 2.6 filesystems using extents
5651 Description: Heavy concurrent write and delete load may cause data corruption.
5652 Details : It was possible under high-load situations to have an extent
5653 metadata block in the block device cache from a just-unlinked
5654 file overwrite a newly-allocated data block. We now unmap any
5655 metadata buffers that alias just-allocated data blocks.
5659 Frequency : filesystems with default stripe_count larger than 77
5660 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5661 Details : lconf specifies an inode size of 4096 bytes when the default
5662 stripe_count is larger than 77. This conflicts with the default
5663 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5664 this case to avoid pinning too much memory for large EAs.
5666 ------------------------------------------------------------------------------
5668 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5673 Frequency : rare (extremely heavy IO load with hundreds of clients)
5675 Description: Client is evicted, gets IO error writing to file
5676 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5677 caused clients to be evicted for AST timeouts. The fixes in
5678 bug 5192 mean we no longer need to have such short AST timeouts
5679 so ldlm_timeout has been increased.
5682 Frequency : occasional during --force or --failover shutdown under load
5683 Bugzilla : 5949, 4834
5684 Description: Server oops/LBUG if stopped with --force or --failover under load
5685 Details : a collection of import/export refcount and cleanup ordering
5686 issues fixed for safer force cleanup
5689 Frequency : only filesystems larger than 120 OSTs
5690 Bugzilla : 5990, 6223
5691 Description: lfs getstripe would oops on a very large filesystem
5692 Details : lov_getconfig used kfree on vmalloc'd memory
5695 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5696 Bugzilla : 6242, 6243
5697 Description: reading from files that had been truncated to a non-zero size
5698 but never opened returned no data
5699 Details : ll_file_read() reads zeros from no-object files to EOF
5704 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5705 Details : There's little harm in aborting MDS/OSS recovery and letting it
5706 try again, so I removed the LASSERT and return an error instead.
5708 Severity : enhancement
5710 Description: New debugging infrastructure for tracking down data corruption
5711 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5712 (b) cover more of the client-side code path, and (c) try to narrow
5713 down where problems occurred
5717 Bugzilla : 3819, 4364, 4397, 6313
5718 Description: Racing close and eviction MDS could cause assertion in mds_close
5719 Details : It was possible to get multiple mfd references during close and
5720 client eviction, leading to one thread referencing a freed mfd.
5722 Severity: : enhancement
5723 Bugzilla : 3262, 6359
5724 Description: Attempts to reconnect to servers are now more aggressive.
5725 Details : This builds on the enhanced upcall-less recovery that was added
5726 in 1.4.2. When trying to reconnect to servers, clients will
5727 now try each server in the failover group every 10 seconds. By
5728 default, clients would previously try one server every 25 seconds.
5733 Description: After recovery, certain operations trigger a failed
5734 assertion on a client.
5735 Details : Failing over an mds, using lconf -d --failover, while a
5736 client was doing a readdir() call would cause the client to
5737 LBUG after recovery completed and the readdir() was resent.
5739 Severity : enhancement
5741 Description: Default groups are now added by lconf
5742 Details : You can now run lconf --group <servicename> without having to
5743 manually add groups with lmc.
5746 Frequency : occasional
5748 Description: Nodes with an elan id of 0 trigger a failed assertion
5751 Frequency : always when accessing e.g. tty/console device nodes
5753 Description: tty and some other devices nodes cannot be used on lustre
5754 Details : file's private_data field is used by device data and lustre
5755 values in there got lost. New field was added to struct file to
5756 store fs-specific private data.
5759 Frequency : when exporting Lustre via NFS
5761 Description: NFSD failed occasionally when looking up a path component
5762 Details : NFSD is looking up ".." which was broken in ext3 directories
5763 that had grown large enough to become hashed.
5766 Frequency : Clusters with multiple interfaces not on the same subnet
5768 Description: Nodes will repeatedly try to reconnect to an interface which it
5769 cannot reach and report an error to the log.
5770 Details : Extra peer list entries will be created by lconf with some peers
5771 unreachable. lconf now validates the peer before adding it.
5774 Frequency : Only if a default stripe is set on the filesystem root.
5776 Description: Setting a default stripe on the filesystem root prevented the
5777 filesystem from being remounted.
5778 Details : The client was sending extra request flags in the root getattr
5779 request and did not allocate a reply buffer for the dir EA.
5782 Frequency : occasional, higher if lots of files are accessed by one client
5783 Bugzilla : 6159, 6097
5784 Description: Client trips assertion regarding lsm mismatch/magic
5785 Details : While revalidating inodes the VFS looks up inodes with ifind()
5786 and in rare cases can find an inode that is being freed.
5787 The ll_test_inode() code will free the lsm during ifind()
5788 when it finds an existing inode and then the VFS later attaches
5789 this free lsm to a new inode.
5793 Bugzilla : 6422, 7030
5794 Description: MDS deadlock between mkdir and client eviction
5795 Details : Creating a new file via mkdir or mknod (starting a transaction
5796 and getting the ns lock) can deadlock with client eviction
5797 (gets ns lock and trying to finish a synchronous transaction).
5800 Frequency : occasional
5801 Description: While starting a server, the fsfilt_ext3 module could not be
5803 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5804 kernels. Previously, lconf would still use the ext3 name
5805 when trying to load modules. Now, it will correctly use
5806 ext3 on 2.4 and ldiskfs on 2.6.
5808 Severity : enhancement
5809 Description: The default stripe count has been changed to 1
5810 Details : The interpretation of the default stripe count (0, to lfs
5811 or lmc) has been changed to mean striping across a single
5812 OST, rather than all available. For general usage we have
5813 found a stripe count of 1 or 2 works best.
5815 Severity : enhancement
5816 Description: Add support for compiling against Cray portals.
5817 Details : Conditional compiling for some areas that are different
5821 Frequency : occasional
5822 Bugzilla : 6409, 6834
5823 Description: Creating files with an explicit stripe count may lead to
5824 a failed assertion on the MDS
5825 Details : If some OSTs are full or unavailable, creating files may
5826 trigger a failed assertion on the MDS. Now, Lustre will
5827 try to use other servers or return an error to the
5831 Frequency : occasional
5833 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5834 caused lock timeout during readahead (since 1.4.2).
5835 Details : Processes doing readahead might match a lock that hasn't been
5836 granted yet if there are overlapping and conflicting lock
5837 requests. The readahead process waits on ungranted lock
5838 (original lock is CBPENDING), while OST waits for that process
5839 to cancel CBPENDING read lock and eventually evicts client.
5841 Severity : enhancement
5843 Description: Initial enabling of flock support for clients
5844 Details : Implements fcntl advisory locking and file status functions.
5845 This feature is provided as an optional mount flag (default
5846 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5847 locking are implemented yet, and those that are are not guaranteed
5848 to be completely correct in production environments.
5849 mount -t lustre -o [flock|noflock] ...
5852 Frequency : occasional
5854 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5855 assertion in the ext3 JBD (journaling) layer.
5856 Details : The b_committed_data struct is protected by the big kernel lock
5857 in 2.4 kernels, serializing journal_commit_transaction() and
5858 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5859 access to this struct. In 2.6 kernels there is finer grained
5860 locking to improve SMP performance of the JBD layer.
5864 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5869 Description: merge and cleanup kernel patches.
5872 -----------------------------------------------------------------------------
5874 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5876 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5877 versions of Lustre. Please update all servers and clients to
5878 version 1.4.2 or later at the same time. You must also run
5879 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5880 to update the configuration logs.
5882 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5883 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5884 - swab llog records of type '0' so we get proper header size/idx (5861)
5885 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5886 - fix rename of one directory over another leaking an inode (5953)
5887 - avoid SetPageDirty on 2.6 (5981)
5888 - don't re-add just-being-destroyed locks to the waiting list (5653)
5889 - when creating new directories, inherit the parent's custom
5890 striping settings if present parent (3048)
5891 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5892 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5893 - don't hold client locks on temporary worklist from l_lru (5666)
5894 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5895 - automatically evict dead clients (5921)
5896 - Update file size properly in create+truncate+fstat case (6196)
5897 - Do not unhash mountpoint dentries, do not allow removal of
5899 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5900 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5901 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5902 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5903 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5904 - Correctly handle reads of files with no objects (6243)
5905 - lctl recover will also mark a device active if deactivate used (5933)
5907 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5908 - allow --write-conf on an MDS with different nettype than client (5619)
5909 - don't write config llogs to MDS for mounts not from that MDS (5617)
5910 - lconf should create multiple TCP connections from a client (5201)
5911 - init scripts are now turned off by default; run chkconfig --on
5912 lustre and chkconfig --on lustrefs to use them
5913 - upcalls are no longer needed for clients to recover to failover
5915 - add --abort-recovery option to lconf to abort recovery on device
5917 - add support for an arbitrary number of OSTs (3026)
5918 - Quota support protocol changes.
5919 - forward compatibility changes to wire structs (6007)
5920 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5921 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5922 - improved "lustre status" script
5923 - initialize blocksize for non-regular files (6062)
5924 - added --disable-server and --disable-client configure options (5782)
5925 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5926 - Vanilla 2.4.29 support
5927 - increase maximum number of obd devices to 520 (6242)
5928 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5929 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5930 SLES 9 2.6.5 kernels
5931 - limit stripes per file to 160 (the maximum EA size) (6093)
5933 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5936 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5937 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5938 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5939 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5940 - small ext3 extents cleanups and fixes (5733)
5941 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5942 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5943 - avoid needless client->OST connect, fix handle mismatch (5317)
5944 - fix DLM error path that led to out-of-sync client, long delays (5779)
5945 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5946 - fix several locking issues related to i_size (5492,5624,5654,5672)
5947 - don't move pending lock onto export if it is already evicted (5683)
5948 - fix kernel oops when creating .foo in unlinked directory (5548)
5949 - fix deadlock in obdfilter statistics vs. object create (5811)
5950 - use time_{before,after} to avoid timer jiffies wrap (5882)
5951 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5952 - Do not leak request if server was not able to process it (5154)
5953 - If mds_open unable to find parent dir, make that negative lookup(5154)
5954 - don't create new directories with extent-mapping (5909, 5936)
5956 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5957 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5958 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5959 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5960 - support for RHEL4 kernel on i686 (5773)
5961 - provide error messages when incompatible logs are encountered (5898)
5963 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5964 * version 1.4.0.10 (1.4.1 release candidate 1)
5966 - don't keep a lock reference when lock is not granted (4238)
5967 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5968 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5969 - drop import inflight refcount on signal_completed_replay error (5255)
5970 - unlock page after async write error during send (3677)
5971 - handle missing objects in filter_preprw_read properly (5265)
5972 - no transno return for symlink open, don't save no-trasno open (3440)
5973 - don't try to complete elan receive that already failed (4012)
5974 - free RPC server reply state on error (5406)
5975 - clean up thread from ptlrpc_start_thread() on error (5160)
5976 - readahead could read extra page into cache that wasn't ejected (5388)
5977 - prevent races in class_attach/setup/cleanup/detach (5260)
5978 - don't dereference de->d_inode after l_dput of de (5458)
5979 - use "int" for stripe value returned from lock_to_stripe (5544)
5980 - mballoc allocation and error-checking fixes in 2.6 (5504)
5981 - block device patches to fix I/O request sizes in 2.6 (5482)
5982 - look up hostnames for IB nals (5602)
5983 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5984 - don't start multiple acceptors for the same port (5277)
5985 - fix incorrect LASSERT in mds_getattr_name (5635)
5986 - export a proc file for general "ping" checking (5628)
5987 - fix "lfs check" to not block when the MDS is down (5628)
5989 - service request history (4965)
5990 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5991 - create an "evict_client" /proc entry on OSTs, like the MDS has
5992 - fix mount usage message, return errors per mount(8) (5168)
5993 - change grep [] to grep "[]" in tests so they work in more UMLs
5994 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5995 - remove ancient LOV_MAGIC_V0 EA support (5047)
5996 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5997 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5998 - allow readahead allocations to fail when low on memory (5383)
5999 - mmap locking landed again, after considerable improvement (2828)
6000 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6002 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6005 - send OST transaction number in read/write reply to free req (4966)
6006 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6007 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6008 - the watchdog thread now runs as interruptible (5246)
6009 - flock/lockf fixes (but it's still disabled, pending 5135)
6010 - don't use EXT3 constants in llite code (5094)
6011 - memory shortage at startup could cause assertion (5176)
6013 - reorganization of lov code
6014 - single portals codebase
6016 - add extents/mballoc support (5025)
6017 - direct I/O reads in the obdfilter (4048)
6018 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6020 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6023 - send OST transaction number in read/write reply to free req (4966)
6024 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6025 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6026 - the watchdog thread now runs as interruptible (5246)
6027 - handle missing objects in filter_preprw_read properly (5265)
6028 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6029 - drop import inflight refcount on signal_completed_replay error (5255)
6030 - unlock page after async write error during send (3677)
6031 - return original error code on reconstructed replies (3761)
6032 - no transno return for symlink open, don't save no-trasno open (3440)
6034 - add pid to ldlm debugging output (4922)
6035 - bump the watchdog timeouts -- we can't handle 30sec yet
6036 - extra debugging for orphan dentry/inode bug (5259)
6038 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6041 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6042 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6043 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6044 - fix NULL dereference in /proc/sys/portals/routes (4827)
6045 - allow failed mdc_close() operations to be interrupted (4561)
6046 - stop precreate on OST before MDS would time out on it (4778)
6047 - don't send partial-page writes before EOF from client (4410)
6048 - discard client grant for sub-page writes on large-page clients (4520)
6049 - don't free dentries not owned by NFS code, check generation (4806)
6050 - fix lsm leak if mds_create_objects() fails (4801)
6051 - limit debug_daemon file size, always print CERROR messages (4789)
6052 - use transno after validating reply (3892)
6053 - process timed out requests if import state changes (3754)
6054 - update mtime on OST during writes, return in glimpse (4829)
6055 - add mkfsoptions to LDAP (4679)
6056 - use ->max_readahead method instead of zapping global ra (5039)
6057 - don't interrupt __l_wait_event() during strace
6059 - add software watchdogs to catch hung threads quickly (4941)
6060 - make lustrefs init script start after nfs is mounted
6061 - fix CWARN/ERROR duplication (4930)
6062 - return async write errors to application if possible (2248)
6063 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6064 - print NAL number in %x format (4645)
6065 - update barely-supported suse-2.4.21-171 series (4842)
6066 - support for sles 9 %post scripts
6067 - support for building 2.6 kernel-source packages
6068 - support for sles km_* packages
6070 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6073 - ignore -ENOENT errors in osc_destroy (3639)
6074 - notify osc create thread that OSC is being cleaned up (4600)
6075 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6076 - reconstruct ost_handle() like mds_handle() (4657)
6077 - create a new thread to do import eviction to avoid deadlock (3969)
6078 - let lconf resolve symlinked-to devices (4629)
6079 - don't unlink "objects" from directory with default EA (4554)
6080 - hold socknal file ref over connect in case target is down (4394)
6081 - allow more than 32000 subdirectories in a single directory (3244)
6082 - fix blocks count for O_DIRECT writes (3751)
6083 - OST returns ENOSPC from object create when no space left (4539)
6084 - don't send truncate RPC if file size isn't changing (4410)
6085 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6086 - bind to privileged port in socknal and tcpnal (3689)
6088 - rate limit CERROR/CWARN console message to avoid overload (4519)
6089 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6090 - basic mmap support (3918)
6091 - kernel patch series update from b1_4 (4711)
6093 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6096 - avoid crash during MDS cleanup with OST shut down (2775)
6097 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6098 - don't use bad inodes on the MDS (3744)
6099 - dynamic object preallocation to improve recovery speed (4236)
6100 - don't hold spinlock over lock dumping or change debug flags (4401)
6101 - don't zero obd_dev when it is force cleaned (3651)
6102 - print grants to console if they go negative (4431)
6103 - "lctl deactivate" will stop automatic recovery attempts (3406)
6104 - look for existing locks in ldlm_handle_enqueue() (3764)
6105 - don't resolve lock handle twice in recovery avoiding race (4401)
6106 - revalidate should check working dir is a directory (4134)
6108 - don't always mark "slow" obdfilter messages as errors (4418)
6110 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6113 - don't close LustreDB during write_conf until it is done (3860)
6114 - fix typo in lconf for_each_profile (3821)
6115 - allow dumping logs from multiple threads at one time (3820)
6116 - don't allow multiple threads in OSC recovery (3812)
6117 - fix debug_size parameters (3864)
6118 - fix mds_postrecov to initialize import for llog ctxt (3121)
6119 - replace config semaphore with spinlock (3306)
6120 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6121 - don't allow enqueue to complete on a destroyed export (3822)
6122 - down write_lock before checking llog header bitmap (3825)
6123 - recover from lock replay timeout (3764)
6124 - up llog sem before sending rpc (3652)
6125 - reduce ns lock hold times when setting kms (3267)
6126 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6127 - fix NULL deref and obd_dev leak on setup error (3312)
6128 - replace some LBUG about llog ops with error handling (3841)
6129 - don't match INVALID dentries from d_lookup and spin (3784)
6130 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6131 - fix invalid assertion in ptlrpc_set_wait (3880)
6133 - add libwrap support for the TCP acceptor (3996)
6134 - add /proc/sys/portals/routes for non-root route listing (3994)
6135 - allow setting MDS UUID in .xml (2580)
6136 - print the stack of a process that LBUGs (4228)
6138 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6141 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6142 - make sure to unset replay flag from failed open requests (3440)
6143 - if default stripe count is 0, use OST count for inode size (3636)
6144 - update parent mtime/ctime on client for create/unlink (2611)
6145 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6146 - free recovery state on server during a forced cleanup (3571)
6147 - unregister_reply for resent reqs (3063)
6148 - loop back devices mounting and status check on 2.6 (3563)
6149 - fix resource-creation race that can provoke i_size == 0 (3513)
6150 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6151 - more debugging for page-accounting assertion (3746)
6152 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6153 - avoid deadlock after precreation failure (3758)
6154 - fix race and lock order deadlock in orphan handling (3450, 3750)
6155 - add validity checks when grabbing inodes from l_ast_data (3599)
6157 - add /proc/.../recovery_status to obdfilter (3428)
6158 - lightweight CDEBUG infrastructure, debug daemon (3668)
6159 - change default OSC RPC parameters to be better on small clusters
6160 - turn off OST read cache for files smaller than 32MB
6161 - install man pages and include them in rpms (3100)
6162 - add new init script for (un)mounting lustre filesystems (2593)
6163 - run chkconfig in %post for init scripts (3701)
6164 - drop scimac NAL (unmaintained)
6166 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6169 - clean kiobufs before and after use (3485)
6170 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6171 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6172 - change init script to fail more clearly if not run as root (1528)
6173 - allow clients to reconnect during replay (1742)
6174 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6175 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6176 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6177 - fix cond_resched (3554)
6178 - only evict unfinished clients after recovery (3515)
6179 - allow bulk resend, prevent data loss (3570)
6180 - dynamic ptlrpc request buffer allocation (2102)
6181 - don't allow unlinking open directory if it isn't empty (2904)
6182 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6183 - remove extraneous obd dereference causing LASSERT failure (3334)
6184 - don't use get_cycles() when creating temp. files on the mds (3156)
6185 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6186 - handle EEXIST for set-stripe, set proper directory name (3336)
6188 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6189 - fix ksocknal_fmb_callback() error messages (2918)
6191 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6194 - don't copy lvb into (possibly NULL) reply on error (2983)
6195 - don't deref dentry after dput, don't free lvb on error (2922)
6196 - use the kms to determine writeback rpc length (2947)
6197 - increment oti_logcookies when osc is inactive (2948)
6198 - update client's i_blocks count via lvb messages (2543)
6199 - handle intent open/close of special files properly (1557)
6200 - mount MDS with errors=remount-ro, like obdfilter (2009)
6201 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6202 - don't use cancelling-locks' kms values (2947)
6203 - use highest lock extent for kms, not last one (2925)
6204 - don't dereference ERR_PTR() dentry in error handling path (3107)
6205 - fix thread race in portals_debug_dumplog() (3122)
6206 - create lprocfs device entries at setup instead of at attach (1519)
6207 - common AST error handler, don't evict client on completion race (3145)
6208 - zero nameidata in detach_mnt in 2.6 (3118)
6209 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6210 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6211 - initialize RPC timeout timer earlier for 2.6 (3219)
6212 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6213 - print nal/nid for unknown nid (3258)
6214 - additional checks for oscc recovery before doing precreate (3284)
6215 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6216 - don't crash in mdc_close for bad permissions on open (3285)
6217 - zero i_rdev for non-device files (3147)
6218 - clear page->private before handing to FS, better assertion (3119)
6219 - tune the read pipeline (3236)
6220 - fix incorrect decref of invalidated dentry (2350)
6221 - provide read-ahead stats and refine rpc in flight stats (3328)
6222 - don't hold journal transaction open across create RPC (3313)
6223 - update atime on MDS at close time (3265)
6224 - close LDAP connection when recovering to avoid server load (3315)
6225 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6226 - don't leak open file on MDS after open resend (3325)
6227 - serialize filter_precreate and filter_destroy_precreated (3329)
6228 - loop device shouldn't call sync_dev() for nul device (3092)
6229 - clear page cache after eviction (2766)
6230 - resynchronize MDS->OST in background (2824)
6231 - refuse to mount the same filesystem twice on same mountpoint (3394)
6232 - allow llmount to create routes for mounting behind routers (3320)
6233 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6234 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6235 - fix rare problem with rename on htree directories (3417)
6237 - allow default OST striping configuration per directory (1414)
6238 - fix compilation for qswnal for 2.6 kernels (3125)
6239 - increase maximum number of MDS request buffers for large systems
6240 - change liblustreapi to be useful for external progs like lfsck (3098)
6241 - increase local configuration timeout for slow disks (3353)
6242 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6244 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6247 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6248 - don't overwrite extent policy data in reply if lock was blocked (2901)
6249 - drop filter export grants atomically with removal from device (2663)
6250 - del obd_self_export from work_list in class_disconnect_exports (2908)
6251 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6252 - swab reply message in mdc_close, other PPC fixes (2464)
6253 - fix destroying of named logs (2325)
6254 - overwrite old logs when running lconf --write_conf (2264)
6255 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6256 - fix race in target_handle_connect (2898)
6257 - mds_reint_create() should take same inode create lock (2926)
6258 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6259 - don't close files for self_export to avoid uninitialized obd (2936)
6260 - allow MDS with the same name as client node (2939)
6261 - hold dentry reference for closed log files for unlink (2325)
6262 - reserve space for all logs during transactions (2059)
6263 - don't evict page beyond end of stripe extent (2925)
6264 - don't oops on a deleted current working directory (2399)
6265 - handle hard links to targets without a parent properly (2517)
6266 - don't dereference NULL lock when racing during eviction (2867)
6267 - don't grow lock extents when lots of conflicting locks (2919)
6269 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6272 - account for cache space usage on clients to avoid data loss (974)
6273 - lfsck support in lustre kernel code (2349)
6274 - reduce journal credits needed for BRW writes (2370)
6275 - orphan handling to avoid losing space on client/server crashes
6276 - ptlrpcd can be blocked, stopping ALL progress (2477)
6277 - use lock value blocks to assist in proper KMS, faster stat (1021)
6278 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6279 - recovery for initial connections (2355)
6280 - fixes for mds_cleanup_orphans (1934)
6281 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6282 - block all file creations until orphan recovery completes (1901)
6283 - client remove rq_connection from request struct (2423)
6284 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6285 - recovery timer race (2670)
6286 - mdc_close recovey bug (2532)
6287 - ptlrpc cleanup bug (2710)
6288 - mds timeout on local locks (2588)
6289 - namespace lock held during RPCs (2431)
6290 - handle interrupted sync write properly (2503)
6291 - don't try to handle a message that hasn't been replied to (2699)
6292 - client assert failure during cleanup after abort recovery (2701)
6293 - leak mdc device after failed mount (2712)
6294 - ptlrpc_check_set allows timedout requests to complete (2714)
6295 - wait for inflight reqs when ptlrpcd finishes (2710)
6296 - make sure unregistered services are removed from the srv_list
6297 - reset bulk XID's when resending them (caught by 1138 test)
6298 - unregister_bulk after timeout
6299 - fix lconf error (2694)
6300 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6301 - readahead locks pages, leaves pending causing memory pressure (2673)
6302 - increase OST request buffers to 4096 on large machines (2729)
6303 - fix up permission of existing directories in simple_mkdir (2661)
6304 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6305 - don't assign transno to errored transactions (2742)
6306 - don't delete objects on OST if given a bogus objid from MDS (2751)
6307 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6308 - if rq_no_resend, then timeout request after recovery (2432)
6309 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6310 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6311 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6312 - MDS recovery shouldn't delete objects that it has given out (2730)
6313 - if enqueue arrives after completion, don't clobber LVB (2819)
6314 - don't unlock pages twice when trigger_group_io returns error (2814)
6315 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6316 - don't write pages to disk if there was an error (1450)
6317 - don't ping imports that have recovery disabled (2676)
6318 - take buffered bytes into account when balancing socknal conn (2817)
6319 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6320 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6321 - remove little-endian swabbing of llog records (1987)
6322 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6323 - retry reposting request buffers if they fail (1191)
6324 - grow extent at grant time to avoid granting a revoked lock (2809)
6325 - lock revoke doesn't evict page if covered by a second lock (2765)
6326 - disable VM readahead to avoid reading outside lock extents (2805)
6328 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6329 - updated kernel patches for hp-2.4.20 kernel (2681)
6331 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6334 - fix truncated write corruption (2366)
6335 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6337 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6338 - ensure len(uuid) < 37 in lmc (1171)
6339 - fix ia64 OOPS in llog_test (2255)
6340 - zero end of page at obdfilter for partial page writes (2648)
6341 - don't leave stale dentries around after renames (bug 2428)
6342 - fix timeouts when evicting a client with a single lock held (2642)
6343 - set deadline for the initial HELLO message to drain (2634)
6344 - print out dotted-quad IP addresses in the socknal (2302)
6346 - additional debugging for MDS client eviction problem (2443)
6347 - fix mkfsoptions support for osts (2603, 2604)
6349 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6352 - add series for the vanilla 2.6.0 kernel
6353 - add series for the vanilla 2.4.24 kernel
6354 - add series for a cray x86/64 UL kernel drop
6355 - fix xattr patches for the vanilla 2.4.19 series
6357 - generate true UUIDs in lmc (1171)
6358 - have portals stack dumping break in UML (2466)
6359 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6360 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6361 - wake callback waiting threads on client eviction (2460)
6362 - Add --ptldebug and --subsystem to lmc (1719)
6363 - update assertion to allow safe interrupt allocation
6364 - set rq_no_resend for cancel requests (2432)
6365 - recalculate ptlrpcd timeout after resend (2494)
6366 - call vfs_rmdir when removing pending directories (2368)
6367 - fix renaming a file to itself (2429)
6368 - lmc creates a default one-stripe lov (2454)
6369 - expand procfs space to handle large clusters (2326)
6370 - increase UML stack to avoid overflow
6371 - update lconf's list of debug and subsystem masks
6372 - fix lfs find --obd (2510)
6373 - /proc tunable for disabling filter read caching (2591)
6374 - stop rpm packages from altering slapd.conf (2301)
6375 - disable nagle in the socknal under 0conf (2578)
6376 - choose mds inode size based on stripe count (2572)
6377 - fix kernel-source rpm problems (2516)
6379 - add --disable-doc to avoid pdf generation (2421)
6380 - update documentation, tests, type-os, comments
6381 - avoid format warnings on ia64
6382 - remove the TOE NAL
6383 - tiny code cleanups by removing unused fields
6385 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6388 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6389 - remove the most blatant lies from BUILDING file (2371)
6390 - change default debug level to reasonable production setting
6391 - reduce client side cache size to reduce cache flush time
6392 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6393 - make TCP zerocopy and pinger support enabled by default (2476)
6394 - sync writes completed after process exits caused crashes (2319)
6395 - maintain correct mount count on the MDS (2356)
6396 - backout 1557, because 2316 wasn't really fixed
6397 - better file I/O statistics gathering in /proc
6398 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6399 - another kernel patch to fix zero-copy TCP function export
6400 - don't take duplicate lock when processing re-sent getattr (2420)
6401 - lctl uses obd_self_export instead of creating new conn (2353)
6402 - MDS/OST recovery case which requires object creation asserted (2425)
6403 - move lfs from /usr/sbin to /usr/bin in packages
6404 - fix race between mds_client_add and mds_client_free (2417)
6405 - use kmalloc instead of slabs in portals (2430)
6406 - don't create duplicate records when a failover MDS is present (2442)
6407 - remove unnecessary mount age check (2332)
6408 - don't remove directory inodes from locks prematurely (2451)
6409 - don't break if MDS service name is the same as hostname (2103)
6410 - fix races in client write RPC generation when cache full (2482)
6412 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6415 - remove now-unused request->rq_obd (278)
6416 - if an allocation fails, print out how much memory we've used (1933)
6417 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6418 - add the "configurable stack size" patch to most series files (1256)
6419 - ability to write large log records, for 100+ OST configs (2306)
6420 - fix NULL deref when filter_prep fails (2314)
6421 - fix operator precedence error in filter_sync
6422 - dynamic allocation of socknal TX descriptors (2315)
6423 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6424 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6425 - ensure that CWARN messages in Portals always get to the syslog
6426 - __init/__exit are not for prototype decls (ldlm_init/exit)
6427 - x86-64 compile warning fixes
6428 - fix gateway LMC keyword conflict (2318)
6429 - fix MDS lock inversions in getattr/reint paths (1844)
6430 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6431 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6432 - DLM race condition prevented some lock evictions (2328)
6433 - ENOMEM detection and retry on socknal sends (2230)
6434 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6435 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6436 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6437 - better error messages when a client is rejected during recovery (1505)
6438 - avoid cancelling locks which were never granted, after failure (2330)
6439 - fix i_sem/journal inversion in mds_client_add (2333)
6440 - fix truncate/getattr lock cycle deadlock (2334)
6441 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6442 - fix two rare exit paths which could leak an l_lock() ref (2321)
6443 - fencepost error in MDS/OST orphan recovery (2226)
6444 - make log record alignment 8 bytes (1988)
6445 - lstripe now fails when requested offset > ost_count (2237)
6446 - ensure that all kernel series have a complete list.h (1607)
6447 - fix crashes in special-file operations (2316)
6448 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6450 - allow configurable automake binary, for testing new versions
6451 - small update to the lfs documentation
6453 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6455 * fix negative export reference count in fsfilt_sync (2312)
6457 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6458 * release candidate 0.9.1
6460 - orphans are moved into the PENDING directory for possible recovery
6461 - replayed opens now open by fid for orphan/rename safety (1042)
6462 - last close of an orphan inode generates a transno (683)
6463 - chdir() and mount() now pin the directory entry (1020)
6464 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6465 - discard very old requests without processing them (1502)
6466 - remove some common, well-understood CERRORs (1505)
6467 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6468 - clear "grant" flags in OST replies until OST grant code lands (1644)
6469 - fix read performance by not clobbering i_blksize on client (1598)
6470 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6471 - make LRU size a /proc tunable, clears locks when reduced (707)
6472 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6473 - server locks take references on exports now (1558)
6474 - build fixes for 2.4.20-rh trees (1663)
6475 - return an error from lov_create if all OSCs are inactive (1751)
6476 - fix import levels when a reconnect happens without a timeout (1597)
6477 - exit early from mds_open if we get a lookup error (1749)
6478 - partial page read at EOF wouldn't wait for disk before sending (1642)
6479 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6480 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6481 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6482 - write/truncate lock inversion (1639)
6483 - Don't auto-load obdclass, portals modules during cleanup (1495)
6484 - fix timestamps from jumping to "now" (1763)
6485 - extra journal assertions (1648)
6486 - add an extra multiunlink test (1771)
6487 - fix read_record/write_record API (1776)
6488 - fix leak of offset_extent, possible incorrect i_size later (1772)
6489 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6490 - Debugging for the kqswnal_get_idle_tx problems (1820)
6491 - Allow recovery to be attempted multiple times (1536)
6492 - Write out MDS last_rcvd file after it is first created (1600)
6493 - Fix tx_descriptor leak in failed transmit situations (1827)
6494 - ext3 journaling fixes for assertion failure after IO error (1871)
6495 - class_export_put() on freed export after completion AST error (1896)
6496 - Fix revalidate looping in VFS (1322)
6497 - Don't access a freed export during MDS_REINT timeout (1521)
6498 - Add open-unlink recovery support on the MDS (1673,1764)
6499 - Return an error if no MDS data was read from last_rcvd (1946)
6500 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6501 - Don't setup a disk device that doesn't match exported UUID (317)
6502 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6503 - avoid committing NULL handle in force close
6504 - local.sh is now a one-stripe LOV configuration
6505 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6506 - fix ext3 htree duplicate directory entry corruption (1516)
6507 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6508 - update to new LOV EA format (2097)
6509 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6511 2003-06-15 Phil Schwan <phil@clusterfs.com>
6514 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6515 - per-import/export recovery handling (958, 931, 959)
6516 - multiple last-rcvd slots, for serving multiple FSes (949)
6517 - connections are again shared between multiple imp/exports (963, 964)
6518 - "umount -f" would hang if any requests needed to be sent (393, 978)
6519 - avoid pinning large req buffer by copying for queued messages (989)
6520 - add "uuid" to "lctl device" command to help upcalls (991)
6521 - "open" RPCs with transnos would confuse recovery counters (1037)
6522 - do proper endian conversion of all wire messages (288, 340, 891)
6523 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6524 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6525 - fix ptlrpc_connection leak in target_handle_connect (1174)
6526 - fix import refcounting bug in OST and MDS cleanup (1134)
6527 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6528 - fix very unlikely obd_types race condition (501)
6529 - remove osc_open hack for echo_client (1187)
6530 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6531 - a failure in read_inode2 leads to deadlock (1139)
6532 - cancel ack-locks as soon as transaction is committed (1072)
6533 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6534 - make sure to commitrw after any preprw to avoid deadlock (1162)
6535 - failing to execute a file in a lustre FS would lock inode (1203)
6536 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6537 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6538 - obd_uuid2tgt would walk off the end of the list (1255)
6539 - on IA64 the osc would give portals incorrect bulk size (1258)
6540 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6541 - fix lock inversion caused by new llite matching code (1282)
6542 - limit the number of dirty pages on a client to 10MB (1286)
6543 - timed out locks were not being corrected cancelled (1289)
6544 - fix O_DIRECT above 4GB on IA-32 (1292)
6545 * major user-visible changes
6546 - fail out/fail over policy now controlled by the upcall (993)
6548 - add OBD_PING to check server availability and failure (954)
6549 - lustre messages are now sent in sending host order (288, 340, 891)
6550 - add eadatalen to MDS getattr reply (340)
6551 - OST read replies may contain second buffer, with per-page status (593)
6553 2003-03-11 Phil Schwan <phil@clusterfs.com>
6556 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6557 - failed open()s could cause deadlock; fixed (867, 869)
6558 - stop cancelling OST locks when files are closed (481)
6559 - overlapping XID spaces caused network corruption (851, 853)
6560 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6561 - setattr_raw added, to do single-RPC, server-side setattrs
6562 - lmc/lconf syntax change for OST UUIDs
6563 - fix crashy race condition between ptlrpc_free_req and osc_close
6564 - don't use request in mdc_enqueue if we hit a timeout (889)
6565 - don't set the inode i_size for regular files from the MDS (896)
6566 - handle out of order completion AST (842)
6567 - don't LBUG if a lock request times out after receiving AST (913)
6568 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6569 - if a bad lock AST arrives, send an error instead of dropping entirely
6570 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6571 - fix leak in bulk IO when only partially completed (899, 900, 926)
6572 - fix O_DIRECT for ia64 (55)
6573 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6574 - C-z after timeout could hang a process forever; fixed (977)
6576 - client-side I/O cache (678, 924, 929, 941, 970)
6578 - READPAGE and SETATTRs which don't take server-side locks get
6581 2003-02-11 Phil Schwan <phil@clusterfs.com>
6584 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6585 - Fix fsfilt_extN_readpage() to read a full page of directory
6586 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6587 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6588 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6589 - Most intent operations were removed, in favour of a new RPC mode
6590 that does a single RPC to the server and bypasses most of the VFS
6591 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6592 - Aggressively cancel local locks on DLM servers
6593 - mds_reint_unlink sends EA to the client if it's the last nlink.
6594 client uses that EA to unlink OST objects.
6595 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6596 - recursive symlinks were fixed (439)
6597 - fixed NULL deref in DEBUG_REQ
6598 - filter_update_lastobjid no longer calls sync, which annoyed extN
6599 - fixed multi-client small-writes to a single file problem (445)
6600 - fixed mtime updates during file writes (607)
6601 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6602 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6603 - fixed Linux OST size reporting problem (444, 656)
6604 - OST now updates object mtime with writes or setattr (607, 619)
6605 - client verifies file size before zeroing page past EOF (445)
6606 - OST now writes last allocated objid to disk with allocation (108)
6607 - LOV on echo now works (409)
6609 - mds_reint_unlink sends a new buffer, with the EA included. this
6610 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6611 set if a regular file was being unlinked, and it was the last link
6612 - use PtlGet from the target for bulk writes (315)
6613 - OST now updates object mtime with writes or setattr (607, 619)
6614 - LDLM now has a grant-time callback to revalidate locked items, if
6616 - Many MDS operations were reorganized to combat race conditions
6618 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6619 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6621 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6624 - Fully reactivate OST imports after reconnection (512, others)
6625 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6626 - More graceful error handling for truncating on dead OST (515)
6627 - Don't error out unless we're actually accessing dead stripes (474)
6628 - Fix garbage sizes when stripes are missing (410)
6629 - LRU counters were broken, causing constant lock purge (433, 432)
6630 - garbage on read from stripes with failed OSTs (441)
6631 - mark OSCs as active before reconnecting during recovery (438)
6632 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6633 - lfind did not preserve OST order in output (443)
6634 - symlinks cause hung clients, incorrect data (439)
6635 - stop dereferencing request after dropping refcount (457)
6636 - don't LASSERT(spin_is_locked) on non-SMP (455)
6637 - fixes for many rename() bugs
6638 - fstat didn't correctly synchronize attributes (399)
6639 - server must handle lock cancellation during blocking AST prep (487)
6640 - bulk descriptors were free()d too soon (511)
6641 - fix paths in lconf, which would load incorrect modules (451, 507)
6642 - fix confusing lconf 'host not found' error message (386)
6643 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6644 - fix race condition in mdc_blocking_ast() for inode access (526)
6645 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6646 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6647 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6648 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6649 - fix lstripe in several ways (kernel side) (527)
6650 - fix request leak in ldlm_cli_enqueue (262)
6651 - incorrect OSC was marked inactive after OST failure
6652 - call mds_fs_cleanup before unmounting filesystem (524)
6653 - fix races between taking ns_lock and ldlm_lock_change_resource
6654 - fix races updating LOV export open file list
6655 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6656 - fix recovery NULL deref in ldlm_cli_cancel_unused
6657 - fix some DLM races by using new hash table for lock handles (419)
6658 - permit the client to specify desired inodes, at replay
6659 - duplicate requests when we queue them for replay reintegration
6660 - fix last_rcvd offset calculation
6661 - sync after each recovered transaction, so we always make progress
6662 - never, not always, ERESTART requests without transnos
6663 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6665 - skip replay if the MDS says that the client is already connected
6666 - don't check for a recovery-enabled export to match lctl's UUID
6667 - don't INC_USE_COUNT for phantom exports
6668 - don't crash when cleaning up phantom exports (567)
6669 - don't double-finish or set replay data for errored mdc_open requests
6670 - abort requests when they time out, so we don't get old replies
6671 - send/receive replies for AST messages again
6672 - if the client says that it doesn't have the lock, cancel it on the
6674 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6675 mark it as destroyed, it will all work out when decref is called
6676 - fix module use counts (22, 581)
6678 - ASTs now expect a reply (server cancels lock on error reply)
6680 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6683 - fix many simultaneous client startup (392)
6684 - fix dentry->d_it clobbering
6685 - credentials weren't being shipped for readdir/getattr operations
6686 - remove invalid assertions triggered during some concurrent MD
6688 - proper Lustre versions added (336, 389)
6689 - fix memory leak for create error case (398)
6690 - fix LOV locking bug that would get cli/srv out of sync
6691 - fix echo client over LOV (409)
6692 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6693 - fix double-O_EXCL intent crash (424)
6694 - avoid sending multiple lock CANCELs (352)
6696 - MDS can do multi-client recovery (modulo bugs in new code)
6698 - many updates, edits, cleanups
6700 2002-11-18 Phil Schwan <phil@clusterfs.com>
6703 - fix null d_it dereference (346)
6704 - fix full OST/dbench hang (333)
6705 - fix permission problem with file removal (286)
6706 - fix removal of OSCs from LOV when they fail
6707 - fix NULL deref during bulk timeout (214)
6708 - fix problems related to multiple filesystems on one MDS (241)
6709 - fixed serious subtle metadata locking bugs
6710 - free locks on clients when inodes are removed due to memory
6712 - fix inode pointer in lock data (285)
6713 - partial support for multiple MDS on a single host (241)
6714 - data locks weren't cancelled at clear_inode time (290, 311)
6715 - intent locks could lead to unbounded lock growth (205)
6716 - added a maximum lock count, an LRU list, and a flusher
6717 - fix multiple rename (365)
6718 - properly abstracted the echo client
6719 - OSC locked 1 byte too many; fixed
6720 - rewrote brw callback code:
6721 - fixed recovery bugs related to LOVs (306)
6722 - fixed too-many-pages-in-one-write crash (191)
6723 - fixed (again) crash in sync_io_timeout (214)
6724 - probably fixed callback-related race (385)
6726 - Add capability to MDS protocol
6727 - LDLM cancellations and callbacks on different portals
6729 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6732 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6733 - defer open object destruction to close time (601981, 138)
6734 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6735 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6736 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6737 - don't call obd_setattr in ll_file_release for destroyed objects
6738 * protocol change to lustre_msg: move |version| and add |flags|
6739 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6740 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6741 * added replay of create, unlink, link and rename operations during
6742 MDS failover; recovery should be much more robust now
6743 * remove failed OSCs from LOVs (only lov_create uses this so far)
6744 * the lustre-HOWTO was brought (more) up to date (582544)
6746 2002-10-23 Phil Schwan <phil@clusterfs.com>
6749 - in-use dentries weren't being reused properly (617851)
6750 - prevent multiple LDLM setup (599178)
6751 - fix LOV size calculations for truncate (617853)
6752 - fix client handling of MDS intent errors (POSIX)
6753 - fix permission bug in lovstripe.c test (624321)
6754 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6755 - truncate past end of file could corrupt data
6756 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6757 - a race in recovery could return ETIMEDOUT to apps (623947)
6758 - building outside the source directory was fixed
6759 * the lustre-HOWTO was brought (more) up to date (582544)
6760 * major progress was made on recovery functionality
6762 2002-10-10 Phil Schwan <phil@clusterfs.com>
6765 - recovery deadlock fix
6766 - rm -rf causes LBUG fix (617817)
6767 - file open by multiple tasks fix (618962)
6768 - directory permissions bugs (602707 and 620007)
6769 - journal_stop fixed with locking (611313)
6770 - O_APPEND failures resolved (618273, perhaps 614459)
6771 - lconf PATH fix (619770)
6772 - IA64 build fix (621450)
6773 - RPC buffer sizes scale with amount of memory
6775 2002-10-01 Phil Schwan <phil@clusterfs.com>
6778 - locks would be cancelled without throwing away data pages,
6779 resulting in inconsistent data (605627)
6780 - inode attributes were not always being refreshed (605627, 612449)
6781 - lconf now continues to cleanup after lctl reports an error
6782 - MDS now enforces user permissions (602707)
6783 - lprocfs cleanup fixed, but not yet enabled (614157)
6784 - fixed infinite server hang, should a client not respond to an AST
6785 - avoid going into recovery if user calls readlink() with a buffer
6786 that's too small (613941)
6787 - AST RPCs no longer require replies (614867) -- this may be changed
6788 - don't crash server if client sends an IOV that's too big (611336)
6789 - fixed lock conversion deadlock (611892)
6790 - fixed the following of symlinks (614622)
6791 * recovery: the server can remove locks from a client that dies, other
6792 clients can make progress
6793 * more extN patch fixes
6794 * compile-time configurable ptlrpc buffer allocations
6796 - collaborative read cache document
6797 - Lustre Lite Performance CDR document-in-progress
6799 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6802 - fix typo in patch-2.4.18
6804 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6807 - clear ptlrpc request each time in handle_incoming_request()
6808 - unlink of files now destroys the object on the OST
6810 2002-09-19 Peter Braam <braam@clusterfs.com>
6812 * add hard link support
6813 * change obdfile creation method
6814 * kernel patch changed
6816 2002-09-19 Peter Braam <braam@clusterfs.com>
6819 - stack overflow bug in extN fixed
6821 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6823 * documentation updates
6824 - add man pages for config tools
6825 - update tests/README to describe testing with new config tools
6826 - finish metadata API descriptions
6827 * bug fixes and cleanups
6828 - statfs workaround for 16TB limit
6829 - LOV stripe allocation improved, can stripe on subset of OSTs
6830 - LOV file size/IO offset was wrong for files > 4GB in size
6831 - object EA data was being dropped, caused files to be unreadable
6832 - memory overflow with non-LOV OST caused memory corruption
6833 - fixed regression tests to work with new config tools, obdfilter
6834 - fixed bug when directory size became larger than 1 block
6835 - fixed bug (for single client case) when PWD was deleted
6836 - invalidate local directory pages when doing intent-based ops
6837 - avoid LDLM oops when lock callback contained bad data
6839 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6841 * documentation updates
6842 * bug fixes and cleanups
6843 - configuration tools
6846 - 64-bit compile warnings
6847 - 64-bit internal statfs data
6849 * test_brw on persistent OST devices
6851 * lprocfs (disabled)
6853 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6855 * documentation updates
6856 * bug fixes and cleanups
6857 * configuration tools
6859 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6862 * many small fixes to 0_5_4
6863 * io/network handling
6864 * thinkos in MDS operations
6866 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6869 * crucial basic fixes to 0.5.3
6870 * IOR, Iozone work over Elan
6873 2002-08-07 Phil Schwan <phil@clusterfs.com>
6874 * version 0_5_3, our first alpha
6875 * we use the new Portals iovs
6876 * documentation updates
6877 * bug fixes and cleanups
6878 * small changes in the DLM wire protocol
6880 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6881 * version 0_5_1 with some initial stability,
6882 * locking on MD and file I/O.
6883 * documentation updates
6884 * several bug fixes since 0.5.0
6885 * small changes in wire protocol
6887 2002-07-18 Phil Schwan <phil@clusterfs.com>
6889 * delivered as Lustre Light Alpha
6890 * fixed a crash after handling invalid MDS requests
6891 * fixed directory pages for architectures with non-4k pages sizes
6893 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6894 * release version v0_4_4
6895 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6897 * Fixes a number of interruption problems with OST operations.
6898 * Update documentation for portals header changes
6899 * Move all wire protocol structs/defines to lustre_idl.h
6900 * Fixes symlink length bug.
6901 * Add tcpdump to repository.
6903 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6904 * release version v0_4_3
6905 * Fixes statfs for inodes on extN.
6906 * Fixes bug in runtests which would delete /etc/hosts.
6907 * Use 64-bit object IDs wherever possible (not into VFS though)
6908 Remove ost_get_info, which is unused by lustre, and out of date.
6910 2002-07-03 Peter Braam <braam@clusterfs.com>
6911 * release version v0_4_2 Fixes a lookup error (type not passed)
6912 * move forward to head of Portals
6913 * move forward to latest Lustre kernel
6915 2002-06-25 Peter Braam <braam@clusterfs.com>
6916 * release version v0_4_1. Hopefully stable on single node use.