3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server primary kernels built and tested during release cycle:
6 4.18.0-348.2.1.el8 (RHEL8.5)
7 * Other server kernels known to build and work at some point (others may also work):
8 3.10.0-862.14.4.el7 (RHEL7.5)
9 3.10.0-957.27.2.el7 (RHEL7.6)
10 3.10.0-1062.18.1.el7 (RHEL7.7)
11 3.10.0-1127.19.1.el7 (RHEL7.8)
12 3.10.0-1160.49.1.el7 (RHEL7.9)
13 4.18.0-80.11.2.el8 (RHEL8.0)
14 4.18.0-147.8.1.el8 (RHEL8.1)
15 4.18.0-193.28.1.el8 (RHEL8.2)
16 4.18.0-240.22.1.el8 (RHEL8.3)
17 4.18.0-305.25.1.el8 (RHEL8.4)
18 4.4.120-92.70 (SLES12 SP2)
19 4.4.180-94.100 (SLES12 SP3)
20 4.4.0-131 (Ubuntu 16.04)
21 4.15.0-32 (Ubuntu 18.04)
22 5.4.0-48 (Ubuntu 20.04)
23 vanilla linux 5.4.0 (ZFS + ldiskfs)
24 vanilla linux 5.4.21 (ZFS + ldiskfs)
25 vanilla linux 5.4.136 (ZFS + ldiskfs)
26 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
27 * Client primary kernels built and tested during release cycle:
28 4.18.0-348.2.1.el8 (RHEL8.5)
29 5.4.0-37 (Ubuntu 20.04)
30 5.3.18-59.27 (SLES15 SP3)
31 * Other clients known to build on these kernels at some point (others may also work):
32 3.10.0-862.14.4.el7 (RHEL7.5)
33 3.10.0-957.27.2.el7 (RHEL7.6)
34 4.14.0-49.13.1.el7a (RHEL7.5)
35 4.14.0-115.2.2.el7a (RHEL7.6)
36 3.10.0-1062.18.1.el7 (RHEL7.7)
37 3.10.0-1127.19.1.el7 (RHEL7.8)
38 3.10.0-1160.49.1.el7 (RHEL7.9)
39 4.18.0-80.11.2.el8 (RHEL8.0)
40 4.18.0-147.8.1.el8 (RHEL8.1)
41 4.18.0-193.28.1.el8 (RHEL8.2)
42 4.18.0-240.22.1.el8 (RHEL8.3)
43 4.18.0-305.25.1.el8 (RHEL8.4)
44 4.4.120-92.70 (SLES12 SP2)
45 4.4.180-94.100 (SLES12 SP3)
46 4.12.14-95.48 (SLES12 SP4)
47 4.12.14-122.91 (SLES12 SP5)
48 4.12.14-197.75 (SLES15 SP1)
49 5.3.18-24.96 (SLES15 SP2)
50 4.4.0-131 (Ubuntu 16.04)
51 4.15.0-48 (Ubuntu 18.04)
52 5.8.0-53 (Ubuntu 20.04.2 HWE)
53 5.11.0-31 (Ubuntu 20.04.3 HWE)
54 5.11.0 (vanilla kernel.org)
55 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
56 * Recommended ZFS version: 2.0.7
57 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
58 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
59 stack overflow. For more information, please refer to bugzilla 17630.
60 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
61 the re-exporting nodes to be mounted with "32bitapi" mount option
63 --------------------------------------------------------------------------------
68 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
69 for currently supported client and server kernel versions.
70 * Server primary kernels built and tested during release cycle:
71 4.18.0-348.2.1.el8 (RHEL8.5)
72 * Other server kernels known to build and work at some point (others may also work):
73 3.10.0-862.14.4.el7 (RHEL7.5)
74 3.10.0-957.27.2.el7 (RHEL7.6)
75 3.10.0-1062.18.1.el7 (RHEL7.7)
76 3.10.0-1127.19.1.el7 (RHEL7.8)
77 3.10.0-1160.49.1.el7 (RHEL7.9)
78 4.18.0-80.11.2.el8 (RHEL8.0)
79 4.18.0-147.8.1.el8 (RHEL8.1)
80 4.18.0-193.28.1.el8 (RHEL8.2)
81 4.18.0-240.22.1.el8 (RHEL8.3)
82 4.18.0-305.25.1.el8 (RHEL8.4)
83 4.4.120-92.70 (SLES12 SP2)
84 4.4.180-94.100 (SLES12 SP3)
85 4.4.0-131 (Ubuntu 16.04)
86 4.15.0-32 (Ubuntu 18.04)
87 5.4.0-48 (Ubuntu 20.04)
88 vanilla linux 5.4.0 (ZFS + ldiskfs)
89 vanilla linux 5.4.21 (ZFS + ldiskfs)
90 vanilla linux 5.4.136 (ZFS + ldiskfs)
91 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
92 * Client primary kernels built and tested during release cycle:
93 4.18.0-348.2.1.el8 (RHEL8.5)
94 5.4.0-37 (Ubuntu 20.04)
95 5.3.18-59.27 (SLES15 SP3)
96 * Other clients known to build on these kernels at some point (others may also work):
97 3.10.0-862.14.4.el7 (RHEL7.5)
98 3.10.0-957.27.2.el7 (RHEL7.6)
99 4.14.0-49.13.1.el7a (RHEL7.5)
100 4.14.0-115.2.2.el7a (RHEL7.6)
101 3.10.0-1062.18.1.el7 (RHEL7.7)
102 3.10.0-1127.19.1.el7 (RHEL7.8)
103 3.10.0-1160.49.1.el7 (RHEL7.9)
104 4.18.0-80.11.2.el8 (RHEL8.0)
105 4.18.0-147.8.1.el8 (RHEL8.1)
106 4.18.0-193.28.1.el8 (RHEL8.2)
107 4.18.0-240.22.1.el8 (RHEL8.3)
108 4.18.0-305.25.1.el8 (RHEL8.4)
109 4.4.120-92.70 (SLES12 SP2)
110 4.4.180-94.100 (SLES12 SP3)
111 4.12.14-95.48 (SLES12 SP4)
112 4.12.14-122.91 (SLES12 SP5)
113 4.12.14-197.75 (SLES15 SP1)
114 5.3.18-24.96 (SLES15 SP2)
115 4.4.0-131 (Ubuntu 16.04)
116 4.15.0-48 (Ubuntu 18.04)
117 5.8.0-53 (Ubuntu 20.04.2 HWE)
118 5.11.0-31 (Ubuntu 20.04.3 HWE)
119 5.11.0 (vanilla kernel.org)
120 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
121 * Recommended ZFS version: 2.0.7
122 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
123 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
124 stack overflow. For more information, please refer to bugzilla 17630.
125 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
126 the re-exporting nodes to be mounted with "32bitapi" mount option
128 --------------------------------------------------------------------------------
132 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
133 for currently supported client and server kernel versions.
134 * Server primary kernels built and tested during release cycle:
135 3.10.0-1062.18.1.el7 (RHEL7.7)
136 3.10.0-1127.19.1.el7 (RHEL7.8)
137 3.10.0-1160.6.1.el7 (RHEL7.9)
138 * Other server kernels known to build and work at some point (others may also work):
139 3.10.0-862.14.4.el7 (RHEL7.5)
140 3.10.0-957.27.2.el7 (RHEL7.6)
141 4.18.0-80.11.2.el8 (RHEL8.0)
142 4.18.0-147.8.1.el8 (RHEL8.1)
143 4.18.0-193.28.1.el8 (RHEL8.2)
144 4.18.0-240.1.1.el8 (RHEL8.3)
145 4.4.120-92.70 (SLES12 SP2)
146 4.4.180-94.100 (SLES12 SP3)
147 4.4.0-131 (Ubuntu 16.04)
148 4.15.0-32 (Ubuntu 18.04)
149 5.4.0-48 (Ubuntu 20.04)
150 vanilla linux 5.4.0 (ZFS + ldiskfs)
151 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
152 * Client primary kernels built and tested during release cycle:
153 3.10.0-1062.18.1.el7 (RHEL7.7)
154 3.10.0-1127.19.1.el7 (RHEL7.8)
155 3.10.0-1160.6.1.el7 (RHEL7.9)
156 4.12.14-95.48 (SLES12 SP4)
157 4.12.14-122.20 (SLES12 SP5)
158 4.12.14-197.75 (SLES15 SP1)
159 5.3.18-24.24 (SLES15 SP2)
160 4.15.0-48 (Ubuntu 18.04)
161 * Other clients known to build on these kernels at some point (others may also work):
162 3.10.0-862.14.4.el7 (RHEL7.5)
163 3.10.0-957.27.2.el7 (RHEL7.6)
164 4.14.0-49.13.1.el7a (RHEL7.5)
165 4.14.0-115.2.2.el7a (RHEL7.6)
166 4.18.0-80.11.2.el8 (RHEL8.0)
167 4.18.0-147.8.1.el8 (RHEL8.1)
168 4.18.0-193.28.1.el8 (RHEL8.2)
169 4.18.0-240.1.1.el8 (RHEL8.3)
170 4.4.120-92.70 (SLES12 SP2)
171 4.4.180-94.100 (SLES12 SP3)
172 4.4.0-131 (Ubuntu 16.04)
173 5.4.0-37 (Ubuntu 20.04)
174 5.4.0 (vanilla kernel.org)
175 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
176 * Recommended ZFS version: 2.0.0
177 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
178 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
179 stack overflow. For more information, please refer to bugzilla 17630.
180 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
181 the re-exporting nodes to be mounted with "32bitapi" mount option
183 --------------------------------------------------------------------------------
187 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
188 for currently supported client and server kernel versions.
189 * Server primary kernels built and tested during release cycle:
190 3.10.0-1062.1.1.el7 (RHEL7.7)
191 * Other server kernels known to build and work at some point (others may also work):
192 3.10.0-862.14.4.el7 (RHEL7.5)
193 3.10.0-957.27.2.el7 (RHEL7.6)
194 4.18.0-80.7.1.el8 (RHEL8.0)
195 4.4.120-92.70 (SLES12 SP2)
196 4.4.180-94.100 (SLES12 SP3)
197 4.4.0-131 (Ubuntu 16.04)
198 4.15.0-32 (Ubuntu 18.04)
199 vanilla linux 4.6.7 (ZFS only)
200 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
201 * Client primary unpatched kernels built and tested during release cycle:
202 3.10.0-1062.1.1.el7 (RHEL7.7)
203 4.12.14-95.29 (SLES12 SP4)
204 4.15.0-48 (Ubuntu 18.04)
205 * Other clients known to build on unpatched kernels at some point (others may also work):
206 3.10.0-862.14.4.el7 (RHEL7.5)
207 3.10.0-957.27.2.el7 (RHEL7.6)
208 4.14.0-49.13.1.el7a (RHEL7.5)
209 4.14.0-115.2.2.el7a (RHEL7.6)
210 4.18.0-80.7.1.el8 (RHEL8.0)
211 4.4.120-92.70 (SLES12 SP2)
212 4.4.180-94.100 (SLES12 SP3)
213 4.4.0-131 (Ubuntu 16.04)
214 vanilla linux 4.15.0 (ZFS only)
215 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
216 * Recommended ZFS / SPL version: 0.7.13
217 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
218 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
219 stack overflow. For more information, please refer to bugzilla 17630.
220 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
221 the re-exporting nodes to be mounted with "32bitapi" mount option
222 * Note for Ubuntu 14.04.5:
223 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
224 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
226 --------------------------------------------------------------------------------
230 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
231 for currently supported client and server kernel versions.
232 * Server known to build on the following kernels (others may also work):
233 2.6.32-696.18.7.el6 (RHEL6.9)
234 3.10.0-862.14.4.el7 (RHEL7.5)
235 3.10.0-957.el7 (RHEL7.6)
236 4.4.120-92.70 (SLES12 SP2)
237 4.4.162-94.69 (SLES12 SP3)
238 4.4.0-131 (Ubuntu 16.04)
239 4.15.0-32 (Ubuntu 18.04)
240 vanilla linux 4.6.7 (ZFS only)
241 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
242 * Client known to build on unpatched kernels (others may also work):
243 2.6.32-696.18.7.el6 (RHEL6.9)
244 2.6.32-754.6.3.el6 (RHEL6.10)
245 3.10.0-862.14.4.el7 (RHEL7.5)
246 3.10.0-957.el7 (RHEL7.6)
247 4.14.0-49.13.1.el7a (RHEL7.5)
248 4.14.0-115.2.2.el7a (RHEL7.6)
249 4.4.120-92.70 (SLES12 SP2)
250 4.4.162-94.69 (SLES12 SP3)
251 4.4.0-131 (Ubuntu 16.04)
252 4.15.0-32 (Ubuntu 18.04)
253 vanilla linux 4.15.0 (ZFS only)
254 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
255 * Recommended ZFS / SPL version: 0.7.9
256 * Tested with ZFS / SPL version: 0.7.9
257 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
258 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
259 stack overflow. For more information, please refer to bugzilla 17630.
260 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
261 the re-exporting nodes to be mounted with "32bitapi" mount option
262 * Note for Ubuntu 14.04.5:
263 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
264 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
266 --------------------------------------------------------------------------------
268 2018-04-03 Intel Corporation
270 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
271 for currently supported client and server kernel versions.
272 * Server known to build on patched kernels:
273 2.6.32-431.29.2.el6 (RHEL6.5)
274 2.6.32-504.30.3.el6 (RHEL6.6)
275 2.6.32-573.26.1.el6 (RHEL6.7)
276 2.6.32-642.15.1.el6 (RHEL6.8)
277 2.6.32-696.18.7.el6 (RHEL6.9)
278 3.10.0-693.21.1.el7 (RHEL7.4)
279 3.0.101-0.47.71 (SLES11 SP3)
280 3.0.101-107 (SLES11 SP4)
281 3.12.74-60.64.40 (SLES12 SP1)
282 4.4.114-92.64 (SLES12 SP2)
283 4.4.114-94.11 (SLES12 SP3)
284 3.13.0-101 (Ubuntu 14.04, ZFS only)
285 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
286 4.4.0-116 (Ubuntu 16.04)
287 vanilla linux 4.6.7 (ZFS only)
288 * Client known to build on unpatched kernels:
289 2.6.32-431.29.2.el6 (RHEL6.5)
290 2.6.32-504.30.3.el6 (RHEL6.6)
291 2.6.32-573.26.1.el6 (RHEL6.7)
292 2.6.32-642.15.1.el6 (RHEL6.8)
293 2.6.32-696.18.7.el6 (RHEL6.9)
294 3.10.0-693.21.1.el7 (RHEL7.4)
295 3.0.101-0.47.71 (SLES11 SP3)
296 3.0.101-107 (SLES11 SP4)
297 3.12.74-60.64.40 (SLES12 SP1)
298 4.4.114-92.64 (SLES12 SP2)
299 4.4.114-94.11 (SLES12 SP3)
300 3.13.0-101 (Ubuntu 14.04)
301 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
302 4.4.0-116 (Ubuntu 16.04)
304 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
305 * Recommended ZFS / SPL version: 0.7.6
306 * Tested with ZFS / SPL version: 0.7.6
307 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
308 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
309 stack overflow. For more information, please refer to bugzilla 17630.
310 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
311 the re-exporting nodes to be mounted with "32bitapi" mount option
312 * Note for Ubuntu 14.04.5:
313 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
314 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
316 --------------------------------------------------------------------------------
318 2017-07-05 Intel Corporation
320 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
321 for currently supported client and server kernel versions.
322 * Server known to build on patched kernels:
323 2.6.32-431.29.2.el6 (RHEL6.5)
324 2.6.32-504.30.3.el6 (RHEL6.6)
325 2.6.32-573.26.1.el6 (RHEL6.7)
326 2.6.32-642.15.1.el6 (RHEL6.8)
327 2.6.32-696.3.1.el6 (RHEL6.9)
328 3.10.0-514.21.1.el7 (RHEL7.3)
329 3.0.101-0.47.71 (SLES11 SP3)
330 3.0.101-100 (SLES11 SP4)
331 3.12.74-60.64.40 (SLES12 SP1)
332 4.4.59-92.17 (SLES12 SP2)
333 vanilla linux 4.6.7 (ZFS only)
334 * Client known to build on unpatched kernels:
335 2.6.32-431.29.2.el6 (RHEL6.5)
336 2.6.32-504.30.3.el6 (RHEL6.6)
337 2.6.32-573.26.1.el6 (RHEL6.7)
338 2.6.32-642.15.1.el6 (RHEL6.8)
339 2.6.32-696.3.1.el6 (RHEL6.9)
340 3.10.0-514.21.1.el7 (RHEL7.3)
341 3.0.101-0.47.71 (SLES11 SP3)
342 3.0.101-100 (SLES11 SP4)
343 3.12.74-60.64.40 (SLES12 SP1)
344 4.4.59-92.17 (SLES12 SP2)
346 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
347 * Recommended ZFS / SPL version: 0.6.5.9
348 * Tested with ZFS / SPL version: 0.6.5.9
349 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
350 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
351 stack overflow. For more information, please refer to bugzilla 17630.
352 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
353 the re-exporting nodes to be mounted with "32bitapi" mount option
355 --------------------------------------------------------------------------------
357 12-07-2016 Intel Corporation
359 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
360 for currently supported client and server kernel versions.
361 * Server known to build on patched kernels:
362 2.6.32-431.29.2.el6 (RHEL6.5)
363 2.6.32-504.30.3.el6 (RHEL6.6)
364 2.6.32-573.26.1.el6 (RHEL6.7)
365 2.6.32-642.11.1.el6 (RHEL6.8)
366 3.10.0-514.el7 (RHEL7.3)
367 3.0.101-0.47.71 (SLES11 SP3)
368 3.0.101-80 (SLES11 SP4)
369 3.12.62-60.64.8 (SLES12 SP1)
370 vanilla linux 4.5.7 (ZFS only)
371 * Client known to build on unpatched kernels:
372 2.6.32-431.29.2.el6 (RHEL6.5)
373 2.6.32-504.30.3.el6 (RHEL6.6)
374 2.6.32-573.26.1.el6 (RHEL6.7)
375 2.6.32-642.11.1.el6 (RHEL6.8)
376 3.10.0-514.el7 (RHEL7.3)
377 3.0.101-0.47.71 (SLES11 SP3)
378 3.0.101-80 (SLES11 SP4)
379 3.12.62-60.64.8 (SLES12 SP1)
381 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
382 * Recommended ZFS / SPL version: 0.6.5.8
383 * Tested with ZFS / SPL version: 0.6.5.8
384 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
385 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
386 stack overflow. For more information, please refer to bugzilla 17630.
387 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
388 the re-exporting nodes to be mounted with "32bitapi" mount option
390 --------------------------------------------------------------------------------
392 02-29-2016 Intel Corporation
394 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
395 for currently supported client and server kernel versions.
396 * Server known to build on patched kernels:
397 2.6.32-431.29.2.el6 (RHEL6.5)
398 2.6.32-504.30.3.el6 (RHEL6.6)
399 2.6.32-573.12.1.el6 (RHEL6.7)
400 3.10.0-327.3.1.el7 (RHEL7.2)
401 3.0.101-0.47.71 (SLES11 SP3)
402 3.0.101-68 (SLES11 SP4)
403 vanilla linux 4.2.1 (ZFS only)
404 * Client known to build on unpatched kernels:
405 2.6.32-431.29.2.el6 (RHEL6.5)
406 2.6.32-504.30.3.el6 (RHEL6.6)
407 2.6.32-573.12.1.el6 (RHEL6.7)
408 3.10.0-327.3.1.el7 (RHEL7.2)
409 3.0.101-0.47.71 (SLES11 SP3)
410 3.0.101-68 (SLES11 SP4)
413 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
414 * Recommended ZFS / SPL version: 0.6.4.2
415 * Tested with ZFS / SPL version: 0.6.4.2
416 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
417 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
418 stack overflow. For more information, please refer to bugzilla 17630.
419 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
420 the re-exporting nodes to be mounted with "32bitapi" mount option
422 --------------------------------------------------------------------------------
424 03-10-2015 Intel Corporation
426 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
427 for currently supported client and server kernel versions.
428 * Server known to build on patched kernels:
429 2.6.32-431.29.2.el6 (RHEL6.5)
430 2.6.32-504.8.1.el6 (RHEL6.6)
431 3.0.101-0.46 (SLES11 SP3)
432 * Client known to build on unpatched kernels:
433 2.6.32-431.29.2.el6 (RHEL6.5)
434 2.6.32-504.8.1.el6 (RHEL6.6)
435 3.10.0-123.20.1.el7 (RHEL7)
436 3.0.101-0.46 (SLES11 SP3)
437 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
438 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
439 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
440 stack overflow. For more information, please refer to bugzilla 17630.
441 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
442 the re-exporting nodes to be mounted with "32bitapi" mount option
444 Severity : enhancement
446 Description: control OST-index in IDIF via ROCOMPAT flag.
447 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
448 last_rcvd file. For new formatted OST device, it will be auto set;
449 for the case of upgrading from old OST device, you can enable it
450 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
451 enabled, for new created OST-object, its IDIF-in-LMA will contain
452 the OST-index; for the existing OST-object, the OSD will convert
453 old format IDIF as new format IDIF with OST-index stored in the
454 LMA EA when accessing such OST-object or via OI scrub. Once such
455 flag is enabled, it cannot be reverted back, so the system cannot
456 be downgraded to the orignal incompatible version.
458 --------------------------------------------------------------------------------
460 07-30-2014 Intel Corporation
462 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
463 for currently supported client and server kernel versions.
464 * Server known to build on patched kernels:
465 2.6.32-431.20.3.el6 (RHEL6.5)
466 3.0.101-0.31 (SLES11 SP3)
467 * Client known to build on unpatched kernels:
468 2.6.32-431.20.3.el6 (RHEL6.5)
469 3.0.101-0.31 (SLES11 SP3)
472 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
473 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
474 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
475 stack overflow. For more information, please refer to bugzilla 17630.
476 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
477 the re-exporting nodes to be mounted with "32bitapi" mount option
479 --------------------------------------------------------------------------------
481 10-31-2013 Intel Corporation
483 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
484 * Server builds for kernels:
485 2.6.32-358.18.1.el6 (RHEL6.4)
486 * Client builds for unpatched kernels:
487 2.6.32-358.18.1.el6 (RHEL6.4)
488 3.0.92-0.8 (SLES11 SP3)
490 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
491 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
492 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
493 stack overflow. For more information, please refer to bugzilla 17630.
494 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
495 the re-exporting nodes to be mounted with "32bitapi" mount option
497 --------------------------------------------------------------------------------
499 05-30-2013 Intel Corporation
501 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
502 * Server builds for kernels:
503 2.6.32-358.6.2.el6 (RHEL6.4)
504 * Client builds for unpatched kernels:
505 2.6.32-358.6.2.el6 (RHEL6.4)
506 3.0.74-0.6.6 (SLES11 SP2)
508 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
509 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
510 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
511 stack overflow. For more information, please refer to bugzilla 17630.
512 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
513 the re-exporting nodes to be mounted with "32bitapi" mount option
515 --------------------------------------------------------------------------------
517 9-30-2012 Whamcloud, Inc.
519 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
520 * Server support for kernels:
521 2.6.32-279.5.1.el6 (RHEL6)
522 * Client support for unpatched kernels:
523 2.6.32-279.5.1.el6 (RHEL6)
525 2.6.18-238.19.1.el5 (RHEL5)
526 2.6.16.60-0.69.1 (SLES10)
527 * Recommended e2fsprogs version: 1.41.90.wc4
528 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
529 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
530 stack overflow. For more information, please refer to bugzilla 17630.
531 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
532 the re-exporting nodes to be mounted with "32bitapi" mount option
534 Severity : enhancement
536 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
538 Severity : enhancement
540 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
542 Severity : enhancement
544 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
546 Severity : enhancement
548 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
550 Severity : enhancement
552 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
554 --------------------------------------------------------------------------------
556 04-30-2012 Whamcloud, Inc.
558 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
559 * Server support for kernels:
560 2.6.32-220.4.2.el6 (RHEL6)
561 * Client support for unpatched kernels:
562 2.6.32-220.4.2.el6 (RHEL6)
564 2.6.18-238.19.1.el5 (RHEL5)
565 2.6.16.60-0.69.1 (SLES10)
566 * Recommended e2fsprogs version: 1.41.90.wc4
567 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
568 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
569 stack overflow. For more information, please refer to bugzilla 17630.
570 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
571 the re-exporting nodes to be mounted with "32bitapi" mount option
573 Severity : enhancement
575 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
577 Severity : enhancement
579 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
581 Severity : enhancement
583 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
585 Severity : enhancement
587 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
589 --------------------------------------------------------------------------------
591 09-30-2011 Whamcloud, Inc.
593 * Server support for kernels:
594 2.6.18-238.19.1.el5 (RHEL5)
595 2.6.32-131.6.1.el6 (RHEL6)
596 * Client support for unpatched kernels:
597 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
598 2.6.18-238.19.1.el5 (RHEL5)
599 2.6.32-131.6.1.el6 (RHEL6)
600 2.6.29.4-167.fc11 (FC11)
601 2.6.16.60-0.69.1 (SLES10)
603 * Recommended e2fsprogs version: 1.41.90.wc3
604 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
605 Lustre filesystem with 4K stack may cause a stack overflow. For more
606 information, please refer to bugzilla 17630.
607 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
608 mounted with "32bitapi" mount option
610 Severity : enhancement
611 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
613 Severity : enhancement
614 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
616 Severity : enhancement
617 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
620 Frequency : only with SLES10
622 Description: Use OFED "KMP" provided by Novell
623 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
624 Lustre is now built against this package. That means you need to
625 install the ofed-kmp package from Novell for the patchless client.
626 Note that the ofed-kmp that Novell ships may not exactly match the
627 kernel version but should still be compatible.
629 Severity : enhancement
631 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
632 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
634 --------------------------------------------------------------------------------
636 2010-07-15 Oracle, Inc.
638 * Server support for kernels:
639 2.6.18-164.11.1.el5 (RHEL 5)
640 2.6.18-164.11.1.0.1.el5 (OEL 5)
641 * Client support for unpatched kernels:
642 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
643 2.6.18-164.11.1.el5 (RHEL 5),
644 2.6.18-164.11.1.0.1.el5 (OEL 5)
645 2.6.16.60-0.42.8 (SLES 10),
647 2.6.29.4-167.fc11 (FC11)
648 * Recommended e2fsprogs version: 1.41.10-sun2
650 Severity : enhancement
652 Description: Distribute lfs_migrate script with lustre
653 Details : The lfs_migrate mentioned in the manual is now included in the
654 lustre tree in lustre/scripts and will distributed in lustre RPMs.
656 Severity : enhancement
658 Description: Failover nid list added to import proc file.
660 Severity : enhancement
662 Description: "som_preview" mount option required to use SOM
663 Details : The clients will need to use som_preview mount option
667 Description: Lustre does not do 1MB IOs to HW RAID
668 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
670 Severity : enhancement
672 Description: lfs quota output is not very convenient for awk/sed-parsing
673 Details : Some positions in lfs quota output table could be empty or
674 non-empty which made it hard to parse it with scripts, now
675 a dash is put instead of space where there is not supposed
680 Description: Update to OFED 1.5.1
681 Details : OFED 1.5.1 will be used in environments that don't use
682 vendor supplied OFED.
686 Description: Remove lnet. prefix from lctl params display
687 Details : Also renamed lnet.memused to lnet_memused to
692 Description: Use conf_param -d to remove permanant settings.
696 Description: Add list_param -R
697 Details : lctl list_param now supports recursive mode.
700 Bugzilla : 21829, 22312
701 Description: Fix llobdstat.
702 Details : 1. Now updating the write_bytes counter
703 2. Port script fixes from 1.8.
707 Description: decrease the usage of memory on clients.
708 Details : 1. On clients, recycle dentries and inodes unused.
709 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
714 Description: llobdstat fix and enhancement
715 Details : add a counter to set a limit to how many samples will be returned
716 fix a wildcard in the path to limit to obdfilter stats only
720 Description: lfs check only allowed for root
721 Details : Code cleanups and added a regresion test of non-root use of
726 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
727 Details : Fix hr->hr_index overload.
731 Description: Fix FIEMAP functionality.
735 Description: FC11 patchless client support added
739 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
740 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
744 Description: New acc-sm script and various test-framework enhancments.
748 Description: Previously landed patch was reverted.
752 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
756 Description: Add ldiskfs maxdirsize mount option.
760 Description: Fix mkfs.lustre support for 16TB LUNs.
764 Description: Print mmp_check_interval
765 Details : print mmp_check_interval and make it possible to abort
766 mount operation in case it takes too long.
770 Description: Add cascading_rw to lustre mpi tests.
774 Description: Store SOM data in EA.
775 Details : Also several SOM related cleanups.
779 Description: Add refcount to md_open_data
783 Description: Use refcount on lov_request_set to prevent memory corruption.
787 Description: Statahead: use parent dir's i_mutex for synchronization.
791 Description: loadgen improvements.
795 Description: Enable peer health detection in ksocklnd and ko2iblnd.
799 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
803 Description: Fix race on obd_devcie:md_stats
807 Description: Fix SLES 11 client builds.
808 Details : Resolved uninitalized pointer gcc warnings.
812 Description: Allow non-root access for "lfs check".
813 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
817 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
819 Severity : enhancement
821 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
822 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
826 Description: Sanity test_27p fails.
827 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
828 Sanity tests 27 were cleaned up and improved.
832 Description: Unify req format on client/servers
833 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
834 buffers and swabbers approach..
839 Description: drop join file stuff.
841 Severity : enhancement
843 Description: Add 2.6.27 and SLES11 (patchless client) support
847 Description: Fix in ptlrpc_expire_one_request() to print the signed time
850 Severity : enhancement
852 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
855 Severity : enhancement
857 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
859 Severity : enhancement
861 Description: correctly handle big reply message.
862 Details : send LNet event if reply is bigger than buffer and adjust this buffer
867 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
869 Severity : enhancement
871 Description: Add support for OEL5.
875 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
877 Severity : enhancement
879 Description: "weak-modules" support
880 Details : Implement "weak-modules" support which enables kernel modules
881 to be used with any kernel that implements the same kABI. In
882 order to achieve this modules are now installed in
883 /lib/modules/$(uname -r)/updates/kernel on all distributions.
887 Description: client could not reconnect to OST because of an active request.
888 Details : The client now retries to reconnect to the same server, if a
889 connect request failed with EBUSY or -EAGAIN. Test case 35b
890 conf-sanity.sh has been created.
894 Description: Don't allow make backward step on assignin osc next id.
895 Details : race between allocation next id and ll_sync thread can be cause
896 of set wrong osc next id and can be kill valid ost objects.
900 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
904 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
908 Description: Error handling in osc_statfs_interpret() has been improved.
909 Details : Check in osc_statfs_interpret() for EBADR.
913 Description: Do not send statfs() requests to OSTs disabled by administrator.
914 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
916 Severity : enhancement
918 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
920 Severity : enhancement
922 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
924 Severity : enhancement
926 Description: Enhancement for lfs(1) command to use numeric uid/gid.
930 Description: Improved cleanup procedure for conf-sanity test_46a.
934 Description: Do not start bulk transfer, if client tried to reconnect.
935 Details : Do not start bulk transfer, if client tried to reconnect,
936 but failed due to active rpcs. Check for exp_abort_active_req
937 has been implemented.
941 Description: Prevent inconsistences between linux and lustre mount structures.
942 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
943 Make the sleep interruptible.
945 Severity : enhancement
947 Description: provide server to client comms path
948 Details : server to client communications path via new LDLM_SET_INFO rpc
952 Description: truncate starts GFP_FS allocation under transaction causing deadlock
953 Details : ldiskfs_truncate calls grab_cache_page which may start page
954 allocation under an open transaction. This may lead to
955 calling prune_icache with consequent lustre reentrance.
959 Description: send correctly lsm on open replay
960 Details : MDS is trust to LSM size on replay open, but client can set wrong size
965 Description: enable client interface failover
966 Details : When a child reconnects from another NID, properly update export
967 nid hash position and ldlm reverse import.
969 Severity : enhancement
971 Description: Communicate OST degraded/readonly state via statfs to MDS
972 Details : Flags in the statfs returned from OSTs indicate whether the
973 OST is in a degraded RAID state, or if the filesystem has
974 turned read-only after a filesystem error is detected.
979 Description: don't panic if EPROTO was hit when reading symlink
980 Details : correctly handling request reference in error cases.
982 Severity : enhancement
984 Description: Add LustreNetLink, a kernel-userspace communcation path.
986 Severity : enhancement
988 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
993 Description: access to llog context before init.
994 Details : move handling CATALOGS file at osc layer and forbid access to llog
998 Frequency : rare, only when client is writing to OST on local node
1000 Description: Avoid deadlock for local client writes
1001 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1002 memory freeing context. This allows OST threads to set the
1003 PF_MEMALLOC flag on task structures in order to allocate memory
1004 from reserved pools and complete IO.
1005 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1006 so that the OST threads generate memory pressure and allow
1007 inactive pages to be reclaimed.
1009 Severity : enhancement
1011 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1016 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1017 Details : move ldlm namespace creation in setup phase to avoid grab
1018 _lprocfs_lock with cli_sem held.
1022 Description: Temporarily disable grant shrink.
1023 Details : Disable the feature for debugging.
1026 Frequency : only when formatting test filesystems
1028 Description: Unable to run several mkfs.lustre on loop devices at the same time
1029 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1030 formatting. The solution is to proper handle the error.
1032 Severity : enhancement
1034 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1036 Severity : enhancement
1038 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1043 Description: resolve race between obd_disconnect and class_disconnect_exports
1044 Details : if obd_disconnect will be called to already disconnected export he
1045 forget release one reference and osc module can't unloaded.
1047 Severity : enhancement
1049 Description: Allow tuning service thread via /proc
1050 Details : For each service a new
1051 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1052 created that can be used to set min/max thread counts, and get the
1053 current number of running threads.
1058 Descriptoin: don't return error if have partially created objects for file.
1059 Details : lov_update_create_set uses set->set_success as index for created
1060 objects, so if some requests will be failed, they will have hole
1061 at end of array and qos_shrink_lsm for allocate correct lsm.
1063 Severity : enhancement
1065 Description: Update OFED support to 1.4.1
1070 Description: Reduce small size read RPC
1071 Details : Set read-ahead limit for every file and only do read-ahead when
1072 available read-ahead pages are bigger than 1M to avoid small size
1075 Severity : enhancement
1077 Description: move AT tunable parameters for more consistent usage
1078 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1080 Severity : enhancement
1082 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1083 Details : allow skip disconnected ost for send statfs request and hide error
1089 Description: fix racy locking of mballoc block bitmaps causing BUG
1090 Details : The locking of the mballoc buddy bitmap and the in-memory
1091 block bitmap was using two different spin locks in some
1092 cases. This made it possible to incorrectly access the
1093 mballoc bitmap while another process was modifying it,
1094 causing a sanity assertion to fail. While no on-disk corruption
1095 was reported, there was some risk of this happening.
1098 Frequency : rare, on llog test 6
1100 Descriptoin: don't allow connect to already connected import
1101 Details : allowing connect to already connected import is hide connecting problem.
1106 Descriptoin: allow kill process which wait statahead result
1107 Details : for some reasons 'ls' can stick in waiting result from statahead,
1108 in this case need way for kill this process.
1110 Severity : enhancement
1112 Description: Add state history info file, enhance import info file
1113 Details : Track import connection state changes in a new osc/mdc proc file;
1114 add overview-type data to the osc/mdc import proc file.
1116 Severity : enhancement
1118 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1120 Severity : enhancement
1122 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1127 Descriptoin: OST grant too much space to client even there are not enough space.
1128 Details : Client will shrink its grant cache to OST if there are no write
1129 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1130 this grant cache if there are already not enough avaible space
1131 (left_space < total_clients * 32M).
1134 Frequency : start MDS on uncleanly shutdowned MDS device
1136 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1137 Details : stay in waiting mds<>ost recovery finished produce random bugs
1138 due race between two ll_sync thread for one lov target. send
1139 ACTIVATE event only if connect realy finished and import have
1143 Frequency : rare, connect and disconnect target at same time
1145 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1146 Details : don't call obd_disconnect under lov_lock. this long time
1147 operation and can block ptlrpcd which answer to connect request.
1152 Descriptoin: don't lose wakeup for imp_recovery_waitq
1153 Details : recover_import_no_retry or invalidate_import and import_close can
1154 both sleep on imp_recovery_waitq, but we was send only one wakeup
1158 Frequency : always with long access acl
1160 Descriptoin: mds can't pack reply with long acl.
1161 Details : mds don't control size of acl but they limited by reint/getattr
1164 Severity : enhancement
1166 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1168 Severity : enhancement
1170 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1173 Frequency : start MDS on uncleanly shutdowned MDS device
1175 Descriptoin: aborting recovery hang on MDS
1176 Details : don't throttle destroy RPCs for the MDT.
1179 Frequency : on remount
1181 Description: external journal device not working after the remount
1182 Details : clear dev_rdonly flag for external journal devices in
1188 Description: shutdown vs evict race
1189 Details : client_disconnect_export vs connect request race.
1190 if client will evicted at this time - we start invalidate
1191 thread without referece to import and import can be freed
1195 Frequency : rare, need acl's on inode.
1197 Description: client can't handle ost additional correctly
1198 Details : if ost was added after client connected to mds client can have
1199 hit lnet_try_match_md ... to big messages to wide striped files.
1200 in this case need teach client to handle config events about add
1201 lov target and update client max ea size at that event.
1203 Severity : enhancement
1205 Description: Changelogs
1206 Details : Changelogs are a lightweight mechanism to track filesystem
1207 metadata and namespace changes. The changelog is recorded
1208 permanently on the MDTs, and is periodically "consumed" / purged
1209 when records are no longer needed.
1211 Severity : enhancement
1213 Description: compact fld format with extents
1214 Details : Store range of seq rather than every seq in FLD. Seq
1215 controller update FLD rather than clients. In Case of CMD, mdt0
1216 has FLD, all other metadata server act as non persistent proxy
1217 for FLD queries and cache fld entries in fld cache.
1222 Description: don't skip ost target if they assigned to file
1223 Details : Drop slow OSCs if we can, but not for requested start idx.
1224 This means "if OSC is slow and it is not the requested
1225 start OST, then it can be skipped, otherwise skip it only
1226 if it is inactive/recovering/out-of-space.
1230 Description: more cleanup in mds_lov
1231 Details : not send LOV EA under replay, we can't know about they size at this
1232 time. Don't allow client connect to mds before any ost connected,
1233 for avoid problems with LOV EA size and returning EIO to client.
1235 Severity : enhancement
1237 Description: Interoperability at server side (Disk interoperability)
1239 Severity : enhancement
1241 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1243 Severity : enhancement
1245 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1247 Severity : enhancement
1249 Description: New client IO stack (CLIO).
1251 Severity : enhancement
1253 Description: Commit on sharing. Eliminate inter-client dependencies between
1254 uncommitted transactions by doing transaction commits.
1255 Thereby clients may recovery independently.
1258 Frequency : Create a symlink file with a very long name
1260 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1261 Details : If there is no extra space in the request for early cancels,
1262 ldlm_req_handles_avail() returns 0 instead of a negative value.
1264 Severity : enhancement
1266 Description: Add /proc entry for import status
1267 Details : The mdc, osc, and mgc import directories now have
1268 an import directory that contains useful import data for debugging
1269 connection problems.
1271 Severity : enhancement
1273 Description: Re-disable certain /proc logging
1274 Details : Enable and disable client's offset_stats, extents_stats and
1275 extents_stats_per_process stats logging on the fly.
1278 Frequency : Only on FC kernels 2.6.22+
1280 Description: oops in statahead
1281 Details : Do not drop reference count for the dentry from VFS when lookup,
1282 VFS will do that by itself.
1284 Severity : enhancement
1286 Description: Generic /proc file permissions
1287 Details : Set /Proc file permissions in a more generic way to enable non-
1288 root users operate on some /proc files.
1292 Description: Hitting mdc_commit_close() ASSERTION
1293 Details : Properly handle request reference release in
1294 ll_release_openhandle().
1298 Description: quota recovery deadlock during mds failover
1299 Details : This patch includes att18982, att18236, att18237 in bz14840.
1301 1. fix osts hang when mds does failover with quotaon
1302 2. prevent watchdog storm when osts threads wait for the
1307 Frequency : only patchless client
1308 Description: add workaround for race between add/remove dentry from hash
1310 Severity : enhancement
1312 Description: Allow OST glimpses to return PW locks
1316 Description: LBUG when llog conf file is full
1317 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1321 Description: Prevent import from entering FULL state when server in recovery
1325 Description: service mount cannot take device name with ":"
1326 Details : Only when device name contains ":/" will mount treat it as
1332 Description: replace ptlrpcd with the statahead thread to interpret the async
1333 statahead RPC callback
1337 Frequency : on recovery
1338 Description: I/O failures after umount during fail back
1339 Details : if client reconnected to restarted server we need join to recovery
1340 instead of find server handler is changed and process self
1341 eviction with cancel all locks.
1343 Severity : enhancement
1345 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1347 Severity : enhancement
1349 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1351 Severity : enhancement
1353 Description: Upcall on Lustre log has been dumped
1354 Details : Allow for a user mode script to be called once a Lustre log has
1355 been dumped. It passes the filename of the dumped log to the
1356 script, the location of the script can be specified via
1357 /proc/sys/lnet/debug_log_upcall.
1362 Description: avoid idr_remove called for id which is not allocated.
1363 Details : Move assigment s_dev for clustered nfs to end of initialization,
1364 to avoid problem with error handling.
1369 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1370 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1371 into CONN_USED_HASH and this prodice warning when put connection
1372 again in unused hash.
1375 Severity : enhancement
1377 Description: File striping can now be set to use an arbitrary pool of OSTs.
1379 Severity : enhancement
1381 Description: Export bytes_read/bytes_write count on OSC/OST.
1385 Description: Early reply size mismatch, MGC loses connection
1386 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1387 the connect flags are properly negotiated.
1392 Description: quotas are not honored with O_DIRECT
1393 Details : all writes with the flag O_DIRECT will use grants which leads to
1394 this problem. Now using OBD_BRW_SYNC to guard this.
1398 Description: add quota statistics
1399 Details : 1. sort out quota proc entries and proc code.
1400 2. add quota statistics
1402 Severity : enhancement
1404 Description: enable quota support for HEAD.
1408 Description: Properly propagate oinfo flags from lov to osc for statfs
1409 Details : restore missing copy oi_flags to lov requests.
1411 Severity : enhancement
1413 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1414 : l_getgroups(8), lst(8), routerstat(8)
1415 Details : included man pages for llobdstat(8), llstat(8),
1416 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1418 Severity : enhancement
1420 Description: configure's --enable-quota should check the
1421 : kernel .config for CONFIG_QUOTA
1422 Details : configure is terminated if --enable-quota is passed but
1423 : no quota support is in kernel
1427 Description: Remove portals compatibility
1428 Details : Remove portals compatibility, not interoperable with releases
1433 Description: Resolve device initialization race
1434 Details : Prevent proc handler from accessing devices added to the
1435 obd_devs array but yet be intialized.
1437 Severity : enhancement
1439 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1441 Severity : enhancement
1443 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1448 Description: Using wrong pointer in osc_brw_prep_request
1449 Details : Access to array[-1] can produce panic if kernel compiled with
1450 CONFIG_PAGE_ALLOC enabled
1454 Description: Client runs out of low memory
1455 Details : Consider only lowmem when counting initial number of llap pages
1459 Description: *optional* service tags registration
1460 Details : if the "service tags" package is installed on a Lustre node
1461 When the filesystem is mounted, a local-node service tag will
1462 be created. See http://inventory.sun.com/ for more information
1463 about the Service Tags asset management system.
1467 Description: Kernel BUG tries to release flock
1468 Details : Lustre does not destroy flock lock before last reference goes
1469 away. So always drop flock locks when client is evicted and
1470 perform unlock regardless of successfulness of speaking to MDS.
1474 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1478 Description: sanity test 65a fails if stripecount of -1 is set
1479 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1484 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1485 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1488 Severity : enhancement
1490 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1494 Description: do not process already freed flock
1495 Details : flock can possibly be freed by another thread before it reaches
1496 to ldlm_flock_completion_ast.
1500 Description: LBUG during stress test
1501 Details : Need properly lock accesses the flock deadlock detection list.
1505 Description: oops in page fault handler
1506 Details : kernel page fault handler can return two special 'pages' in error
1507 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1511 Description: timeout with invalidate import.
1512 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1513 should be handled by ptlrpcd. This produce long age waiting and
1514 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1516 Severity : enhancement
1518 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1522 Description: enable MGS and MDT services start separately
1523 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1524 a co-located MGS without starting the MGS, which is a complement
1525 to 'nosvc' mount option.
1529 Frequency : after recovery
1530 Description: precreate to many object's after del orphan.
1531 Details : del orphan st in oscc last_id == next_id and this triger growing
1532 count of precreated objects. Set flag LOW to skip increase count
1533 of precreated objects.
1537 Frequency : rare, on clear nid stats
1538 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1539 Details : when clean nid stats sometimes try destroy live entry,
1540 and this produce panic in free.
1544 Description: Stack overflow during MDS log replay
1545 ease stack pressure by using a thread dealing llog_process.
1549 Description: wait until IO finished before start new when do lock cancel.
1550 Details : VM protocol want old IO finished before start new, in this case
1551 need wait until PG_writeback is cleared until check dirty flag and
1552 call writepages in lock cancel callback.
1554 Severity : enhancement
1556 Description: using special macro for print time and cleanup in includes.
1560 Description: mds_mfd_close() ASSERTION(rc == 0)
1561 Details : In mds_mfd_close(), we need protect inode's writecount change
1562 within its orphan write semaphore to prevent possible races.
1566 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1570 Frequency : rare, on shutdown ost
1571 Description: don't hit live lock with umount ost.
1572 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1573 use shrink_dcache_sb instead.
1577 Description: don't panic with use echo client
1578 Details : echo client pass NULL as client nid pointer and this produce null
1579 pointer dereference.
1583 Description: fix build on ppc32
1584 Details : compile code with -m64 flag produce wrong object file for ppc32.
1588 Description: add message levels for liblustreapi
1592 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1593 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1594 will result in an -EAGAIN and in the exit call path we call
1595 alloc_rr() which will with increasing aggressiveness attempt to
1596 aquire precreated objects on the minimum number of required OSCs.
1600 Description: Use old size assignment to avoid deadlock
1601 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1602 the scheduling while holding a spinlock. We do not need locking
1603 for size in ll_update_inode() because size is only updated from
1604 the MDS for directories or files without objects, so there is no
1605 other place to do the update, and concurrent access to such inodes
1606 are protected by the inode lock.
1610 Description: resolve "_IOWR redefined" build error on SLES10
1614 Description: dump the memory debugging after all modules are unloaded to
1615 suppress false negative in conf_sanity test 39
1617 Severity : enhancement
1619 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1624 Description: fix SLES kernel versioning
1625 Details : the kernel version for our SLES 10 kernel did not include a "-"
1626 before the "smp" at the end. while this was not a problem in
1627 general, it did mean that software trying to use the kernel
1628 version to try to detect a vendor specific kernel would fail.
1629 this was most evident by the OFED build scripts.
1633 Description: Don't update lov_desc members until making sure they are valid
1634 Details : When updating lov_desc members via proc fs, need fix their
1635 validities before doing the real update.
1639 Description: don't put request into delay list while invalidate in flight.
1640 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1641 import in flight. this produce timeout for invalidate and sometimes
1642 can cause stale data.
1646 Frequency : on ppc only
1647 Description: not convert ost objects for directory because it's not exist.
1648 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1652 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1654 Severity : enhancement
1656 Description: Update to support 2.6.22.14 vanilla kernel.
1660 Frequency : rare, on recovery
1661 Description: read procfs can produce deadlock in some situation
1662 Details : Holding lprocfs lock which send rpc can produce block for destroy
1663 obd objects and this also block reconnect with -EALREADY. This
1664 isn't fix all lprocfs bugs - but make it rare.
1666 Severity : enhancement
1668 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1671 Frequency : frequent on X2 node
1673 Description: mdc_set_open_replay_data LBUG
1674 Details : Set replay data for requests that are eligible for replay.
1678 Description: lustre_mgs: operation 101 on unconnected MGS
1679 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1680 MGC, and late on MGC cannot successfully connect to MGS and a lot
1681 of the error messages complaining that MGS is not connected.
1685 Frequency : on network error
1686 Description: panic with double free request if network error
1687 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1688 but it's true only for synchronus enqueue, for async enqueue
1689 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1692 Severity : enhancement
1694 Description: client-side metadata stat-ahead during readdir(directory readahead)
1695 Details : perform client-side metadata stat-ahead when the client detects
1696 readdir and sequential stat of dir entries therein
1699 Frequency : on start mds
1701 Description: Implement get_info(last_id) in obdfilter.
1704 Frequency : occasional
1706 Description: Correctly check stale fid, not start epoch if ost not support SOM
1707 Details : open with flag O_CREATE need set old fid in op_fid3 because
1708 op_fid2 was overwritten with new generated fid, but mds can answer
1709 with one of these two fids and both is not stale. Setattr
1710 incorrectly started an epoch and assume will be called
1711 done_writing, but without SOM done_writing ever being called.
1714 Frequency : rare, depends on device drivers and load
1716 Description: MDS or OSS nodes crash due to stack overflow
1717 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1718 In some cases, in conjunction with device drivers that use a lot
1719 of stack the MDS (or possibly OSS) service threads could overflow
1720 the stack. One change which was identified to consume additional
1721 stack has been reworked to avoid the extra stack usage.
1724 Frequency : occasional
1726 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1727 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1728 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1729 oscc_has_objects and see OSC has no objests and call
1730 oscc_internal_create to resend create request.
1732 Severity : enhancement
1734 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1736 Severity : enhancement
1738 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1741 Frequency : very rare
1743 Description: Fix replay if there is an un-replied request and open
1744 Details : In some cases, older replay request will revert the
1745 mcd->mcd_last_xid on MDS which is used to record the client's
1746 latest sent request.
1748 Severity : enhancement
1750 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1752 Severity : enhancement
1754 Description: Add rhel5 support to HEAD.
1756 Serverity : enhancement
1758 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1763 Description: Don't allow skipping OSTs if index has been specified.
1764 Details : Don't allow skipping OSTs if index has been specified, make
1765 locking in internal create lots better.
1769 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1770 Details : access to bitfield in structure is always rounded to long
1771 and this produce problem with not atomic change any bit.
1775 Description: Lustre make rpms failed.
1776 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1777 builds Lustre rpms from tarball.
1780 Frequency : rare on shutdown ost
1782 Description: If llog cancel was not send before clean_exports phase, this can
1783 produce deadlock in llog code.
1784 Details : If llog thread has last reference to obd and call class_import_put
1785 this produce deadlock because llog_cleanup_commit_master wait when
1786 last llog_commit_thread exited, but this never success because was
1787 called from llog_commit_thread.
1791 Description: allow userland application know is lost one of stripes.
1792 Details : fill lvb_blocks with error code on ost and return it to
1793 application if error flag found.
1797 Description: NULL lov_tgts causing MDS oops
1798 Details : more safe checks for NULL lov_tgts for avoid oops.
1800 Severity : enhancement
1802 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1806 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1807 Details : If an OST is down the MDS will hang indefinitely in
1808 obd_statfs_rqset() waiting for the statfs data. While for
1809 MDS QOS usage of statfs, it should not stuck in waiting.
1811 Severity : enhancement
1813 Description: remote_acl support
1814 Details : Support ACL-based permission check for remote user.
1815 Support setfacl/getfacl for remote user with the utils
1816 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1817 the system "{s,g}etfacl" utils.
1819 Severity : enhancement
1821 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1823 Severity : enhancement
1825 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1829 Description: make mgs_setparam() handle fsname containing dash
1830 Details : fsname containing a dash does not work with lctl conf_param
1832 Severity : enhancement
1834 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1836 Severity : enhancement
1838 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1842 Frequency : rare, at shutdown
1843 Description: access already free / zero obd_namespace.
1844 Details : if client_disconnect_export was called without force flag set,
1845 and exist connect request in flight, this can produce access to
1846 NULL pointer (or already free pointer) when connect_interpret
1847 store ocd flags in obd_namespace.
1851 Frequency : only at startup
1852 Description: not alloc memory with spinlock held.
1853 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1854 if any spinlock held.
1856 Severity : enhancement
1858 Description: make lustre randomly fail allocating memory
1859 Details : Make lustre randomly failed allocating memory for testing purpose.
1861 Severity : enhancement
1863 Description: lost problems with lov objid file
1864 Details : Fixes some scability and access to not inited memory problems
1865 in work with lov objdid file.
1870 Description: lfs find does not continue on file error
1871 Details : Continue other files processing when a file/dir is absent.
1875 Description: Inconsistent usage of lustre_pack_reply()
1876 Details : Standardize the usage of lustre_pack_reply() such that it
1877 always generate a CERROR on failure.
1880 Frequency : occasional
1882 Description: MDS hang or stay in waiting lock
1883 Details : If client receive lock with CBPENDING flag ldlm need send lock
1884 cancel as separate rpc, to avoid situation when cancel request
1885 can't processed due all i/o threads stay in wait lock.
1889 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1893 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1897 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1901 Description: Update to RHEL4 latest kernel.
1903 Severity : enhancement
1905 Description: Build SLES10 patchless client fails
1906 Details : The configure was broken by run ./configure with
1907 --with-linux-obj=.... argument for patchless client. When the
1908 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1909 can't search header adequately. Use absolute path such as
1910 -I($LINUX)/include instead.
1914 Description: interrupt oig_wait produce painc on resend.
1915 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1916 private set, and this produce situation when rq_ptlrpcd_data not
1917 copyed to new allocated request and triggered LBUG on assert
1918 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1919 ptlrpcd set if request is changed and can be safety changed to use
1924 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1925 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1926 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1927 these macros and replace them with their difinitions instead.
1931 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1932 Details : Update which_patch & target file for SLES10 latest kernel.
1935 Description: Remove client patches from SLES 10 kernel.
1936 Details : This causes SLES 10 clients to behave as patchless clients
1937 even on a Lustre-patched (server) kernel.
1939 Severity : enhancement
1941 Description: self-adjustable client's lru lists
1942 Details : use adaptive algorithm for managing client cached locks lru
1943 lists according to current server load, other client's work
1944 pattern, memory activities, etc. Both, server and client
1945 side namespaces provide number of proc tunables for controlling
1948 Severity : enhancement
1950 Description: light-weight GSS support
1951 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1952 and reduce performance overhead.
1954 Severity : enhancement
1956 Description: Linux keyring support
1957 Details : Support using service of Linux keyring for Lustre GSS internal
1958 context refresh/cache mechanism.
1962 Description: Fix errors in lfs documentation
1963 Details : Fixes man pages
1967 Description: don't use GFP_* in generic Lustre code.
1968 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1972 Description: obdclass is limited by single OBD_ALLOC(idarray)
1973 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1977 Description: warnings with build patchless client with vanila 2.6.19 and up
1978 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1979 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1983 Frequency : Only for SLES
1984 Description: sanity_quota fail test_1
1985 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1986 file, which makes TSTID[2] inunique.
1990 Frequency : Only for SLES9
1991 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1992 Details : Update target/ChangeLog/which_patch .
1996 Frequency : Only for test_10 in sanity-quota.sh
1997 Description: a bug in quota test code
1998 Details : checking a test flag in a wrong place
2002 Frequency : Only for test_13 in sanity-quota.sh
2003 Description: a bug in quota test script
2004 Details : shouldn't check overfull allocated quota
2008 Description: jbd statistics
2009 Details : Port older jbd statistics patch for sles10
2011 Severity : enhancement
2013 Description: Add printing inode info into message about error in writepage.
2018 Description: reply_lock_interpret crash due to race with it and lock cancel.
2019 Details : Do not replay locks that are being cancelled. Do not reference
2020 locks by their address during replay, just by their handle.
2024 Frequency : When flocks are used.
2025 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2026 Details : Flock locks might destroy just granted lock if it could be merged
2027 with another existing flock, this is done in completion handler,
2028 so teach ldlm_cli_enquque_fini that this is a valid case for
2034 Description: Oops in read and write path when failing to allocate lock.
2035 Details : Check if lock allocation failed and return error back.
2039 Description: lstripe command fails for valid OST index
2040 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2041 instead of lov->desc.ld_active_tgt_count.
2045 Description: sanity.sh failed test 103
2046 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2051 Description: df doesn't work properly if diskfs blocksize != 4K
2052 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2056 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2057 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2059 Severity : enhancement
2061 Description: Adaptive timeouts
2062 Details : RPC timeouts adapt to changing server load and network
2063 conditions to reduce resend attempts and improve recovery time.
2067 Description: llapi_file_create() does not allow some changes
2068 Details : add llapi_file_open() that allows specifying the mode and
2069 open flags, and also returns an open file handle.
2073 Description: merge and cleanup kernel patches.
2074 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2078 Description: Add journal checksum support.(Kernel part)
2079 Details : The journal checksum feature adds two new flags i.e
2080 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2081 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2082 indicates that the commit block contains the checksum for
2083 the blocks described by the descriptor blocks. Now commit
2084 record can be sent to disk without waiting for descriptor
2085 blocks to be written to disk. This behavior is controlled
2086 using JBD2_FEATURE_ASYNC_COMMIT flag.
2090 Description: OSS needs mutliple precreate threads
2091 Details : Add ability to start more than one create thread per OSS.
2095 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2096 Details : execute lfs setstripe on client
2100 Description: mds_obd_create error creating tmp object
2101 Details : When the user sets quota on root, llog will be affected and can't
2102 create files and write files.
2105 Frequency : Always on ia64 patchless client, and possibly others.
2107 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2108 Details : This allows the patchless client to be loaded on architectures
2109 without this export.
2113 Description: RedHat Update kernel for RHEL5
2114 Details : Add the kernel config file for RHEL5.
2118 Description: RedHat Update kernel for RHEL5
2119 Details : Modify the kernel config file more closer RHEL5.
2123 Description: Build failure against Centos5 (RHEL5)
2124 Details : Define PAGE_SIZE when it isn't present.
2128 Description: test_7 on sanity-quota.sh can't be used on separate machines
2129 Details : add facet to handle it
2133 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2134 Details : with NFS, the anon dentry's parent was set to itself in
2135 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2136 corresponding dentry other than use rec->ur_name.
2138 Severity : enhancement
2140 Description: lfs setstripe enhancement
2141 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2145 Description: enable data checksumming by default
2146 Details : enable checksum by default, allow --disable-checksum
2147 configure option and "-o nochecksum" mount option. Checksums
2148 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2152 Description: lustre support for RHEL5
2153 Details : Add support for RHEL5.
2157 Description: Client eviction due to failover config
2158 Details : after a connection loss, the lustre client should attempt to
2159 reconnect to the last active server first before trying the
2160 other potential connections.
2164 Description: when mds and osts use different quota unit(32bit and 64bit),
2165 quota will be released repeatly.
2166 Details : void sending multiple quota reqs to mds, which will keep the
2167 status between the reqs.
2171 Description: osts not allocated evenly to files
2172 Details : change the condition to increase offset_idx
2175 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2177 Description: Data corruption for OSTs that are formatted larger than 2TB
2179 Details : When generating the bio request for lustre file writes the
2180 sector number would overflow a temporary variable before being
2181 used for the IO. The data reads correctly from Lustre (which
2182 will overflow in a similar manner) but other file data or
2183 filesystem metadata may be corrupted in some cases.
2187 Description: Tune the kernel for good SCSI performance.
2188 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2189 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2194 Description: rewrite ext2-derived code in obdclass/uuid.c
2195 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2196 from scratch preserving functionality.
2200 Description: "lfs find -obd UUID" prints directories
2201 Details : "lfs find -obd UUID" will return all directory names instead
2202 of just file names. It is incorrect because the directories
2203 do not reside on the OSTs.
2207 Description: use i_size_read and i_size_write in 2.6 port
2208 Details : replace inode->i_size access with i_size_read/write()
2211 Frequency : when removing large files
2213 Description: scheduling issue during removal of large Lustre files
2214 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2215 It causes scheduling issues when removing large files (17TB in the
2219 Frequency : only with liblustre clients on XT3
2221 Description: evictions taking too long
2222 Details : allow llrd to evict clients directly on OSTs
2225 Frequency : only on ppc
2227 Description: /proc/fs/lustre/devices broken on ppc
2228 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2229 We should make sure the type of 'index' is loff_t and then cast
2230 explicitly as needed below. Do not assign an explicitly cast
2234 Frequency : only for rhel5
2236 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2237 Details : Modify the target file & which_kernel.
2239 Severity : enhancement
2241 Description: omit set fsid for export NFS
2242 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2243 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2247 Frequency : Always, for kernels after 2.6.16
2248 Description: Fix warning idr_remove called for id=.. which is not allocated.
2249 Details : Last kernels save old s_dev before kill super and not allow
2250 to restore from callback - restore it before call kill_anon_super.
2254 Description: buffer overruns could theoretically occur
2255 Details : llapi_semantic_traverse() modifies the "path" argument by
2256 appending values to the end of the origin string, and a
2257 overrun may occur. Adding buffer overrun check in liblustreapi.
2261 Description: Fix error on 'ls .' at the top of the Lustre mount.
2262 Details : Don't revalidate dentry if it is a root dentry.
2266 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2267 Details : Modify vm-tunables-rhel4.patch.
2271 Description: Kernel config for 2.6.18-vanilla.
2272 Details : Modify targets/2.6-vanilla.target.in.
2273 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2274 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2275 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2276 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2280 Description: improve handling recoverable errors
2281 Details : if request processig with error which can be recoverable on server
2282 request should be resend, otherwise page released from cache and
2287 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2288 Details : Modify target file & which_patch.
2289 A flaw was found in the IA32 system call emulation provided
2290 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2291 value could be stored in the %RAX register, which could trigger an
2292 out-of-bounds system call table access. An untrusted local user
2293 could exploit this flaw to run code in the kernel
2294 (ie a root privilege escalation). (CVE-2007-4573).
2298 Description: O_DIRECT bypasses client statistics.
2299 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2300 read_ahead_stats were updated. copied the stats section from
2301 osc_send_oap_rpc() into async_internal().
2305 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2309 Description: change order of libsysio includes
2310 Details : '#include sysio.h' should always come before '#include xtio.h'
2314 Description: handle "serious error: objid * already exists" more gracefully
2315 Details : If LAST_ID value on disk is smaller than the objects existing in
2316 the O/0/d* directories, it indicates disk corruption and causes an
2317 LBUG(). If the object is 0-length, then we should use the existing
2318 object. This will help to avoid a full fsck in most cases.
2320 Severity : enhancement
2322 Description: adapt the lustre_config script to support the upgrade case
2323 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2324 server targets to 1.6 in parallel.
2328 Description: To avoid grant space > avaible space when the disk is almost
2329 full. Without this patch you might see the error "grant XXXX >
2330 available" or some LBUG about grant, when the disk is almost
2332 Details : In filter_check_grant, for non_grant cache write, we should
2333 check the left space by if (*left > ungranted + bytes), instead
2334 of (*left > ungranted), because only we are sure the left space
2335 is enough for another "bytes", then the ungrant space should be
2336 increase. In client, we should update cl_avail_grant only there
2337 is OBD_MD_FLGRANT in the reply.
2341 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2342 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2343 described in CVE-2007-4573. This problem could allow untrusted
2344 local users to gain root access.
2347 Frequency : when using O_DIRECT and quotas
2349 Description: Incorrect file ownership on O_DIRECT output files
2350 Details : block usage reported by 'lfs quota' does not take into account
2351 files that have been written with O_DIRECT.
2356 Description: touch file failed when fs is not full
2357 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2358 otherwise we can get ENOSP while fs is not full.
2362 Description: parallel lock callbacks
2363 Details : Instead of sending blocking and completion callbacks as separated
2364 requests, adding them to a set and sending in parallel.
2367 Frequency : only for Cray XT3
2368 Bugzilla : 12829/13455
2369 Description: Changing primary group doesn't change the group lustre assigns to
2371 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2372 overridden with the primary group provided by the group upcall,
2373 whereas the supplied fsgid can be trusted if it is in the list of
2374 supplementary groups returned by the group upcall.
2376 Severity : enhancement
2378 Description: Allow masking D_WARNING, D_ERROR messages from console
2379 Details : Console messages can now be disabled via lnet.printk.
2383 Description: User code with malformed file open parameter crashes client node
2384 Details : Before packing join_file req, all the related reference should be
2385 checked carefully in case some malformed flags cause fake
2386 join_file req on client.
2390 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2391 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2392 mode is equal to the granted mode. However, it should also check
2393 that LDLM_FL_CP_REQD is not set.
2397 Description: Heavy nfs access might result in deadlocks
2398 Details : After ELC code landed, it is now improper to enqueue any mds
2399 locks under och_sem, because enqueue might want to decide to
2400 cancel open locks for same inode we are holding och_sem for.
2404 Description: Client eviction while running blogbench
2405 Details : A lot of unlink operations with concurrent I/O can lead to a
2406 deadlock causing evictions. To address the problem, the number of
2407 oustanding OST_DESTROY requests is now throttled to
2408 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2409 callbacks are processed in priority.
2413 Description: enable ACLs on MDS by default
2414 Details : ACLs must be enabled on MDS by default.
2417 Frequency : PPC/PPC64 only
2419 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2420 Details : fix duplicated definitions between asm-ppc64/types.h and
2421 lustre_types.h on PPC.
2424 Frequency : PPC/PPC64 only
2426 Description: asm-ppc/segment.h does not exist
2427 Details : fix compile issue on PPC.
2431 Description: better handle error messages in extents code
2434 Frequency : RHEL4 only
2436 Description: mkfs is very slow on IA64/RHEL4
2437 Details : A performance regression has been discovered in the MPT Fusion
2438 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2439 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2440 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2442 Severity : enhancement
2444 Description: SNMP support enhancement
2445 Details : Adding total number of sampled request for an MDS node in snmp
2448 Severity : enhancement
2450 Description: Optimize ldlm waiting list processing for PR extent locks
2451 Details : When processing waiting list for read extent lock and meeting read
2452 lock that is same or wider to it that is not contended, skip
2453 processing rest of the list and immediatelly return current
2454 status of conflictness, since we are guaranteed there are no
2455 conflicting locks in the rest of the list.
2459 Description: Time out and refuse to reconnect
2460 Details : When the failover node is the primary node, it is possible
2461 to have two identical connections in imp_conn_list. We must
2462 compare not conn's pointers but NIDs, otherwise we can defeat
2463 connection throttling.
2467 Description: port llog fixes from b1_6 into HEAD
2468 Details : Port llog reference couting and some llog cleanups from b1_6
2469 (bug 10800) into HEAD, for protect from panic and access to already
2470 free llog structures.
2474 Description: Detect stride IO mode in read-ahead
2475 Details : When a client does stride read, read-ahead should detect that and
2476 read-ahead pages according to the detected stride pattern.
2480 Description: data checksumming impacts single node performance
2481 Details : add support for several checksum algorithm. Currently, only CRC32
2482 and Adler-32 are supported. The checksum type can be changed on
2483 the fly via /proc/fs/lustre/osc/*/checksum_type.
2487 Description: use adler32 for page checksums
2488 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2493 Description: build for x2 fails
2494 Details : fix compile issue on Cray systems.
2498 Description: Properly match for duplicate locks
2499 Details : Due to different lock order from skiplists code, we need to
2500 traverse entire list for now
2503 Frequency : only on PPC/SLES10
2505 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2506 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2507 be defined. Add a hack in mkfs_lustre.c to work around this compile
2512 Description: LASSERT on MDS when client holding flock lock dies
2513 Details : ldlm pool logic depends on number of granted locks equal to
2514 number of released locks which is not true for flock locks, so
2515 just exclude such locks from consideration.
2519 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2520 Details : Use fsfilt_sync() for both the whole filesystem sync and
2521 individual file sync to eliminate dangerous inode locking
2522 with I_LOCK that can lead to a deadlock.
2526 Description: performance in 1.6.3
2527 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2528 sure that 1MB requests are not fragmented by the block layer.
2530 Severity : enhancement
2532 Description: organize the server-side client stats on per-nid basis
2533 Details : Change the structure of stats under obdfilter and mds to
2541 The "uuid"s file would list the uuids of _active_ exports.
2542 And the clear entry is to clear all stats and stale nids.
2544 Severity : enhancement
2546 Description: eliminate client locks in face of contention
2547 Details : file contention detection and lockless i/o implementation
2548 for contended files.
2552 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2555 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2557 Description: MDS may lose file striping (and hence file data) in some cases
2558 Details : If there are additional extended attributes stored on the MDS,
2559 in particular ACLs, SELinux, or user attributes (if user_xattr
2560 is specified for the client mount options) then there is a risk
2561 of attribute loss. Additionally, the Lustre file striping
2562 needs to be larger than default (e.g. striped over all OSTs),
2563 and an additional attribute must be stored initially in the
2564 inode and then increase in size enough to be moved to the
2565 external attribute block (e.g. ACL growing in size) for file
2570 Description: skiplist implementation simplification
2571 Details : skiplists are used to group compatible locks on granted list
2572 that was implemented as tracking first and last lock of each
2573 lock group the patch changes that to using doubly linked lists
2577 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2578 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2579 because LASSERT on both the data supplied by a client, and the
2580 data on disk is dangerous and incorrect.
2582 Severity : enhancement
2584 Description: Slow truncate/writes to huge files at high offsets.
2585 Details : Directly associate cached pages to lock that protect those pages,
2586 this allows us to quickly find what pages to write and remove
2587 once lock callback is received.
2591 Description: more ldlm soft lockups
2592 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2593 starve other threads from the resource lock for a long time in
2594 case of long waiting queue, so change the debug level from
2595 D_OTHER to the less frequently used D_INFO.
2597 Severity : enhancement
2599 Description: add -gid, -group, -uid, -user options to lfs find
2603 Description: Hung threads in invalidate_inode_pages2_range
2604 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2605 once one is completed. As a result, some RPCs are stuck in the
2606 queue and are never sent.
2610 Description: filter threads hungs on waiting journal commit
2611 Details : Cleanup filter group llog code, then only filter group llog will
2612 be only created in the MDS/OST syncing process.
2616 Description: Procfs and llog threads access destoryed import sometimes.
2617 Details : Sync the import destoryed process with procfs and llog threads by
2618 the import refcount and semaphore.
2620 Severity : enhancement
2622 Description: openlock cache of b1_6 port to HEAD
2627 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2628 Details : If insertion of an extent fails, then discard the inode
2629 preallocation and free data blocks else it can lead to duplicate
2634 Description: don't always update ctime in ext3_xattr_set_handle()
2635 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2636 In some cases the ctime should not be updated, for example for
2637 2.0->1.8 compatibility it is necessary to delete an xattr and it
2638 should not update the ctime.
2642 Bugzilla : 15713/16362
2643 Description: Assertion in iopen_connect_dentry in 1.6.3
2644 Details : looking up an inode via iopen with the wrong generation number can
2645 populate the dcache with a disconneced dentry while the inode
2646 number is in the process of being reallocated. This causes an
2647 assertion failure in iopen since the inode's dentry list contains
2648 both a connected and disconnected dentry.
2652 Description: assertion failure in ldlm_handle2lock()
2653 Details : fix a race between class_handle_unhash() and class_handle2object()
2654 introduced in lustre 1.6.5 by bug 13622.
2659 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2660 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2665 Description: oops in sd_iostats_seq_show()
2666 Details : unloading/reloading the scsi low level driver triggers a kernel
2667 bug when trying to access the sd iostat file.
2672 Description: Kernel panics during QLogic driver reload
2673 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2674 patch, causing memory corruption.
2679 Description: journal_dev option does not work in b1_6
2680 Details : pass mount option during pre-mount.
2682 Severity : enhancement
2684 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2685 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2686 extent information of a file. It can be used to map logical blocks
2687 in a file to physical blocks in the block device.
2691 Description: LDLM soft lockups - improvement
2692 Details : It is be possible to send the lock handle along with each read
2693 or write request because the client is already doing a lock match
2694 itself so there isn't any reason the OST should have to re-do that
2698 Frequency : only on Cray X2
2700 Description: X2 build failures
2701 Details : fix build failures on Cray X2.
2705 Description: xid & resent requests
2706 Details : Initialize RPC XID from clock at startup (randomly if clock is
2709 Severity : enhancement
2711 Description: Add lustre_start utility to start or stop multiple Lustre servers
2716 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2717 Details : In case of memory pressure, list_del() can be called twice on
2718 req->rq_history_list, causing a kernel oops.
2722 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2723 Details : fix stack overflow in the distributed lock manager by defering
2724 export eviction after a failed AST to the elt thread instead of
2725 handling it in the dlm interpret routine.
2729 Description: Convert some comments to new format.
2730 Details : Update documenting comments to match doxygen conventions.
2734 Description: Grammar fixes.
2735 Details : A couple of trivial spelling fixes.
2739 Description: OSD_COUNTERS-mandatory
2740 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2741 to simplify the code.
2745 Description: simplify cmm_device freeing logic.
2746 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2750 Description: Add lockdep support to dt_object_operations locking interface.
2751 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2752 parameter indicating lock ordering. Update mdd code to use new
2757 Description: Introduce failloc constants for lockless IO tests.
2758 Details : Add two new failloc constants to test lockless IO. Only one of
2759 them in implemented---another is checked in yet to be landed
2764 Description: Add lockdep support for inode mutex.
2765 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2766 in the situations where Lustre has to lock more than one inode
2771 Description: Add optional invariants checking support.
2772 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2773 configure switch. This macro is to be used for consistency and
2774 sanity checks that are too expensive to be left in `production'
2779 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2780 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2785 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2786 Details : Use spin_lock_nested() in (the only) situation where more than
2787 one ldlm_lock is locked simultaneously. Also, fix possible
2788 dead-lock in ldlm_lock_change_resource() by enforcing particular
2793 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2794 Details : Instead of specifying each ldlm_lock call-back through separate
2795 parameter, wrap them into struct ldlm_callback_suite.
2799 Description: Kill join_lru obd method and its callers.
2800 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2801 in memory---a requirement implemented through ->o_join_lru() obd
2802 method in HEAD. Remove this method and its users.
2806 Description: Add asynchronous ldlm ENQUEUE completion handler.
2807 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2808 case that stock ldlm_completion_ast() cannot handle as it waits
2809 until lock is granted. Introduce new ldlm_completion_ast_async()
2810 for this. Also comment ldlm_completion_ast().
2814 Description: ldlm_error <-> errno conversion.
2815 Details : Add functions to map (rather arbitrary) between LDLM error codes
2816 and standard errno values. CLIO needs this to prevent LDLM specific
2817 constants from escaping ldlm and osc.
2821 Description: Kill unused ldlm_handle2lock_ns() function.
2822 Details : Kill unused ldlm_handle2lock_ns() function.
2826 Description: Add lu_ref support to ldlm_lock
2827 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2828 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2829 and ldlm_resource. LDLM interface has to be changed, because code that
2830 releases a reference on a lock, has to "know" what reference this is.
2831 In the most frequent case
2833 lock = ldlm_handle2lock(handle);
2835 LDLM_LOCK_PUT(lock);
2837 no changes are required. When any other reference (received _not_ from
2838 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2839 instead of LDLM_LOCK_PUT().
2841 Arguably, changes are pervasive, and interface requires some discipline
2842 for proper use. On the other hand, it was very instrumental in finding
2843 a few leaked lock references.
2847 Description: Add ldlm_lock_addref_try().
2848 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2849 attempts to addref a lock that might be being canceled
2854 Description: Add ldlm_weigh_callback().
2855 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2856 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2857 hard-coded `number of pages' logic.
2861 Description: Add lockdep annotations to llog code.
2862 Details : Use appropriately tagged _nested() locking calls in the places
2863 where llog takes more than one ->lgh_lock lock.
2867 Description: Add loi_kms_set().
2868 Details : Wrap kms updates into a helper function.
2872 Description: Constify instances of struct lsm_operations.
2873 Details : Constify instances of struct lsm_operations.
2877 Description: lu_conf support.
2878 Details : On a server, a file system object is uniquely identified
2879 by a fid, which is sufficient to locate and load all object
2880 state (inode). On a client, on the other hand, more data are
2881 necessary instantiate an object. Change lu_object_find() and
2882 friends to take additional `lu_conf' argument describing object.
2883 Typically this includes layout information.
2887 Description: lu_context fixes.
2888 Details : Introduce new lu_context functions that are needed on the client
2889 side, where some system threads (ptlrpcd) are shared by multiple
2890 modules, and so cannot be stopped during module shutdown.
2894 Description: Add start and stop methods to lu_device_type_operations.
2895 Details : Introduce two new methods in lu_device_type_operations, that are
2896 invoked when first instance of a given type is created and last one
2897 is destroyed respectively. This is need by CLIO.
2901 Description: Add lu_ref support to struct lu_device.
2902 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2903 track leaked references.
2907 Description: Introduce lu_kmem_descr.
2908 Details : lu_kmem_descr and its companion interface allow to create
2909 and destroy a number of kmem caches at once.
2913 Description: Fix lu_object finalization race.
2914 Details : Fix a race between lu_object_find() finding an object and its
2915 concurrent finalization. This race is (most likely) not possible
2916 on the server, but might happen on the client.
2920 Description: Introduce lu_ref interface.
2921 Details : lu_ref is a debugging module allowing to track references to
2922 a given object. It is quite cpu expensive, and has to be
2923 explicitly enabled with --enable-lu_ref. See usage description
2928 Description: Factor lu_site procfs stats into a separate function.
2929 Details : Separate lu_site stats printing code into a separate function
2930 that can be reused on a client.
2934 Description: Constify instances of struct {lu,dt,md}_device_operations.
2935 Details : Constify instances of struct {lu,dt,md}_device_operations.
2939 Description: Introduce struct md_site and move meta-data specific parts of
2940 struct lu_site here.
2941 Details : Move md-specific fields out of struct lu_site into special struct
2942 md_site, so that lu_site can be used on a client.
2946 Description: Kill mdd_lov_destroy().
2947 Details : Remove unused mdd code.
2951 Description: Add st_block checking to multistat.c.
2952 Details : Add st_block checking to multistat.c.
2956 Description: Add lu_ref support to struct obd_device.
2957 Details : Add lu_ref tracking to obd_device.
2961 Description: Kill obd_set_fail_loc().
2962 Details : Remove unused code.
2966 Description: Add special type for ptlrpc_request interpret functions.
2967 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2968 there. Actual usage will be in CLIO.
2972 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2973 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2974 doesn't work with lockdep.
2978 Description: Add rwv.c test program.
2979 Details : New testing program exercising readv(2) and writev(2) (Qian).
2983 Description: Add sendfile.c test program.
2984 Details : New testing program exercising sendfile(2) (Jay).
2988 Description: Ratelimit a message that can be very frequent.
2989 Details : Ratelimit a memory allocation failure message that can
2994 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2995 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2999 Description: Liblustre build fixes.
3000 Details : Liblustre build fixes.
3004 Description: libcfs: add cfs_{need,cond}_resched() interface.
3005 Details : libcfs: add cfs_{need,cond}_resched() definition and
3006 implementations for Linux, NT, and liblustre.
3008 Severity : enhancement
3010 Description: More exported tunables for mballoc
3011 Details : Add support for tunable preallocation window and new tunables for
3012 large/small requests
3016 Description: Detect corruption of block bitmap and checking for preallocations
3017 Details : Checks validity of on-disk block bitmap. Also it does better
3018 checking of number of applied preallocations. When corruption is
3019 found, it turns filesystem readonly to prevent further corruptions.
3023 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3024 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3025 window to 0. If it does hit cache miss, and read-pattern is still
3026 stride-io mode, does not reset the stride window, but also does not
3027 increase the stride window length in this case.
3031 Frequency : only for big-endian servers
3032 Description: Check if system is big-endian while mounting fs with extents feature
3033 Details : Mounting a filesystem with extents feature will fail on big-endian
3034 systems since ext3-based ldiskfs is not supported on big-endian
3035 systems. This can be over-riden with "bigendian_extents" mount option.
3037 Severity : enhancement
3039 Description: The root squash functionality
3040 Details : A security feature, which is to prevent users from being able
3041 to mount lustre on their desktop, run as root, and delete
3042 all of the files in the filesystem. The goal is accomplished by
3043 remapping user id (UID) and group id (GID) of the root user to
3044 a UID and GID specified by the system administartor via Lustre
3045 configuration management server (MGS). The functionality also
3046 allows to specify sets of clients for which the remapping does
3051 Description: Excessive recovery window
3052 Details : With AT enabled, the recovery window can be excessively long (6000+
3053 seconds). To address this problem, we no longer use
3054 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3055 timeout no longer depends on the service time, it is set to
3056 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3057 time via pb_service_time.
3061 Description: Watchdog triggered on MDS failover
3062 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3063 the OSTs know that the MDS "UUID" can be reused for the same export
3064 from a different NID, so we do not need to wait for the export to be
3068 Frequency : rare, only if using MMP with Linux RAID
3070 Description: MMP doesn't work with Linux RAID
3071 Details : While using HA for Lustre servers with Linux RAID, it is possible
3072 that MMP will not detect multiple mounts. To make this work we
3073 need to unplug the device queue in RAID when the MMP block is being
3074 written. Also while reading the MMP block, we should read it from
3075 disk and not the cached one.
3077 Severity : enhancement
3079 Description: open file using fid
3080 Details : A file can be opened using just its fid, like
3081 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3084 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3087 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3088 Details : add proper locking for creation of jbd_4k slab cache
3092 Description: MMP check in ext3_remount() fails without displaying any error
3093 Details : When multiple mount protection fails during remount, proper error
3096 Severity : enhancement
3098 Description: Allow stripe size to be up to 4G-64k
3099 Details : Fix math logic to allow large stripe sizes.
3103 Description: add check for >8TB ldiskfs filesystems
3104 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3105 Don't allow >8TB ldiskfs filesystems to be mounted without
3106 force_over_8tb mount option
3111 Descriptoin: OSS complains "too many blocks" and crashes
3112 Details : In filter_commitrw_write():
3113 1. lquota_chkquota();
3114 2. obdo_from_inode();
3115 3. lquota_pending_commit();
3116 "2." has changed uid and gid in this case, then caused
3117 lquota_chkquota() and lquota_pending_commit() to use
3118 different uid and gid. That is the root of the bug.
3122 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3123 Details : Detect and print out the 'stripe count = -1' case properly.
3127 Description: Out or order replies might be lost on replay
3128 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3129 request with tid smaller than one currently being added, add it
3130 to the start, not end of the list.
3132 --------------------------------------------------------------------------------
3134 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3136 * Support for kernels:
3137 2.6.5-7.283 (SLES 9),
3138 2.6.9-55.EL (RHEL 4),
3139 2.6.16.46-0.14 (SLES 10),
3140 2.6.18.8 vanilla (kernel.org)
3141 * Client support for unpatched kernels:
3142 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3143 2.6.16 - 2.6.22 vanilla (kernel.org)
3144 * Due to recently discovered recovery problems, we do not recommend
3145 using patchless RHEL 4 clients with this or any earlier release.
3146 * Recommended e2fsprogs version: 1.39.cfs8
3147 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3148 * Starting with this release, the ldiskfs backing filesystem required
3149 by Lustre is now in its own package, lustre-ldiskfs. This package
3150 should be installed. It is versioned separately from Lustre and
3151 may be released separately in future.
3155 Description: block reactivating mgc import until all deactivates complete
3156 Details : Fix race when failing back MDT/MGS to itself (testing)
3159 Frequency : at statup only
3161 Description: mds_lov_synchronize race leads to various problems
3162 Details : simultaneous MDT->OST connections at startup can cause the
3163 sync to abort, leaving the OSC in a bad state.
3165 Severity : enhancement
3167 Description: add optional extra BUILD_VERSION info
3168 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3169 to override the lustre version.
3172 Frequency : 2.6.18 servers only
3174 Description: ll_kern_mount() doesn't release the module reference
3175 Details : The ldiskfs module reference count never drops down to 0
3176 because ll_kern_mount() doesn't release the module reference.
3181 Description: server LBUG when using old ost_num_threads parameter
3182 Details : Accept the old ost_num_threads parameter but warn that it
3183 is deprecated, and fix an off-by-one error that caused an LBUG.
3188 Description: Transient SCSI error results in persistent IO issue
3189 Details : iobuf->dr_error is not reinitialized to 0 between two
3193 Frequency : sometimes when underlying device returns I/O errors
3195 Description: OSTs not going read-only during write failures
3196 Details : OSTs are not remounted read-only when the journal commit threads
3197 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3198 instead of the ext3 wrappers.
3202 Description: poor connect scaling with increasing client count
3203 Details : Don't run filter_grant_sanity_check for more than 100 exports
3204 to improve scaling for large numbers of clients.
3207 Frequency : SLES10 only
3209 Description: sanity-quota.sh quotacheck failed: rc = -22
3210 Details : Quotas cannot be enabled on SLES10.
3213 Frequency : liblustre clients only
3215 Description: getdirentries does not give error when run on compute nodes
3216 Details : getdirentries does not fail when the size specified as an argument
3217 is too small to contain at least one entry
3219 Severity : enhancement
3221 Description: Add LNET router traceability for debug purposes
3222 Details : If a checksum failure occurs with a router as part of the
3223 IO path, the NID of the last router that forwarded the bulk data
3224 is printed so it can be identified.
3229 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3230 Details : Due to a race condition, liblustre clients were occasionally
3231 evicted incorrectly.
3233 Severity : enhancement
3235 Description: lfs setstripe use optional parameters instead of postional
3238 Severity : enhancement
3240 Description: Nanosecond timestamp support for ldiskfs
3241 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3242 resolution timestamps. There is not yet support for this at
3243 the Lustre filesystem level.
3246 Frequency : during server recovery
3248 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3249 Details : request with rq_no_resend flag not awake l_wait_event if they get
3253 Frequency : nfs export on patchless client
3255 Description: connectathon hang when test nfs export over patchless client
3256 Details : Disconnected dentry cannot be found with lookup, so we do not need
3257 to unhash it or make it invalid
3260 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3261 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3262 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3267 Description: open req refcounting wrong on reconnect
3268 Details : If reconnect happened between getting open reply from server and
3269 call to mdc_set_replay_data in ll_file_open, we will schedule
3270 replay for unreferenced request that we are about to free.
3271 Subsequent close will crash in variety of ways.
3272 Check that request is still eligible for replay in
3273 mdc_set_replay_data().
3278 Description: disable writes to filesystem when reading health_check file
3279 Details : the default for reading the health_check proc file has changed
3280 to NOT do a journal transaction and write to disk, because this
3281 can cause reads of the /proc file to hang and block HA state
3282 checking on a healthy but otherwise heavily loaded system. It
3283 is possible to return to the previous behaviour during configure
3284 with --enable-health-write.
3286 Severity : enhancement
3288 Description: 64-bit inode version
3289 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3290 to the inode. This will be required for version-based recovery.
3295 Description: MDS fails to start if a duplicate client export is detected
3296 Details : in some rare cases it was possible for a client to connect to
3297 an MDS multiple times. Upon recovery the MDS would detect this
3298 and fail during startup. Handle this more gracefully.
3300 Severity : enhancement
3302 Description: Add -o localflock option to simulate old noflock
3304 Details : This will achieve local-only flock/fcntl locks
3310 Description: log_commit_thread vs filter_destroy race leads to crash
3311 Details : Take import reference before releasing llog record semaphore
3316 Description: Wrong request locking in request set processing
3317 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3318 imp_delayed_list, in this place should be used imp_lock.
3321 Frequency : when reconnection
3323 Description: Grant Leak when osc reconnect to OST
3324 Details : When osc reconnect ost, OST(filter) should check whether it
3325 should grant more space to client by comparing fed_grant and
3326 cl_avail_grant, and return the granted space to client instead
3327 of "new granted" space, because client will call osc_init_grant
3328 to update the client grant space info.
3331 Frequency : when client reconnect to OST
3333 Description: Grant Leak when osc do resend and replay bulk write
3334 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3335 bulk write request, because the grant info while be sync between
3336 OSC and OST when reconnect, and we should ignore the grant info
3337 these of resend/replay write req.
3342 Description: Grant space more than available left space sometimes.
3343 Details : When then OST is about to be full, if two bulk writing from
3344 different clients came to OST. Accord the avaliable space of the
3345 OST, the first req should be permitted, and the second one
3346 should be denied by ENOSPC. But if the seconde arrived before
3347 the first one is committed. The OST might wrongly permit second
3348 writing, which will cause grant space > available space.
3351 Frequency : when client is evicted
3353 Description: Grant might be wrongly erased when osc is evicted by OST
3354 Details : when the import is evicted by server, it will fork another
3355 thread ptlrpc_invalidate_import_thread to invalidate the
3356 import, where the grant will be set to 0. While the original
3357 thread will update the grant it got when connecting. So if
3358 the former happened latter, the grant will be wrongly errased
3359 because of this race.
3364 Description: Checking Stale with correct fid
3365 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3366 to check whether it is stale, because sometimes, we want the
3367 enqueue happened anyway, and op_data.fid2 will not be initialized.
3369 Severity : enhancement
3371 Description: update patchless client
3372 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3375 Frequency : only with 2.4 kernel
3377 Description: random memory corruption
3378 Details : size of struct ll_inode_info is to big for union inode.u and this
3379 can be cause of random memory corruption.
3384 Description: Memory leak in recovery
3385 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3386 It should also check obd_fail before fsfilt_start, otherwise if
3387 fsfilt_start return -EROFS,(failover mds during mds recovery).
3388 then the req will return with repmsg->transno = 0 and rc = EROFS.
3389 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3390 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3391 be freed no matter whether fsfilt_commit success or not.
3394 Frequency : only with huge count clients
3396 Description: Prevents from taking the superblock lock in llap_from_page for
3398 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3399 for llap_from_page call in ll_removepage prevents from taking the
3400 superblock lock for a soon died page.
3405 Description: Not check open intent error before release open handle
3406 Details : in some rare cases, the open intent error is not checked before
3407 release open handle, which may cause
3408 ASSERTION(open_req->rq_transno != 0), because it tries to release
3409 the failed open handle.
3414 Description: Set cat log bitmap only after create log success.
3415 Details : in some rare cases, the cat log bitmap is set too early. and it
3416 should be set only after create log success.
3420 Description: Accessing a block bevice can re-enable I/O when Lustre is
3421 tearing down a device.
3422 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3427 Description: service threads may hog cpus when there are a lot of requests
3428 Details : Insert cond_resched to give other threads a chance to use some CPU
3433 Description: the cat log was not initialized in recovery
3434 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3435 unlink log on mds will not be initialized until mds post
3436 recovery. And also in mds post recovery, the unlink log will
3437 initialization will be done asynchronausly, so there will be race
3438 between add unlink log and unlink log initialization.
3442 Description: brw_stats were being printed incorrectly
3443 Details : brw_stats were being printed as log2 but all of them were not
3444 recorded as log2. Also remove some code duplication arising from
3445 filter_tally_{read,write}.
3449 Frequency : rare, only in recovery.
3450 Description: ASSERTION(req->rq_type != LI_POISON) failed
3451 Details : imp_lock should be held while iterating over imp_sending_list for
3452 prevent destroy request after get timeout in ptlrpc_queue_wait.
3456 Description: replay-single.sh test 52 fails
3457 Details : A lock's skiplist need to be cleanup when it being unlinked
3458 from its resource list.
3462 Description: Short directio read returns full requested size rather than
3464 Details : Direct I/O operations should return actual amount of bytes
3465 transferred rather than requested size.
3467 Severity : enhancement
3469 Description: metadata RPC reduction (e.g. for rm performance)
3470 Details : decrease the amount of synchronous RPC between clients and servers
3471 by canceling conflicing lock before the operation on the client
3472 and packing thier handles into the main operation RPC to server.
3474 Severity : enhancement
3476 Description: Async OSC create to avoid the blocking unnecessarily.
3477 Details : If an OST has no remaining object, system will block on the
3478 creation when it needs to create a new object on this OST. Now,
3479 ways use pre-created objects when available, instead of blocking on
3480 an empty osc while others are not empty. If we must block, we
3481 block for the shortest possible period of time.
3485 Description: improve handling recoverable errors
3486 Details : if request processig with error which can be recoverable on server
3487 request should be resend, otherwise page released from cache and
3490 Severity : enhancement
3492 Description: refine locking for avoid write wrong info into lov_objid file
3493 Details : fix possible races with add new target and write/update data in
3496 --------------------------------------------------------------------------------
3498 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3503 Frequency : on some architectures
3505 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3506 Details : Uninitialized flags sometimes cause configuration commands to
3510 Frequency : patchless clients only
3512 Description: missing __iget() symbol export
3513 Details : The __iget() symbol export is missing. To avoid the need for
3514 this on patchless clients the deathrow inode reaper is turned
3515 off, and we depend on the VM to clean up old inodes. This
3516 dependency was during via the fix for bug 12181.
3518 --------------------------------------------------------------------------------
3520 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3522 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3523 INTEROPERATE with older versions automatically. In many cases a
3524 special upgrade step is needed. Please read the
3525 user documentation before upgrading any part of a live system.
3526 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3527 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3528 * WARNING: Lustre configuration and startup changes are required with
3529 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3533 Severity : enhancement
3535 Description: MountConf
3536 Details : Lustre configuration is now managed via mkfs and mount
3537 commands instead of lmc and lconf. New obd types (MGS, MGC)
3538 are added for dynamic configuration management. See
3539 https://mail.clusterfs.com/wikis/lustre/MountConf for
3542 Severity : enhancement
3544 Description: dynamic OST addition
3545 Details : OSTs can now be added to a live filesystem
3547 Severity : enhancement
3549 Description: startup order invariance
3550 Details : MDTs and OSTs can be started in any order. Clients only
3551 require the MDT to complete startup.
3553 Severity : enhancement
3555 Description: parallel, asynchronous orphan cleanup
3556 Details : orphan cleanup is now performed in separate threads for each
3557 OST, allowing parallel non-blocking operation.
3559 Severity : enhancement
3561 Description: optimized stripe assignment
3562 Details : stripe assignments are now made based on ost space available,
3563 ost previous usage, and OSS previous usage, in order to try
3564 to optimize storage space and networking resources.
3566 Severity : enhancement
3568 Description: Permanently set tunables
3569 Details : All writable /proc/fs/lustre tunables can now be permanently
3570 set on a per-server basis, at mkfs time or on a live system.
3572 Severity : enhancement
3574 Description: Lustre message v2
3575 Details : Add lustre message format v2.
3577 Severity : enhancement
3579 Description: client OST exclusion list
3580 Details : Clients can be started with a list of OSTs that should be
3581 declared "inactive" for known non-responsive OSTs.
3585 Description: SPEC SFS validation failure on NFS v2 over lustre.
3586 Details : Changes the blocksize for regular files to be 2x RPC size,
3587 and not depend on stripe size.
3589 Severity : enhancement
3591 Description: Multiple MD RPCs in flight.
3592 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3593 To avoid overly-overloading MDS, introduce a limit on number of
3594 MDS RPCs in flight for a single client and add /proc controls
3595 to adjust this limit.
3597 Severity : enhancement
3599 Description: client read/write statistics
3600 Details : Add client read/write call usage stats for performance
3601 analysis of user processes.
3602 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3603 file access. extents_stats shows chunk size distribution.
3604 extents_stats_per_process show chunk size distribution per
3607 Severity : enhancement
3609 Description: mds statistics
3610 Details : Add detailed mds operations statistics in
3611 /proc/fs/lustre/mds/*/stats.
3615 Description: Failure of copying files with lustre special EAs.
3616 Details : Client side always return success for setxattr call for lustre
3617 special xattr (currently only "trusted.lov").
3622 Description: Refcount LNET uuids
3623 Details : The global LNET uuid list grew linearly with every startup;
3624 refcount repeated list entries instead of always adding to
3627 Severity : enhancement
3629 Description: Dynamic service threads
3630 Details : Within a small range, start extra service threads
3631 automatically when the request queue builds up.
3634 Frequency : mixed-endian client/server environments
3636 Description: mixed-endian crashes
3637 Details : The new msg_v2 system had some failures in mixed-endian
3640 Severity : enhancement
3642 Description: Easy OST removal
3643 Details : OSTs can be permanently deactivated with e.g. 'lctl
3644 conf_param lustre-OST0001.osc.active=0'
3646 Severity : enhancement
3648 Description: MGS proc entries
3649 Details : Added basic proc entries for the MGS showing what filesystems
3652 Severity : enhancement
3654 Description: provide MGS failover
3655 Details : Added config lock reacquisition after MGS server failover.
3657 Severity : enhancement
3659 Description: add Linux 2.4 support
3660 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3664 Description: a large application tries to do I/O to the same resource and dies
3665 in the middle of it.
3666 Details : Check the req->rq_arrival time after the call to
3667 ost_brw_lock_get(), but before we do anything about
3668 processing it & sending the BULK transfer request. This
3669 should help move old stale pending locks off the queue as
3670 quickly as obd_timeout.
3673 Frequency : when an incorrect nid is specified during startup
3675 Description: ptlrpc connect to non-existant node causes kernel crash
3676 Details : LNET can't be re-entered from an event callback, which
3677 happened when we expire a message after the export has been
3678 cleaned up. Instead, hand the zombie cleanup off to another
3681 Severity : enhancement
3683 Description: plain/inodebits lock performance improvement
3684 Details : Grouping plain/inodebits in granted list by their request modes
3685 and bits policy, thus improving the performance of search through
3689 Frequency : only if OST filesystem is corrupted
3691 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3692 Details : for a short time RPCs with bulk IO are in the replay list,
3693 but replay of bulk IOs is unimplemented. If the OST filesystem
3694 is corrupted due to disk cache incoherency and then replay is
3695 started it is possible to trip an assertion. Avoid putting
3696 committed RPCs into the replay list at all to avoid this issue.
3699 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3701 Description: System hang on startup
3702 Details : This bug allowed the liblustre (e.g. catamount) client to
3703 return to the app before handling all startup RPCs. This
3704 could leave the node unresponsive to lustre network traffic
3705 and manifested as a server ptllnd timeout.
3707 Severity : enhancement
3709 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3710 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3711 boolean to control whether to print peer debug info when a
3712 client's RPC times out.
3715 Frequency : only for kernels with patches from Lustre below 1.4.3
3717 Description: Remove old rdonly API
3718 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3721 Frequency : only for devices with external journals
3723 Description: Set external device read-only also
3724 Details : During a commanded failover stop, we set the disk device
3725 read-only while the server shuts down. We now also set any
3726 external journal device read-only at the same time.
3729 Frequency : when upgrading from 1.4 while trying to change parameters
3731 Description: The wrong (new) MDC name was used when setting parameters for
3732 upgraded MDT's. Also allows changing of OSC (and MDC)
3733 parameters if --writeconf is specified at tunefs upgrade time.
3736 Frequency : when setting specific ost indicies
3738 Description: QOS code breaks on skipped indicies
3739 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3740 created with --index need not be sequential.
3742 Severity : enhancement
3744 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3745 Details : The uninit_groups feature works in conjunction with the kernel
3746 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3747 up the pass1 processing of e2fsck. This is a read-only feature
3748 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3749 mount filesystems that have had this feature enabled.
3751 Severity : enhancement
3753 Description: Improve multi-block allocation algorithm to avoid fragmentation
3754 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3755 allocation locality and avoid filesystem fragmentation.
3757 ------------------------------------------------------------------------------
3759 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3761 * Support for kernels:
3762 2.4.21-47.0.1.EL (RHEL 3)
3763 2.6.5-7.283 (SLES 9)
3764 2.6.9-42.0.10.EL (RHEL 4)
3765 2.6.12.6 vanilla (kernel.org)
3766 2.6.16.27-0.9 (SLES 10)
3767 * Recommended e2fsprogs version: 1.39.cfs5
3769 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3773 Frequency : occasional, depends on client load and configuration
3774 Bugzilla : 12181, 12203
3775 Description: data loss for recently-modified files
3777 Details : In some cases it is possible that recently written or created
3778 files may not be written to disk in a timely manner (this should
3779 normally be within 30s unless client IO load is very high).
3780 The problem appears as zero-length files or files that are a
3781 multiple of 1MB in size after a client crash or client eviction
3782 that are missing data at the end of the file.
3784 This problem is more likely to be hit on clients where files are
3785 repeatedly created and unlinked in the same directory, clients
3786 have a large amount of RAM, have many CPUs, the filesystem has
3787 many OSTs, the clients are rebooted frequently, and/or the files
3788 are not accessed by other nodes after being written.
3790 The presence of the problem can be detected by looking at
3791 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3792 smaller than the second (nr_unused) then dirty files will not
3793 be flushed automatically to disk. "sync; sleep 10" should be
3794 run several times on the node before unmounting it to update
3795 Lustre (this is also safe to run on nodes without this problem).
3797 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3798 that can cause this same problem, so customers using that kernel
3799 also need to update the kernel in addition to Lustre. In order
3800 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3802 It is normal that files written just before a client crash (less
3803 than 30s) may not yet have been flushed to disk, even for local
3807 Frequency : frequent on thin XT3 nodes
3809 Description: UUID collision on thin XT3 Linux nodes
3810 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3811 randomly, since we relied on an insufficiently-seeded PRNG.
3816 Description: OSS hangs after "All ost request buffers busy"
3817 Details : A deadlock between quota and journal operations caused OSS
3818 hangs after printing "All ost request buffers busy."
3821 Frequency : always on liblustre builds
3823 Description: Cleanup compiler warnings on liblustre
3826 Frequency : always on liblustre builds on XT3
3828 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3832 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3833 > 32000 subdirectories
3834 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3835 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3836 subdirectory count crosses 32000. This will aid e2fsck to
3837 correctly handle more than 32000 subdirectories.
3840 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3842 Description: System hang on startup
3843 Details : This bug allowed the liblustre (e.g. catamount) client to
3844 return to the app before handling all startup RPCs. This
3845 could leave the node unresponsive to lustre network traffic
3846 and manifested as a server ptllnd timeout.
3848 Severity : enhancement
3850 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3851 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3852 boolean to control whether to print peer debug info when a
3853 client's RPC times out.
3858 Description: make O_SYNC working on 2.6 kernels
3859 Details : 2.6 kernels use different method for mark pages for write,
3860 so need add a code to lustre for O_SYNC work.
3865 Description: Failure to close file and release space on NFS
3866 Details : Put inode details into lock acquired in ll_intent_file_open.
3867 Use mdc_intent_lock in ll_intent_open to properly
3868 detect all kind of errors unhandled by mdc_enqueue.
3873 Description: proc file read during shutdown sometimes raced obd removal,
3875 Details : Add lock to prevent obd access after proc file removal.
3878 Frequency : Only for files larger than 4GB on 32-bit clients.
3880 Description: improperly doing page alignment of locks
3881 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3882 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3887 Description: RPCs being resent when they shouldn't be
3888 Details : Some RPCs that should not be resent are being resent. This
3889 can cause inconsistencies in the RPC state machine. Do not
3890 resend such requests.
3893 Frequency : rare, only with NFS export
3895 Description: Crash on NFS re-export node
3896 Details : under very unusual load conditions an assertion is hit in
3897 ll_intent_file_open()
3900 Frequency : only if OST filesystem is corrupted
3902 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3903 Details : for a short time RPCs with bulk IO are in the replay list,
3904 but replay of bulk IOs is unimplemented. If the OST filesystem
3905 is corrupted due to disk cache incoherency and then replay is
3906 started it is possible to trip an assertion. Avoid putting
3907 committed RPCs into the replay list at all to avoid this issue.
3912 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3913 Details : Large single O_DIRECT read and write calls can fail to allocate
3914 a sufficiently large buffer to process the request. In case of
3915 allocation failure the allocation is retried with a smaller
3916 buffer and broken into smaller requests.
3918 Severity : enhancement
3920 Description: Add -o localflock option to simulate old noflock behaviour.
3921 Details : This will achieve local-only flock/fcntl locks coherentness.
3926 Description: versioning check is incomplete
3927 Details : Checking the version difference of client vs. server, report
3928 error if the gap is too big.
3933 Description: add support PG_writeback bit
3934 Details : add support for PG_writeback bit for Lustre, for more carefull
3935 work with page cache in 2.6 kernel. This also fix some deadlocks
3936 and remove hack for work O_SYNC with 2.6 kernel.
3938 Severity : enhancement
3940 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3941 Details : The uninit_groups feature works in conjunction with the kernel
3942 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3943 up the pass1 processing of e2fsck. This is a read-only feature
3944 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3945 mount filesystems that have had this feature enabled.
3947 Severity : enhancement
3949 Description: Improve multi-block allocation algorithm to avoid fragmentation
3950 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3951 allocation locality and avoid filesystem fragmentation.
3953 ------------------------------------------------------------------------------
3955 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3957 * Support for kernels:
3958 2.6.9-42.0.3.EL (RHEL 4)
3959 2.6.5-7.276 (SLES 9)
3960 2.4.21-47.0.1.EL (RHEL 3)
3961 2.6.12.6 vanilla (kernel.org)
3962 2.6.16.21-0.8 (SLES10)
3963 * Recommended e2fsprogs version: 1.39.cfs2-0
3965 * The backwards-compatible /proc/sys/portals symlink has been removed
3966 in this release. Before upgrading, please ensure that you change
3967 any configuration scripts or /etc/sysctl.conf files that access
3968 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3969 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3970 in advance of the upgrade on any system running Lustre 1.4.6 or
3971 newer, since /proc/sys/lnet was added in that version.
3972 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3976 Frequency : only when quota is used
3978 Description: avoid scanning export list for quota master
3979 Details : Change the algorithms to avoid scanning export list in order
3980 to improve the efficiency.
3983 Frequency : MDS failover only, very rarely
3985 Description: "went back in time" messages on mds failover
3986 Details : The greatest transno may be lost when the current operation
3987 finishes with an error (transno==0) and the client's last_rcvd
3988 record is over-written. Save the greatest transno in the
3989 mds_last_transno for this case.
3992 Frequency : always for specific kernels and striping counts
3994 Description: client may get "Matching packet too big" without ACL support
3995 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3996 when trying to access files in certain configurations. The
3997 clients should in fact be denied when mounting because they do
3998 not understand ACLs.
4001 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4003 Description: many clients connecting with IO in progress causes connect timeouts
4004 Details : Avoid synchronous journal commits to avoid delays caused by many
4005 clients connecting/disconnecting when bulk IO is in progress.
4006 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4007 OST_IO_PORTAL to avoid delays behind potentially many pending
4011 Frequency : occasionally with multiple writers to a single file
4013 Description: shared writes to file may result in wrong size reported by stat()
4014 Details : Allow growing of kms when extent lock is cancelled
4017 Frequency : always with random mmap IO to multi-striped file
4019 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4020 Details : Only if the hole is at the end of OST object so that kms is too
4021 small. Fix is to increase kms accordingly in ll_nopage.
4024 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4026 Description: writes to a missing object would leak memory on the OST
4027 Details : If there is an inconsistency between the MDS and OST filesystems,
4028 such that the MDS references an object that doesn't exist, writes
4029 to that object will leak memory due to incorrect cleanup in the
4030 error handling path, eventually running out of memory on the OST.
4035 Description: Creating too long symlink causes lustre errors
4036 Details : Check symlink and name lengths before sending requests to MDS.
4039 Frequency : only if flock is enabled (not on by default)
4041 Description: posix locks not released on fd closure on 2.6.9+
4042 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4043 kernels, this caused such locks not to be released on fd close and
4044 then assertions on fs unmount about still used locks.
4047 Frequency : MDS failover only, very rarely
4049 Description: clients may get ASSERTION(granted_lock != NULL)
4050 Details : When request was taking a long time, and a client was resending
4051 a getattr by name lock request. The were multiple lock requests
4052 with the same client lock handle and
4053 mds_getattr_name->fixup_handle_for_resent_request found one of the
4054 lock handles but later failed with ASSERTION(granted_lock != NULL).
4059 Description: handle->h_buffer_credits > 0, assertion failure
4060 Details : h_buffer_credits is zero after truncate, causing assertion
4061 failure. This patch extends the transaction or creates a new
4065 Frequency : NFS re-export or patchless client
4066 Bugzilla : 11179, 10796
4067 Description: Crash on NFS re-export node (__d_move)
4068 Details : We do not want to hash the dentry if we don't have a lock.
4069 But if this dentry is later used in d_move, we'd hit uninitialised
4070 list head d_hash, so we just do this to init d_hash field but
4071 leave dentry unhashed.
4074 Frequency : NFS re-export or patchless client
4076 Description: NFS exports has problem with symbolic link
4077 Details : lustre client didn't properly install dentry when re-exported
4078 to NFS or running patchless client.
4081 Frequency : NFS re-export or patchless client
4083 Description: Various nfs/patchless fixes.
4084 Details : fixes reuse disconected alias for lookup process - this fixes
4085 warning "find_exported_dentry: npd != pd",
4086 fix permission error with open files at nfs.
4087 fix apply umask when do revalidate.
4090 Frequency : occasional
4092 Description: Crash on NFS re-export node
4093 Details : calling clear_page() on the wrong pointer triggered oops in
4094 generic_mapping_read().
4097 Frequency : rarely, using O_DIRECT IO
4099 Description: unaligned directio crashes client with LASSERT
4100 Details : check for unaligned buffers before trying any requests.
4103 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4105 Description: stale data returned from RAID cache
4106 Details : If only a small amount of IO is done to the RAID device before
4107 reading it again it is possible to get stale data from the RAID
4108 cache instead of reading it from disk.
4111 Frequency : always for sles10 kernel
4113 Description: sles10 support
4114 Details : ll_follow_link: compile fixes and using of nd_set_link
4115 under newer kernels.
4118 Frequency : depends on arch, kernel and compiler version, always on sles10
4121 Description: recursive or deep enough symlinks cause stack overflow
4122 Details : getting rid of large stack-allocated variable in
4126 Frequency : depends on hardware
4128 Description: lustre write performance loss in the SLES10 kernel
4129 Details : the performance loss is caused by using of write barriers in the
4130 ext3 code. The SLES10 kernel turns barrier support on by
4131 default. The fix is to undo that change for ldiskfs.
4133 ------------------------------------------------------------------------------
4135 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4137 * Support for kernels:
4138 2.6.9-42.0.3EL (RHEL 4)
4139 2.6.5-7.276 (SLES 9)
4140 2.4.21-47.0.1.EL (RHEL 3)
4141 2.6.12.6 vanilla (kernel.org)
4145 Frequency : quota enabled and large files being deleted
4147 Description: releasing more than 4GB of quota at once hangs OST
4148 Details : If a user deletes more than 4GB of files on a single OST it
4149 will cause the OST to spin in an infinite loop. Release
4150 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4155 Description: statfs data retrieved from /proc may be stale or zero
4156 Details : When reading per-device statfs data from /proc, in the
4157 {kbytes,files}_{total,free,avail} files, it may appear
4158 as zero or be out of date.
4161 Frequency : systems with MD RAID1 external journal devices
4163 Description: lconf's call to blkid is confused by RAID1 journal devices
4164 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4165 of returning all block devices that match the journal UUID.
4168 Frequency : always, for aggregate stripe size over 4GB
4170 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4171 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4172 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4175 Frequency : always if "lfs find" used on a local file/directory
4177 Description: "lfs find" segfaults if used on a local file/directory
4178 Details : The case where a directory component was not specified wasn't
4179 handled correctly. Handle this properly.
4182 Frequency : always on ppc64
4184 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4185 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4186 perform endianness conversion.
4189 Frequency : rarely (truncate to non-zero file size after write under load)
4190 Bugzilla : 10730, 10687
4191 Description: Files padded with zeros to next 4K multiple
4192 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4193 it is possible that files that are truncated to a non-zero size
4194 immediately after being written are filled with zero bytes beyond
4195 the truncated size. No file data is lost.
4197 Severity : enhancement
4199 Description: Allow recovery/failover for liblustre clients.
4200 Details : liblustre clients were unaware of failover configurations until
4203 Severity : enhancement
4205 Description: user file locks should fail when not mounting with flock option
4206 Details : Set up an error-returning stub in ll_file_operations.lock field
4207 to prevent incorrect behaviour when client is mounted without
4208 flock option. Also, set up properly f_op->flock field for
4212 Frequency : always on ia64
4214 Description: "lfs df" loops on printing out MDS statfs information
4215 Details : The obd_ioctl_data was not initialized and in some systems
4216 this caused a failure during the ioctl that did not return
4217 an error. Initialize the struct and return an error on failure.
4220 Frequency : SLES 9 only
4222 Description: Error of copying files with lustre special EAs as root
4223 Details : Client side always return success for setxattr call for lustre
4224 special xattr (currently only "trusted.lov").
4227 Frequency : rarely on clusters with both ia64+i386 clients
4229 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4230 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4231 clients are doing concurrent writes to the same file it is
4232 possible that the ia64 clients may LASSERT because the OST
4233 extent locks are not PAGE_SIZE aligned. Ensure that grown
4234 locks are always aligned on the request boundary.
4237 Frequency : specific use, occasional
4239 Description: Overwriting in use executable truncates on-disk binary image
4240 Details : If one node attempts to overwrite an executable in use by
4241 another node, we now correctly return ETXTBSY instead of
4242 truncating the file.
4244 Severity : enhancement
4246 Description: Async OSC create to avoid the blocking unnecessarily.
4247 Details : If an OST has no remaining object, system will block on the creating
4248 when need to create a new object on this OST. Now, ways use
4249 pre-created objects when available, instead of blocking on an
4250 empty osc while others are not empty. If we must block, we block
4251 for the shortest possible period of time.
4256 Description: chmod on Lustre root is propagated to other clients
4257 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4258 invalid by the follow_mount time.
4261 Frequency : liblustre clients only
4263 Description: Race in 'instant cancel' lock handling could lead to such locks
4264 never to be granted in case of SMP MDS
4265 Details : Do not destroy not yet granted but cbpending locks in
4269 Frequency : replay/resend of open
4271 Description: non null lock assetion failure in mds_intent_policy
4272 Details : Trying to replay/resend lockless open requests resulted in
4273 mds_open() returning 0 with no lock. Now it sets a flag if
4274 a lock is going to be returned.
4276 Severity : enhancement
4278 Description: Checksum enhancements
4279 Details : New checksum enhancements allow for resending RPCs that failed
4282 Severity : enhancement
4284 Description: Tunables on number of dirty pages in cacche
4285 Details : Allow to set limit on number of dirty pages cached.
4290 Description: client crash on unmount - lock still has references
4291 Details : In some error handling cases it was possible to leak a lock
4292 reference on a client while accessing a file. This was not
4293 harmful to the client during operation, but would cause the
4294 client to crash when the filesystem is unmounted.
4297 Frequency : specific case, rare
4299 Description: ETXTBSY on mds though file not in use by client
4300 Details : ETXTBSY is no longer incorrectly returned when attempting to
4301 chmod or chown a directory that the user previously tried to
4302 execute or a currently-executing binary.
4305 Frequency : extremely rare except on liblustre-based clients
4307 Description: Lustre space not freed when files are deleted
4308 Details : Clean up open-unlinked files after client eviction. Previously
4309 the unlink was skipped and the files remained as orphans.
4314 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4315 Details : In some cases it was possible to send a blocking callback to
4316 a client doing a glimpse, even though that client didn't get
4317 a lock granted. When the glimpse lock is cancelled on the OST
4318 the freed lock is left on the waiting list and corrupted the list.
4321 Frequency : all core dumps
4323 Description: Broke core dumps to lustre
4324 Details : Negative dentry may be unhashed if parent does not have UPDATE
4325 lock, but some callers, e.g. do_coredump, expect dentry to be
4326 hashed after successful create, hash it in ll_create_it.
4328 ------------------------------------------------------------------------------
4330 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4332 * Support for kernels:
4333 2.6.9-42.0.2.EL (RHEL 4)
4334 2.6.5-7.276 (SLES 9)
4335 2.4.21-40.EL (RHEL 3)
4336 2.6.12.6 vanilla (kernel.org)
4340 Frequency : always on RHEL 3
4342 Description: Number of open files grows over time
4343 Details : The number of open files grows over time, whether or not
4344 Lustre is started. This was due to a filp leak introduced
4345 by one of our kernel patches.
4347 ------------------------------------------------------------------------------
4349 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4351 * Support for kernels:
4352 2.6.9-42.EL (RHEL 4)
4353 2.6.5-7.276 (SLES 9)
4354 2.4.21-40.EL (RHEL 3)
4355 2.6.12.6 vanilla (kernel.org)
4360 Bugzilla : 5719, 9635, 9792, 9684
4361 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4362 Details : If a server is under heavy load and cannot reply to new
4363 connection requests before the client resends the (re)connect,
4364 the connection handling code can behave badly if two service
4365 threads are concurrently handing separate (re)connections from
4366 the same client. Add better locking to the connection handling
4367 code, and ensure that only a single connection will be processed
4368 for a given client UUID, even if the lock is dropped.
4370 Severity : enhancement
4372 Description: add TCP zero-copy support to kernel
4373 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4374 sends if the hardware supports scatter-gather and checksumming.
4375 This allows socklnd to do client-write and server-read more
4376 efficiently and reduce CPU utilization from skbuf copying.
4379 Frequency : only if NFS exporting from client
4381 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4382 Details : If setting attributes on a file created under NFS that had
4383 never been opened it would be possible to oops the client
4384 if the file had no objects.
4387 Frequency : always for liblustre
4389 Description: liblustre client does MDS+OSTs setattr RPC for each write
4390 Details : When doing a write from a liblustre client, the client
4391 incorrectly issued an RPC to the MDS and each OST the file was
4392 striped over in order to update the timestamps. When writing
4393 with small chunks and many clients this could overwhelm the MDS
4394 with RPCs. In all cases it would slow down the write because
4395 these RPCs are unnecessary.
4397 Severity : enhancement
4399 Description: allow number of MDS service threads to be changed at module load
4400 Details : It is now possible to change the number of MDS service threads
4401 running. Adding "options mds mds_num_threads={N}" to the MDS's
4402 /etc/modprobe.conf will set the number of threads for the next
4403 time Lustre is restarted (assuming the "mds" module is also
4404 reloaded at that time). The default number of threads will
4405 stay the same, 32 for most systems.
4410 Description: OST crash if filesystem is unformatted or corrupt
4411 Details : If an OST is started on a device that has never been formatted
4412 or if the filesystem is corrupt and cannot even mount then the
4413 error handling cleanup routines would dereference a NULL pointer.
4418 Description: NULL pointer deref in llap_from_page.
4419 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4420 incorrect) mapping if the page was truncated/reclaimed while it was
4421 searched for. Check for this condition and skip such pages when
4422 doing readahead. Introduce extra check to llap_from_page() to
4423 verify page->mapping->host is non-NULL (so page is not anonymous).
4426 Frequency : Sometimes when using sys_sendfile
4428 Description: "page not covered by a lock" warnings from ll_readpage
4429 Details : sendfile called ll_readpage without right page locks present.
4430 Now we introduced ll_file_sendfile that does necessary locking
4431 around call to generic_file_sendfile() much like we do in
4435 Frequency : with certain MDS communication failures at client mount time
4437 Description: NULL pointer deref after failed client mount
4438 Details : a client connection request may delayed by the network layer
4439 and not be sent until after the PTLRPC layer has timed out the
4440 request. If the client fails the mount immediately it will try
4441 to clean up before the network times out the request. Add a
4442 reference from the request import to the obd device and delay
4443 the cleanup until the network drops the request.
4446 Frequency : occasionally during client (re)connect
4448 Description: assertion failure during client (re)connect
4449 Details : processing a client connection request may be delayed by the
4450 client or server longer than the client connect timeout. This
4451 causes the client to resend the connection request. If the
4452 original connection request is replied in this interval, the
4453 client may trip an assertion failure in ptlrpc_connect_interpret()
4454 which thought it would be the only running connect process.
4457 Frequency : only with obd_echo servers and clients that are rebooted
4459 Description: kernel BUG accessing uninitialized data structure
4460 Details : When running an obd_echo server it did not start the ping_evictor
4461 thread, and when a client was evicted an uninitialized data
4462 structure was accessed. Start the ping_evictor in the RPC
4463 service startup instead of the OBD startup.
4465 Severity : enhancement
4466 Bugzilla : 10193 (patchless)
4467 Description: Remove dependency on various unexported kernel interfaces.
4468 Details : No longer need reparent_to_init, exit_mm, exit_files,
4469 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4472 Frequency : rare (only users of deprecated and unsupported LDAP config)
4474 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4475 Details : LDAP apparently contains 'lustreName' attributes instead of
4476 'name'. A simple remapping of the name is sufficient.
4479 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4481 Description: waiting_locks_callback trips kernel BUG if client is evicted
4482 Details : Running with the dump_on_timeout debug flag turned on makes
4483 it possible that the waiting_locks_callback() can try to dump
4484 the Lustre kernel debug logs from an interrupt handler. Defer
4485 this log dumping to the expired_lock_main() thread.
4487 Severity : enhancement
4489 Description: Support NFS exporting on 2.6 kernels.
4490 Details : Implement non-rawops metadata methods for NFS server to use without
4491 changing NFS server code.
4494 Frequency : very rare (synthetic metadata workload only)
4496 Description: two racing renames might cause an MDS thread to deadlock
4497 Details : Running the "racer" program may cause one MDS thread to rename
4498 a file from being the source of a rename to being the target of
4499 a rename at exactly the same time that another thread is doing
4500 so, and the second thread has already enqueued these locks after
4501 doing a lookup of the target and is trying to relock them in
4502 order. Ensure that we don't try to re-lock the same resource.
4505 Frequency : only very large systems with liblustre clients
4507 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4508 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4509 all OSTs in parallel. This allows the eviction of stale liblustre
4510 clients to proceed much faster than if they were done in series,
4511 and also offers similar improvements for other set_info RPCs.
4516 Description: excessive CPU usage during initial read phase on client
4517 Details : During the initial read phase on a client, it would agressively
4518 retry readahead on the file, consuming too much CPU and impacting
4519 performance (since 1.4.5.8). Improve the readahead algorithm
4520 to avoid this, and also improve some other common cases (read
4521 of small files in particular, where "small" is files smaller than
4522 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4527 Description: MDS crash when receiving packet with unknown intent.
4528 Details : Do not LBUG in unknown intent case, just return -EFAULT
4530 Severity : enhancement
4531 Bugzilla : 9293, 9385
4532 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4533 Details : Do not serialize getattr (non-intent version) and statfs.
4536 Frequency : occasional, when OST network is overloaded/intermittent
4538 Description: client evicted by OST after bulk IO timeout
4539 Details : If a client sends a bulk IO request (read or write) the OST
4540 may evict the client if it is unresposive to its data GET/PUT
4541 request. This is incorrect if the network is overloaded (takes
4542 too long to transfer the RPC data) or dropped the OST GET/PUT
4543 request. There is no need to evict the client at all, since
4544 the pinger and/or lock callbacks will handle this, and the
4545 client can restart the bulk request.
4548 Frequency : Always when mmapping file with no objects
4550 Description: client crashes when mmapping file with no objects
4551 Details : Check that we actually have objects in a file before doing any
4552 operations on objects in ll_vm_open, ll_vm_close and
4558 Description: Request leak when working with deleted CWD
4559 Details : Introduce advanced request refcount tracking for requests
4560 referenced from lustre intent.
4562 Severity : Enhancement
4564 Description: Cache open file handles on client.
4565 Details : MDS now will return special lock along with openhandle, if
4566 requested and client is allowed to hold openhandle, even if unused,
4567 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4568 closing files for every read/write openration.
4570 Severity : Enhancement
4572 Description: Cache open negative dentries on client when possible.
4573 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4574 negative dentries on lock revocation.
4579 Description: Remounting a client read-only wasn't possible with a zconf mount
4580 Details : It wasn't possible to remount a client read-only with llmount.
4582 Severity : enhancement
4583 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4584 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4585 MPICH is included to add Lustre-specific ADIO interfaces.
4586 This is based closely on the UFS ADIO layer and only differs
4587 in file creation, in order to allow the OST striping to be set.
4588 This is user-contributed code and not supported by CFS.
4593 Description: extended inode attributes (immutable, append-only) work improperly
4594 when 2.4 and 2.6 kernels are used on client/server or vice versa
4595 Details : Introduce kernel-independent values for these flags.
4597 Severity : enhancement
4600 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4601 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4602 granularity. Now, Fractional MB granularity is be supported,
4603 this is very useful for low memory system.
4605 Severity : enhancement
4607 Description: Getattr by fid
4608 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4609 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4610 have locks and avoids getattr rpc requests for it on every stat.
4613 Frequency : Always, for filesystems larger than 2TB
4615 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4616 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4617 filesystems larger than 2TB to be mounted with mballoc enabled.
4620 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4622 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4623 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4624 filesystems larger than 2TB is mounted then the kernel will
4625 silently corrupt the start of the filesystem. CONFIG_LBD is
4626 enabled for all CFS-supported kernels, but the possibility of
4627 this happening with a modified kernel config exists.
4629 Severity : enhancement
4631 Description: add client O_DIRECT support for 2.6 kernels
4632 Details : It is now possible to do O_DIRECT reads and writes to files
4633 in the Lustre client mountpoint on 2.6 kernel clients.
4635 Severity : enhancement
4637 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4638 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4639 parallel, not waiting for response from every OST before sending
4640 a rpc to the next OST.
4645 Description: setattr vs write race when updating file timestamps
4646 Details : Client processes that update a file timestamp into the past
4647 right after writing to the file (e.g. tar) it is possible that
4648 the updated file modification time can be reset to the current
4649 time due to a race between processing the setattr and write RPC.
4651 Severity : enhancement
4653 Description: Bring 'lfs find' closer in line with regular Linux find.
4654 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4655 -print0 options and obtains all the needed info through the lustre
4658 Severity : enhancement
4660 Description: support up to 1024 configured devices on one node
4661 Details : change obd_dev array from statically allocated to dynamically
4662 allocated structs as they are first used to reduce memory usage
4667 Description: Flush dirty partially truncated pages during truncate
4668 Details : Immediatelly flush partially truncated pages in filter_setattr,
4669 this way we completely avoid having any pages in page cache on OST
4670 and can retire ugly workarounds during writes to flush such pages.
4675 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4676 Details : protect inode from truncation within vfs_unlink() context
4677 just take a reference before calling vfs_unlink() and release it
4678 when parent's i_sem is free.
4683 Description: last_id value checked outside lock on OST caused LASSERT failure
4684 Details : If there were multiple MDS->OST object precreate requests in
4685 flight, it was possible that the OST's last object id was checked
4686 outside a lock and incorrectly tripped an assertion. Move checks
4687 inside locks, and discard old precreate requests.
4690 Frequency : always, if extents are used on OSTs
4692 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4693 Details : OSTs using the extents format would not zero the high 16 bits of
4694 the index physical block number. This is not a problem for any
4695 OST filesystems smaller than 16TB, and no kernels support ext3
4696 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4697 new/modified files) and can be fixed for existing filesystems
4698 with e2fsprogs-1.39-cfs1.
4703 Description: import connection selection may be incorrect if timer wraps
4704 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4705 import connections to be ignored if the 32-bit jiffies counter
4706 wraps. Use a 64-bit jiffies counter.
4709 Frequency : very large clusters immediately after boot
4711 Description: LNET request buffers exhausted under heavy short-term load
4712 Details : If a large number of client requests are generated on a service
4713 that has previously never seen so many requests it is possible
4714 that the request buffer growth cannot keep up with the spike in
4715 demand. Instead of dropping incoming requests, they are held in
4716 the LND until the RPC service can accept more requests.
4719 Frequency : Sometimes during replay
4721 Description: Assertion failure in ll_local_open after replay.
4722 Details : If replay happened on an open request reply before we were able
4723 to set replay handler, reply will become not swabbed tripping the
4724 assertion in ll_local_open. Now we set the handler right after
4725 recognising of open request
4728 Frequency : very rare
4730 Description: kernel reports "badness in vsnprintf"
4731 Details : Reading from the "recovery_status" /proc file in small chunks
4732 may cause a negative length in lprocfs_obd_rd_recovery_status()
4733 call to vsnprintf() (which is otherwise harmless). Exit early
4734 if there is no more space in the output buffer.
4736 Severity : enhancement
4738 Description: clear OBD RPC statistics by writing to them
4739 Details : It is now possible to clear the OBD RPC statistics by writing
4740 to the "stats" file.
4745 Description: Client mtime is not the same on different clients after utimes
4746 Details : In some cases, the client was using the utimes() syscall on
4747 a file cached on another node. The clients now validate the
4748 ctime from the MDS + OSTs to determine which one is right.
4753 Description: Inability to activate failout mode
4754 Details : lconf script incorrectly assumed that in pythong string's numeric
4755 value is used in comparisons.
4758 Frequency : always with multiple stripes per file
4760 Description: Inefficient object allocation for mutli-stripe files
4761 Details : When selecting which OSTs to stripe files over, for files with
4762 a stripe count that divides evenly into the number of OSTs,
4763 the MDS is always picking the same starting OST for each file.
4764 Return the OST selection heuristic to the original design.
4769 Description: mount failures may take full timeout to return an error
4770 Details : Under some heavy load conditions it is possible that a
4771 failed mount can wait for the full obd_timeout interval,
4772 possibly several minutes, before reporting an error.
4773 Instead return an error as soon as the status is known.
4775 Frequency : quota enabled and large files being deleted
4777 Description: releasing more than 4GB of quota at once hangs OST
4778 Details : If a user deletes more than 4GB of files on a single OST it
4779 will cause the OST to spin in an infinite loop. Release
4780 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4785 Description: statfs data retrieved from /proc may be stale or zero
4786 Details : When reading per-device statfs data from /proc, in the
4787 {kbytes,files}_{total,free,avail} files, it may appear
4788 as zero or be out of date.
4791 Frequency : systems with MD RAID1 external journal devices
4793 Description: lconf's call to blkid is confused by RAID1 journal devices
4794 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4795 of returning all block devices that match the journal UUID.
4798 Frequency : always, for aggregate stripe size over 4GB
4800 Description: assertion fails when trying to use 4GB stripe size
4801 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4802 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4805 Frequency : always on ppc64
4807 Description: the first write on an ext3 filesystem with mballoc got stuck
4808 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4809 perform endianness conversion.
4811 ------------------------------------------------------------------------------
4813 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4815 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4816 INTEROPERATE with older versions automatically. Please read the
4817 user documentation before upgrading any part of a live system.
4818 * WARNING: Lustre networking configuration changes are required with
4819 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4822 * Support for kernels:
4823 2.6.9-22.0.2.EL (RHEL 4)
4824 2.6.5-7.244 (SLES 9)
4825 2.6.12.6 vanilla (kernel.org)
4828 Severity : enhancement
4829 Bugzilla : 7981/8208
4830 Description: Introduced Lustre Networking (LNET)
4831 Details : LNET is new networking infrastructure for Lustre, it includes
4832 a reorganized network configuration mode (see the user
4833 documentation for full details) as well as support for routing
4834 between different network fabrics. Lustre Networking Devices
4835 (LNDs) for the supported network fabrics have also been
4836 created for this new infrastructure.
4838 Severity : enhancement
4839 Description: Introduced Access control lists
4840 Details : clients can set ACLs on files and directories in order to have
4841 more fine-grained permissions than the standard Unix UGO+RWX.
4842 The MDS must be started with the "-o acl" mount option.
4844 Severity : enhancement
4845 Description: Introduced filesystem quotas
4846 Details : Administrators may now establish per-user quotas on the
4849 Severity : enhancement
4851 Description: Configuration change for the XT3
4852 The PTLLND is now used to run Lustre over Portals on the XT3
4853 The configure option(s) --with-cray-portals are no longer used.
4854 Rather --with-portals=<path-to-portals-includes> is used to
4855 enable building on the XT3. In addition to enable XT3 specific
4856 features the option --enable-cray-xt3 must be used.
4861 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4862 Details : A race between allocating a new llcd and re-getting the llcd_lock
4863 allowed another thread to grab newly-allocated llcd.
4865 Severity : enhancement
4867 Description: 2.6 OST async journal commit and locking fix to improve performance
4868 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4869 2.6 kernels are now async as they already were in 2.4 kernels so
4870 that they can commit concurrently with the network bulk transfer.
4871 For block-allocated files the filter allocation semaphore is held
4872 to avoid filesystem fragmentation during allocation. BKL lock
4873 removed for 2.6 xattr operations where it is no longer needed.
4878 Description: lconf incorrectly determined whether two IP networks could talk
4879 Details : In some more complicated routing and multiple-network
4880 configurations, lconf will avoid trying to make a network
4881 connection to a disjoint part of the IP space. It was doing the
4882 math incorrectly for one set of cases.
4887 Description: Fix for potential infinite loop processing records in an llog.
4888 Details : If an llog record is corrupted/zeroed, it is possible to loop
4889 forever in llog_process(). Validate the llog record length
4890 and skip the remainder of the block on error.
4893 Frequency : occasional (liblustre only)
4895 Description: liblustre could not open files whose last component is a symlink
4896 Details : sysio_path_walk() would incorrectly pass the open intent to
4897 intermediate path components.
4900 Frequency : rare (liblustre only with non-standard tuning)
4901 Bugzilla : 7201 (7350)
4902 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4903 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4904 after it has been released, only do this for referenced locks
4906 Severity : enhancement
4908 Description: specifying an (invalid) directory default stripe_size of -1
4909 would reset the directory default striping
4910 Details : stripe_size -1 was used internally to signal directory stripe
4911 removal, now use "all default" to signal dir stripe removal
4912 as a directory striping of "all default" is not useful
4915 Frequency : common for large clusters running liblustre clients
4917 Description: doing an ls when liblustre clients are running is slow
4918 Details : sending a glimpse AST to a liblustre client waits for every AST
4919 to time out, as liblustre clients will not respond. Since they
4920 cannot cache data we refresh the OST lock LVB from disk instead.
4922 Severity : enhancement
4924 Description: doing an ls at the same time as file IO can be slow
4925 Details : enqueue and other "small" requests can be blocked behind many
4926 large IO requests. Create a new OST IO portal for non-IO
4927 requests so they can be processed faster.
4930 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4932 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4933 created file with mode 000
4934 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4935 allow an ftruncate() as a non-root user to a file with mode 000.
4936 Lustre now respects this flag to disable mode checks when
4937 truncating a file owned by the user
4940 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4943 Description: Revoking locks from clients that went dead or catatonic might take
4945 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4946 cancellation of such locks instant on servers without waiting for
4947 any reply from clients. Clients drops these locks when cancel
4948 notification from server is received without replying.
4951 Frequency : liblustre-only, when liblustre client dies or becomes busy
4953 Description: Doing ls on Linux clients can take a long time with active
4955 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4956 granting such locks to it in the first place if possible. Locks
4957 are taken by proxy on the OST during the read or write and
4958 dropped immediately afterward. Add connect flags handling, do
4959 not grant locks to liblustre clients for glimpse ASTs.
4961 Severity : enhancement
4963 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4964 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4965 sequential reads and behaves badly with applications doing random
4966 reads. Improve it by reading ahead at least read region, and
4967 avoiding excessive large RPC for small reads.
4969 Severity : enhancement
4971 Description: Creating more than 1000 files for a single job may cause a load
4972 imbalance on the OSTs if there are also a large number of OSTs.
4973 Details : qos_prep_create() uses an OST index reseed value that is an
4974 even multiple of the number of available OSTs so that if the
4975 reseed happens in the middle of the object allocation it will
4976 still utilize the OSTs as uniformly as possible.
4981 Description: OST or MDS may oops in ping_evictor_main()
4982 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4983 but doesn't restart at beginning of obd_exports_timed list
4986 Severity : enhancement
4988 Description: improve by-nid export eviction on the MDS and OST
4989 Details : allow multiple exports with the same NID to be evicted at one
4990 time without re-searching the exports list.
4993 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4995 Description: MDS may oops in groups_free()
4996 Details : in rare race conditions a newly allocated group_info struct is
4997 freed again, and this can be NULL. The 2.4 compatibility code
4998 for groups_free() checked for a NULL pointer, but 2.6 did not.
5001 Frequency : common for liblustre clients doing little filesystem IO
5002 Bugzilla : 9352, 7313
5003 Description: server may evict liblustre clients accessing contended locks
5004 Details : if a client is granted a lock or receives a completion AST
5005 with a blocking AST already set it would not reply to the AST
5006 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5009 Frequency : lfs setstripe, only systems with more than 160 OSTs
5011 Description: unable to set striping with a starting offset beyond OST 160
5012 Details : llapi_create_file() incorrectly limited the starting stripe
5013 index to the maximum single-file stripe count.
5016 Frequency : LDAP users only
5018 Description: lconf did not handle in-kernel recovery with LDAP properly
5019 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5021 Severity : enhancement
5023 Description: bind OST threads to NUMA nodes to improve performance
5024 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5025 node and do their allocations there to localize memory access
5027 Severity : enhancement
5029 Description: llmount can determine client NID directly from Myrinet (GM)
5030 Details : the client NID code from gmnalnid was moved directly into
5031 llmount, removing the need to use this or specifying the
5032 client NID explicitly when mounting GM clients with zeroconf
5035 Frequency : if client is started with down MDS
5037 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5038 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5039 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5040 ptlrpc_import_delayed_req() because we didn't check if the
5041 request was interrupted, and we also didn't break out of the
5042 event loop if there was no timeout
5047 Description: data loss during non-page-aligned writes to a single file from
5048 both multiple nodes and multiple threads on one node at same time
5049 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5050 inconsistency led to false short-reads, that were cached and later
5051 used by ->prepare_write() to fill in partially written page,
5052 leading to data loss.
5055 Frequency : always, if lconf --abort_recovery used
5057 Description: lconf --abort_recovery fails with 'Operation not supported'
5058 Details : lconf was attempting to abort recovery on the MDT device and not
5061 Severity : enhancement
5063 Description: remove cleanup logs
5064 Details : replace lconf-generated cleanup logs with lustre internal
5065 cleanup routines. Eliminates the need for client-cleanup and
5068 Severity : enhancement
5070 Description: add support for EAs (user and system) on lustre filesystems
5071 Details : it is now possible to store extended attributes in the Lustre
5072 client filesystem, and with the user_xattr mount option it
5073 is possible to allow users to store EAs on their files also
5075 Severity : enhancement
5077 Description: Add possibility (config option) to show minimal available OST free
5079 Details : When compiled with --enable-mindf configure option, statfs(2)
5080 (and so, df) will return least minimal free space available from
5081 all OSTs as amount of free space on FS, instead of summary of
5082 free spaces of all OSTs.
5084 Severity : enhancement
5086 Description: do not expand extent locks acquired on OST-side
5087 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5088 by server: they are not cached anyway.
5091 Frequency : when mmap is used/binaries executed from Lustre
5093 Description: Unmmap pages before throwing them away from read cache.
5094 Details : llap_shrink cache now attempts to unmap pages before discarding
5095 them (if unmapping failed - do not discard). SLES9 kernel has
5096 extra checks that trigger if this unmapping is not done first.
5101 Description: lconf didn't resolve symlinks before checking to see whether a
5102 given mountpoint was already in use
5105 Frequency : when migrating failover services
5106 Bugzilla : 6395, 9514
5107 Description: When migrating a subset of services from a node (e.g. failback
5108 from a failover service node) the remaining services would
5109 time out and evict clients.
5110 Details : lconf --force (implied by --failover) sets the global obd_timeout
5111 to 5 seconds in order to quickly disconnect, but this caused
5112 other RPCs to time out too quickly. Do not change the global
5113 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5115 Severity : enhancement
5116 Frequency : if MDS is started with down OST
5117 Bugzilla : 9439,5706
5118 Description: Allow startup/shutdown of an MDS without depending on the
5119 availability of the OSTs.
5120 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5121 Add appropriate locking and lov-osc refcounts for safe
5122 cleaning. Add osc abort_inflight calls in case the
5123 synchronize never started.
5126 Frequency : occasional (Cray XT3 only)
5128 Description: root not authorized to access files in CRAY_PORTALS environment
5129 Details : The client process capabilities were not honoured on the MDS in
5130 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5131 been accessed by an authorized user then root was able to access
5132 the file on the local client also. The root user capabilities
5133 are now allowed on the MDS, as this environment has secure UID.
5136 Frequency : occasional
5138 Description: ldiskfs "too long searching" message happens too often
5139 Details : A debugging message (otherwise harmless) prints too often on
5140 the OST console. This has been reduced to only happen when
5141 there are fragmentation problems on the filesystem.
5146 Description: Division by zero in statfs when all OSCs are inactive
5147 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5148 lov_statfs divides by value returned by lov_get_stripecnt().
5152 Bugzilla : 9489, 3273
5153 Description: First write from each client to each OST was only 4kB in size,
5154 to initialize client writeback cache, which caused sub-optimal
5155 RPCs and poor layout on disk for the first written file.
5156 Details : Clients now request an initial cache grant at (re)connect time
5157 and so that they can start streaming writes to the cache right
5158 away and always do full-sized RPCs if there is enough data.
5159 If the OST is rebooted the client also re-establishes its grant
5160 so that client cached writes will be honoured under the grant.
5165 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5166 Details : Now I/O RPCs go to different portal number and (presumably) fast
5167 lock requests (and glimses) and other RPCs get their own service
5168 threads pool that should be able to service those RPCs
5171 Severity : enhancement
5173 Description: Ability to exchange lustre version between client and servers and
5174 issue warnings at client side if client is too old. Also for
5175 liblustre clients there is ability to refuse connection of too old
5177 Details : New 'version' field is added to connect data structure that is
5178 filled with version info. That info is later checked by server and
5182 Frequency : rare, liblustre only.
5183 Bugzilla : 9296, 9581
5184 Description: Two simultaneous writes from liblustre at offset within same page
5185 might proceed at the same time overwriting eachother with stale
5187 Details : I/O lock withing llu_file_prwv was released too early, before data
5188 actually was hitting the wire. Extended lock-holding time until
5189 server acknowledges receiving data.
5192 Frequency : extremely rare. Never observed in practice.
5194 Description: avoid generating lustre_handle cookie of 0.
5195 Details : class_handle_hash() generates handle cookies by incrementing
5196 global counter, and can hit 0 occasionaly (this is unlikely, but
5197 not impossible, because initial value of cookie counter is
5198 selected randonly). Value of 0 is used as a sentinel meaning
5199 "unassigned handle" --- avoid it. Also coalesce two critical
5200 sections in this function into one.
5202 Severity : enhancement
5204 Description: allow liblustre clients to delegate truncate locking to OST
5205 Details : To avoid overhead of locking, liblustre client instructs OST to
5206 take extent lock in ost_punch() on client's behalf. New connection
5207 flag is added to handle backward compatibility.
5209 Severity : enhancement
5210 Bugzilla : 4928, 7341, 9758
5211 Description: allow number of OST service threads to be specified
5212 Details : a module parameter allows the number of OST service threads
5213 to be specified via "options ost ost_num_threads={N}" in the
5214 OSS's /etc/modules.conf or /etc/modprobe.conf.
5218 Bugzilla : 6146, 9635, 9895
5219 Description: servers crash with bad pointer in target_handle_connect()
5220 Details : In rare cases when a client is reconnecting it was possible that
5221 the connection request was the last reference for that export.
5222 We would temporarily drop the export reference and get a new
5223 one, but this may have been the last reference and the export
5224 was just destroyed. Get new reference before dropping old one.
5226 Severity : enhancement
5227 Frequency : if client is started with failover MDS
5229 Description: Allow multiple MDS hostnames in the mount command
5230 Details : Try to read the configuration from all specified MDS
5231 hostnames during a client mount in case the "primary"
5234 Severity : enhancement
5236 Description: Stop sending data to evicted clients as soon as possible.
5237 Details : Check if the client we are about to send or are sending data to
5238 was evicted already. (Check is done every second of waiting,
5239 for which l_wait_event interface was extended to allow checking
5240 of exit condition at specified intervals).
5243 Frequency : rare, normally only when NFS exporting is done from client
5245 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5247 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5250 Frequency : always, if srand() is called before liblustre initialization
5252 Description: Liblustre uses system PRNG disturbing its usage by user application
5253 Details : Introduce internal to lustre fast and high-quality PRNG for
5254 lustre usage and make liblustre and some other places in generic
5255 lustre code to use it.
5257 Severity : enhancement
5258 Bugzilla : 9477, 9557, 9870
5259 Description: Verify that the MDS configuration logs are updated when xml is
5260 Details : Check if the .xml configuration logs are newer than the config
5261 logs stored on the MDS and report an error if this is the case.
5262 Request --write-conf, or allow starting with --old_conf.
5264 Severity : enhancement
5266 Description: Handle symlinks in the path when checking if Lustre is mounted.
5267 Details : Resolve intermediate symlinks when checking if a client has
5268 mounted a filesystem to avoid duplicate client mounts.
5273 Description: lconf can hit an error exception but still return success.
5274 Details : The lconf command catches the Command error exception at the top
5275 level script context and will exit with the associated exit
5276 status, but doesn't ensure that this exit status is non-zero.
5281 Description: failure of ptlrpc thread startup can cause oops
5282 Details : Starting a ptlrpc service thread can fail if there are a large
5283 number of threads or the server memory is very fragmented.
5284 Handle this without oopsing.
5287 Frequency : always, only if liblustre and non-default acceptor port was used
5289 Description: liblustre cannot connect to servers with non-default acceptor port
5290 Details : tcpnal_set_default_params() was not called and was therefore
5291 ignoring the environment varaible TCPNAL_PORT, as well as other
5292 TCPNAL_ environment variables
5297 Description: two objects could be created on the same OST for a single file
5298 Details : If an OST is down, in some cases it was possible to create two
5299 objects on a single OST for a single file. No problems other
5300 than potential performance impact and spurious error messages.
5304 Bugzilla : 5681, 9562
5305 Description: Client may oops in ll_unhash_aliases
5306 Details : Client dcache may become inconsistent in race condition.
5307 In some cases "getcwd" can fail if the current directory is
5313 Description: Inode refcounting problems in NFS export code
5314 Details : link_raw functions used to call d_instantiate without obtaining
5315 extra inode reference first.
5319 Bugzilla : 9942, 9903
5320 Description: Referencing freed requests leading to crash, memleaks with NFS.
5321 Details : We used to require that call to ll_revalidate_it was always
5322 followed by ll_lookup_it. Also with revalidate_special() it is
5323 possible to call ll_revalidate_it() twice for the same dentry
5324 even if first occurence returned success. This fix changes semantic
5325 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5326 reference on a request referred from the intent.
5327 ll_intent_release() then releases such a request.
5330 Frequency : rare, normally benchmark loads only
5332 Description: unlinked inodes were kept in memory on the client
5333 Details : If a client is repeatedly creating and unlinking files it
5334 can accumulate a lot of stale inodes in the inode slab cache.
5335 If there is no other client load running this can cause the
5336 client node to run out of memory. Instead flush old inodes
5337 from client cache that have the same inode number as a new inode.
5340 Frequency : SLES9 2.6.5 kernel and long filenames only
5341 Bugzilla : 9969, 10379
5342 Description: utime reports stale NFS file handle
5343 Details : SLES9 uses out-of-dentry names in some cases, which confused
5344 the lustre dentry revalidation. Change it to always use the
5348 Frequency : rare, unless heavy write-truncate concurrency is continuous
5349 Bugzilla : 4180, 6984, 7171, 9963, 9331
5350 Description: OST becomes very slow and/or deadlocked during object unlink
5351 Details : filter_destroy() was holding onto the parent directory lock
5352 while truncating+unlinking objects. For very large objects this
5353 may block other threads for a long time and slow overall OST
5354 responsiveness. It may also be possible to get a lock ordering
5355 deadlock in this case, or run out of journal credits because of
5356 the combined truncate+unlink. Solution is to do object truncate
5357 first in one transaction without parent lock, and then do the
5358 final unlink in a new transaction with the parent lock. This
5359 reduces the lock hold time dramatically.
5362 Frequency : rare, 2.4 kernels only
5364 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5365 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5366 before unmounting their filesystem in order to avoid deadlock.
5367 The kernel_locked() function in 2.4 kernels only checks whether
5368 the kernel lock is held, not whether it is this process that is
5369 holding it as 2.6 kernels do.
5374 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5375 Details : The client ptlrpc code may be trying to reconnect to a down
5376 server before a previous connection attempt has timed out.
5377 Increase the reconnect interval to be longer than the connection
5378 timeout interval to avoid sending duplicate connections to
5382 Frequency : echo_client brw_test command
5384 Description: fix echo_client to work with OST preallocated code
5385 Details : OST preallocation code (5137) didn't take echo_client IO path
5386 into account: echo_client calls filter methods outside of any
5387 OST thread and, hence, there is no per-thread preallocated
5388 pages and buffers to use. Solution: hijack pga pages for IO. As
5389 a byproduct, this avoids unnecessary data copying.
5393 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5394 Description: Client can oops in mdc_commit_close() after open replay
5395 Details : It was possible for the MDS to return an open request with no
5396 transaction number in mds_finish_transno() if the client was
5397 evicted, but without actually returning an error. Clients
5398 would later try to replay that open and may trip an assertion
5399 Simplify the client close codepath, and always return an error
5400 from the MDS in case the open is not successful.
5403 Frequency : rare, 2.6 OSTs only
5405 Description: OST may deadlock under high load on fragmented files
5406 Details : If there was a heavy load and highly-fragmented OST filesystems
5407 it was possible to have all the OST threads deadlock waiting on
5408 allocation of biovecs, because the biovecs were not released
5409 until the entire RPC IO was completed. Instead, release biovecs
5410 as soon as they are complete to ensure forward IO progress.
5412 Severity : enhancement
5414 Description: Support for specifying external journal device at mount
5415 Details : If an OST or MDS device is formatted with an external journal
5416 device, this device major/minor is stored in the ext3 superblock
5417 and may not be valid for failover. Allow detecting and
5418 specifying the external journal at mount time.
5423 Description: Mounting an MDS with pending unlinked files may cause oops
5424 Details : target_finish_recovery() calls mds_postrecov() which returned
5425 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5426 considers this an error and immediately begins cleaning up the
5427 lov, just after starting the mds_lov process
5429 Severity : enhancement
5431 Description: Implement 'lfs df' to report actual free space on per-OST basis
5432 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5433 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5434 usage of objects; '-h' to report in human readable format.
5436 ------------------------------------------------------------------------------
5438 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5445 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5446 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5447 at startup during orphan unlinking. Instead, skip buddy update
5448 before setup, as it will be regenerated after recovery is complete.
5453 Description: If an OST is inactive, its locks might reference stale inodes.
5454 Details : lov_change_cbdata() must iterate over all namespaces, even if
5455 they are inactive to clear inode references from the lock.
5457 Severity : enhancement
5458 Frequency : occasional, if non-standard max_dirty_mb used
5460 Description: Client will block write RPCs if not enough grant
5461 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5462 then the client will block writes while waiting for another RPC
5463 to complete instead of consuming its dirty limit. With change
5464 we get improved performance when max_dirty_mb is small.
5466 Severity : enhancement
5467 Bugzilla : 3389, 6253
5468 Description: Add support for supplementary groups on the MDS.
5469 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5470 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5471 lookups for user supplementary groups into a cache.
5475 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5476 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5477 for liblustre O_EXCL is enforced by MDS.
5481 Description: readdir never returns NULL in liblustre.
5482 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5483 dentry in struct dirent.
5487 Frequency : liblustre only, depends on application IO pattern
5488 Description: liblustre clients evicted if not contacting servers
5489 Details : Don't put liblustre clients into the ping_evictor list, so
5490 they will not be evicted by the pinger ever.
5492 Severity : enhancement
5494 Description: Add ability to evict clients by NID from MDS.
5495 Details : By echoing "nid:$NID" string into
5496 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5497 $NID would be instantly evicted from this MDS and from all active
5498 OSTs connected to it.
5502 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5503 Details : lookup_it_finish() used to query file size from OSTs that was not
5508 Description: service threads change working directory to that of init
5509 Details : Starting lustre service threads may pin the working directory
5510 of the parent thread, making that filesystem busy. Threads
5511 now change to the working directory of init to avoid this.
5515 Frequency : during shutdown only
5516 Description: shutdown with a failed MDS or OST can cause unmount to hang
5517 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5522 Frequency : 2.6 only
5523 Description: chmod/chown may include an extra supplementary group
5524 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5525 supplementary group and if none is specified this is used.
5528 Bugzilla : 5479 (6816)
5530 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5531 Details : If lookup is in progress on a file that is unlinked we might try
5532 to revalidate the inode and fail in revalidate after lookup is
5533 complete and ll_file_open() enqueues the open again but
5534 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5537 Frequency : always, if lconf --abort_recovery used
5539 Description: lconf --abort_recovery fails with 'Operation not supported'
5540 Details : lconf was attempting to abort recovery on the MDT device and not
5543 ------------------------------------------------------------------------------
5545 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5550 Frequency : rare (only unsupported configurations with a node running as an
5552 Bugzilla : 6514, 5137
5553 Description: Mounting a Lustre file system on a node running as an OST could
5555 Details : OSTs now preallocates memory needed to write out data at
5556 startup, instead of when needed, to avoid having to
5557 allocate memory in possibly low memory situations.
5558 Specifically, if the file system is mounted on on OST,
5559 memory pressure could force it to try to write out data,
5560 which it needed to allocate memory to do. Due to the low
5561 memory, it would be unable to do so and the node would
5562 become unresponsive.
5564 Severity : enhancement
5566 Description: Addition of lconf --service command line option
5567 Details : lconf now accepts a '--service <arg>' option, which is
5568 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5570 Severity : enhancement
5572 Description: Failover mode is now the default for OSTs.
5573 Details : By default, OSTs will now run in failover mode. To return to
5574 the old behaviour, add '--failout' to the lmc line for OSTs.
5576 Severity : enhancement
5578 Description: Health checks are now provided for MDS and OSTs
5579 Details : Additional detailed health check information on MSD and OSTs
5580 is now provided through the procfs health_check value.
5583 Frequency : occasional, depends on IO load
5585 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5586 after numerous create/delete cycles
5587 Details : The ext3 inode allocation policy would not allocate new inodes
5588 very well on the OSTs because there are no new directories
5589 being created. Instead we look for groups with free space if
5590 the parent directories are nearly full.
5595 Description: Network or server problems during mount may cause partially
5596 mounted clients instead of returning an error.
5597 Details : The config llog parsing code may overwrite the error return
5598 code during mount error handling, returning success instead
5604 Description: MDS can fail to allocate large reply buffers
5605 Details : After long uptimes the MDS can fail to allocate large reply
5606 buffers (e.g. zconf client mount config records) due to memory
5607 fragmentation or consumption by the buffer cache. Preallocate
5608 some large reply buffers so that these replies can be sent even
5609 under memory pressure.
5613 Frequency : rare (liblustre)
5614 Description: fsx running with liblustre complained that using truncate() to
5615 extend the file doesn't work. This patch corrects that issue.
5616 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5617 ATTR_SIZE and lsm use in llu_setattr_raw.
5621 Frequency : rare, only 2.6 kernels
5622 Description: Unusual file access patterns on the MDS may result in inode
5623 data being lost in very rare circumstances.
5624 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5625 code in the 2.6 kernel caused the inode and/or EA data not to
5626 be read from disk, causing single-file corruption.
5630 Frequency : rare, only 2.6 filesystems using extents
5631 Description: Heavy concurrent write and delete load may cause data corruption.
5632 Details : It was possible under high-load situations to have an extent
5633 metadata block in the block device cache from a just-unlinked
5634 file overwrite a newly-allocated data block. We now unmap any
5635 metadata buffers that alias just-allocated data blocks.
5639 Frequency : filesystems with default stripe_count larger than 77
5640 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5641 Details : lconf specifies an inode size of 4096 bytes when the default
5642 stripe_count is larger than 77. This conflicts with the default
5643 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5644 this case to avoid pinning too much memory for large EAs.
5646 ------------------------------------------------------------------------------
5648 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5653 Frequency : rare (extremely heavy IO load with hundreds of clients)
5655 Description: Client is evicted, gets IO error writing to file
5656 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5657 caused clients to be evicted for AST timeouts. The fixes in
5658 bug 5192 mean we no longer need to have such short AST timeouts
5659 so ldlm_timeout has been increased.
5662 Frequency : occasional during --force or --failover shutdown under load
5663 Bugzilla : 5949, 4834
5664 Description: Server oops/LBUG if stopped with --force or --failover under load
5665 Details : a collection of import/export refcount and cleanup ordering
5666 issues fixed for safer force cleanup
5669 Frequency : only filesystems larger than 120 OSTs
5670 Bugzilla : 5990, 6223
5671 Description: lfs getstripe would oops on a very large filesystem
5672 Details : lov_getconfig used kfree on vmalloc'd memory
5675 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5676 Bugzilla : 6242, 6243
5677 Description: reading from files that had been truncated to a non-zero size
5678 but never opened returned no data
5679 Details : ll_file_read() reads zeros from no-object files to EOF
5684 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5685 Details : There's little harm in aborting MDS/OSS recovery and letting it
5686 try again, so I removed the LASSERT and return an error instead.
5688 Severity : enhancement
5690 Description: New debugging infrastructure for tracking down data corruption
5691 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5692 (b) cover more of the client-side code path, and (c) try to narrow
5693 down where problems occurred
5697 Bugzilla : 3819, 4364, 4397, 6313
5698 Description: Racing close and eviction MDS could cause assertion in mds_close
5699 Details : It was possible to get multiple mfd references during close and
5700 client eviction, leading to one thread referencing a freed mfd.
5702 Severity: : enhancement
5703 Bugzilla : 3262, 6359
5704 Description: Attempts to reconnect to servers are now more aggressive.
5705 Details : This builds on the enhanced upcall-less recovery that was added
5706 in 1.4.2. When trying to reconnect to servers, clients will
5707 now try each server in the failover group every 10 seconds. By
5708 default, clients would previously try one server every 25 seconds.
5713 Description: After recovery, certain operations trigger a failed
5714 assertion on a client.
5715 Details : Failing over an mds, using lconf -d --failover, while a
5716 client was doing a readdir() call would cause the client to
5717 LBUG after recovery completed and the readdir() was resent.
5719 Severity : enhancement
5721 Description: Default groups are now added by lconf
5722 Details : You can now run lconf --group <servicename> without having to
5723 manually add groups with lmc.
5726 Frequency : occasional
5728 Description: Nodes with an elan id of 0 trigger a failed assertion
5731 Frequency : always when accessing e.g. tty/console device nodes
5733 Description: tty and some other devices nodes cannot be used on lustre
5734 Details : file's private_data field is used by device data and lustre
5735 values in there got lost. New field was added to struct file to
5736 store fs-specific private data.
5739 Frequency : when exporting Lustre via NFS
5741 Description: NFSD failed occasionally when looking up a path component
5742 Details : NFSD is looking up ".." which was broken in ext3 directories
5743 that had grown large enough to become hashed.
5746 Frequency : Clusters with multiple interfaces not on the same subnet
5748 Description: Nodes will repeatedly try to reconnect to an interface which it
5749 cannot reach and report an error to the log.
5750 Details : Extra peer list entries will be created by lconf with some peers
5751 unreachable. lconf now validates the peer before adding it.
5754 Frequency : Only if a default stripe is set on the filesystem root.
5756 Description: Setting a default stripe on the filesystem root prevented the
5757 filesystem from being remounted.
5758 Details : The client was sending extra request flags in the root getattr
5759 request and did not allocate a reply buffer for the dir EA.
5762 Frequency : occasional, higher if lots of files are accessed by one client
5763 Bugzilla : 6159, 6097
5764 Description: Client trips assertion regarding lsm mismatch/magic
5765 Details : While revalidating inodes the VFS looks up inodes with ifind()
5766 and in rare cases can find an inode that is being freed.
5767 The ll_test_inode() code will free the lsm during ifind()
5768 when it finds an existing inode and then the VFS later attaches
5769 this free lsm to a new inode.
5773 Bugzilla : 6422, 7030
5774 Description: MDS deadlock between mkdir and client eviction
5775 Details : Creating a new file via mkdir or mknod (starting a transaction
5776 and getting the ns lock) can deadlock with client eviction
5777 (gets ns lock and trying to finish a synchronous transaction).
5780 Frequency : occasional
5781 Description: While starting a server, the fsfilt_ext3 module could not be
5783 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5784 kernels. Previously, lconf would still use the ext3 name
5785 when trying to load modules. Now, it will correctly use
5786 ext3 on 2.4 and ldiskfs on 2.6.
5788 Severity : enhancement
5789 Description: The default stripe count has been changed to 1
5790 Details : The interpretation of the default stripe count (0, to lfs
5791 or lmc) has been changed to mean striping across a single
5792 OST, rather than all available. For general usage we have
5793 found a stripe count of 1 or 2 works best.
5795 Severity : enhancement
5796 Description: Add support for compiling against Cray portals.
5797 Details : Conditional compiling for some areas that are different
5801 Frequency : occasional
5802 Bugzilla : 6409, 6834
5803 Description: Creating files with an explicit stripe count may lead to
5804 a failed assertion on the MDS
5805 Details : If some OSTs are full or unavailable, creating files may
5806 trigger a failed assertion on the MDS. Now, Lustre will
5807 try to use other servers or return an error to the
5811 Frequency : occasional
5813 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5814 caused lock timeout during readahead (since 1.4.2).
5815 Details : Processes doing readahead might match a lock that hasn't been
5816 granted yet if there are overlapping and conflicting lock
5817 requests. The readahead process waits on ungranted lock
5818 (original lock is CBPENDING), while OST waits for that process
5819 to cancel CBPENDING read lock and eventually evicts client.
5821 Severity : enhancement
5823 Description: Initial enabling of flock support for clients
5824 Details : Implements fcntl advisory locking and file status functions.
5825 This feature is provided as an optional mount flag (default
5826 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5827 locking are implemented yet, and those that are are not guaranteed
5828 to be completely correct in production environments.
5829 mount -t lustre -o [flock|noflock] ...
5832 Frequency : occasional
5834 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5835 assertion in the ext3 JBD (journaling) layer.
5836 Details : The b_committed_data struct is protected by the big kernel lock
5837 in 2.4 kernels, serializing journal_commit_transaction() and
5838 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5839 access to this struct. In 2.6 kernels there is finer grained
5840 locking to improve SMP performance of the JBD layer.
5844 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5849 Description: merge and cleanup kernel patches.
5852 -----------------------------------------------------------------------------
5854 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5856 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5857 versions of Lustre. Please update all servers and clients to
5858 version 1.4.2 or later at the same time. You must also run
5859 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5860 to update the configuration logs.
5862 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5863 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5864 - swab llog records of type '0' so we get proper header size/idx (5861)
5865 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5866 - fix rename of one directory over another leaking an inode (5953)
5867 - avoid SetPageDirty on 2.6 (5981)
5868 - don't re-add just-being-destroyed locks to the waiting list (5653)
5869 - when creating new directories, inherit the parent's custom
5870 striping settings if present parent (3048)
5871 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5872 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5873 - don't hold client locks on temporary worklist from l_lru (5666)
5874 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5875 - automatically evict dead clients (5921)
5876 - Update file size properly in create+truncate+fstat case (6196)
5877 - Do not unhash mountpoint dentries, do not allow removal of
5879 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5880 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5881 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5882 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5883 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5884 - Correctly handle reads of files with no objects (6243)
5885 - lctl recover will also mark a device active if deactivate used (5933)
5887 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5888 - allow --write-conf on an MDS with different nettype than client (5619)
5889 - don't write config llogs to MDS for mounts not from that MDS (5617)
5890 - lconf should create multiple TCP connections from a client (5201)
5891 - init scripts are now turned off by default; run chkconfig --on
5892 lustre and chkconfig --on lustrefs to use them
5893 - upcalls are no longer needed for clients to recover to failover
5895 - add --abort-recovery option to lconf to abort recovery on device
5897 - add support for an arbitrary number of OSTs (3026)
5898 - Quota support protocol changes.
5899 - forward compatibility changes to wire structs (6007)
5900 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5901 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5902 - improved "lustre status" script
5903 - initialize blocksize for non-regular files (6062)
5904 - added --disable-server and --disable-client configure options (5782)
5905 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5906 - Vanilla 2.4.29 support
5907 - increase maximum number of obd devices to 520 (6242)
5908 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5909 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5910 SLES 9 2.6.5 kernels
5911 - limit stripes per file to 160 (the maximum EA size) (6093)
5913 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5916 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5917 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5918 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5919 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5920 - small ext3 extents cleanups and fixes (5733)
5921 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5922 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5923 - avoid needless client->OST connect, fix handle mismatch (5317)
5924 - fix DLM error path that led to out-of-sync client, long delays (5779)
5925 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5926 - fix several locking issues related to i_size (5492,5624,5654,5672)
5927 - don't move pending lock onto export if it is already evicted (5683)
5928 - fix kernel oops when creating .foo in unlinked directory (5548)
5929 - fix deadlock in obdfilter statistics vs. object create (5811)
5930 - use time_{before,after} to avoid timer jiffies wrap (5882)
5931 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5932 - Do not leak request if server was not able to process it (5154)
5933 - If mds_open unable to find parent dir, make that negative lookup(5154)
5934 - don't create new directories with extent-mapping (5909, 5936)
5936 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5937 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5938 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5939 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5940 - support for RHEL4 kernel on i686 (5773)
5941 - provide error messages when incompatible logs are encountered (5898)
5943 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5944 * version 1.4.0.10 (1.4.1 release candidate 1)
5946 - don't keep a lock reference when lock is not granted (4238)
5947 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5948 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5949 - drop import inflight refcount on signal_completed_replay error (5255)
5950 - unlock page after async write error during send (3677)
5951 - handle missing objects in filter_preprw_read properly (5265)
5952 - no transno return for symlink open, don't save no-trasno open (3440)
5953 - don't try to complete elan receive that already failed (4012)
5954 - free RPC server reply state on error (5406)
5955 - clean up thread from ptlrpc_start_thread() on error (5160)
5956 - readahead could read extra page into cache that wasn't ejected (5388)
5957 - prevent races in class_attach/setup/cleanup/detach (5260)
5958 - don't dereference de->d_inode after l_dput of de (5458)
5959 - use "int" for stripe value returned from lock_to_stripe (5544)
5960 - mballoc allocation and error-checking fixes in 2.6 (5504)
5961 - block device patches to fix I/O request sizes in 2.6 (5482)
5962 - look up hostnames for IB nals (5602)
5963 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5964 - don't start multiple acceptors for the same port (5277)
5965 - fix incorrect LASSERT in mds_getattr_name (5635)
5966 - export a proc file for general "ping" checking (5628)
5967 - fix "lfs check" to not block when the MDS is down (5628)
5969 - service request history (4965)
5970 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5971 - create an "evict_client" /proc entry on OSTs, like the MDS has
5972 - fix mount usage message, return errors per mount(8) (5168)
5973 - change grep [] to grep "[]" in tests so they work in more UMLs
5974 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5975 - remove ancient LOV_MAGIC_V0 EA support (5047)
5976 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5977 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5978 - allow readahead allocations to fail when low on memory (5383)
5979 - mmap locking landed again, after considerable improvement (2828)
5980 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5982 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5985 - send OST transaction number in read/write reply to free req (4966)
5986 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5987 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5988 - the watchdog thread now runs as interruptible (5246)
5989 - flock/lockf fixes (but it's still disabled, pending 5135)
5990 - don't use EXT3 constants in llite code (5094)
5991 - memory shortage at startup could cause assertion (5176)
5993 - reorganization of lov code
5994 - single portals codebase
5996 - add extents/mballoc support (5025)
5997 - direct I/O reads in the obdfilter (4048)
5998 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6000 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6003 - send OST transaction number in read/write reply to free req (4966)
6004 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6005 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6006 - the watchdog thread now runs as interruptible (5246)
6007 - handle missing objects in filter_preprw_read properly (5265)
6008 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6009 - drop import inflight refcount on signal_completed_replay error (5255)
6010 - unlock page after async write error during send (3677)
6011 - return original error code on reconstructed replies (3761)
6012 - no transno return for symlink open, don't save no-trasno open (3440)
6014 - add pid to ldlm debugging output (4922)
6015 - bump the watchdog timeouts -- we can't handle 30sec yet
6016 - extra debugging for orphan dentry/inode bug (5259)
6018 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6021 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6022 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6023 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6024 - fix NULL dereference in /proc/sys/portals/routes (4827)
6025 - allow failed mdc_close() operations to be interrupted (4561)
6026 - stop precreate on OST before MDS would time out on it (4778)
6027 - don't send partial-page writes before EOF from client (4410)
6028 - discard client grant for sub-page writes on large-page clients (4520)
6029 - don't free dentries not owned by NFS code, check generation (4806)
6030 - fix lsm leak if mds_create_objects() fails (4801)
6031 - limit debug_daemon file size, always print CERROR messages (4789)
6032 - use transno after validating reply (3892)
6033 - process timed out requests if import state changes (3754)
6034 - update mtime on OST during writes, return in glimpse (4829)
6035 - add mkfsoptions to LDAP (4679)
6036 - use ->max_readahead method instead of zapping global ra (5039)
6037 - don't interrupt __l_wait_event() during strace
6039 - add software watchdogs to catch hung threads quickly (4941)
6040 - make lustrefs init script start after nfs is mounted
6041 - fix CWARN/ERROR duplication (4930)
6042 - return async write errors to application if possible (2248)
6043 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6044 - print NAL number in %x format (4645)
6045 - update barely-supported suse-2.4.21-171 series (4842)
6046 - support for sles 9 %post scripts
6047 - support for building 2.6 kernel-source packages
6048 - support for sles km_* packages
6050 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6053 - ignore -ENOENT errors in osc_destroy (3639)
6054 - notify osc create thread that OSC is being cleaned up (4600)
6055 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6056 - reconstruct ost_handle() like mds_handle() (4657)
6057 - create a new thread to do import eviction to avoid deadlock (3969)
6058 - let lconf resolve symlinked-to devices (4629)
6059 - don't unlink "objects" from directory with default EA (4554)
6060 - hold socknal file ref over connect in case target is down (4394)
6061 - allow more than 32000 subdirectories in a single directory (3244)
6062 - fix blocks count for O_DIRECT writes (3751)
6063 - OST returns ENOSPC from object create when no space left (4539)
6064 - don't send truncate RPC if file size isn't changing (4410)
6065 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6066 - bind to privileged port in socknal and tcpnal (3689)
6068 - rate limit CERROR/CWARN console message to avoid overload (4519)
6069 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6070 - basic mmap support (3918)
6071 - kernel patch series update from b1_4 (4711)
6073 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6076 - avoid crash during MDS cleanup with OST shut down (2775)
6077 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6078 - don't use bad inodes on the MDS (3744)
6079 - dynamic object preallocation to improve recovery speed (4236)
6080 - don't hold spinlock over lock dumping or change debug flags (4401)
6081 - don't zero obd_dev when it is force cleaned (3651)
6082 - print grants to console if they go negative (4431)
6083 - "lctl deactivate" will stop automatic recovery attempts (3406)
6084 - look for existing locks in ldlm_handle_enqueue() (3764)
6085 - don't resolve lock handle twice in recovery avoiding race (4401)
6086 - revalidate should check working dir is a directory (4134)
6088 - don't always mark "slow" obdfilter messages as errors (4418)
6090 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6093 - don't close LustreDB during write_conf until it is done (3860)
6094 - fix typo in lconf for_each_profile (3821)
6095 - allow dumping logs from multiple threads at one time (3820)
6096 - don't allow multiple threads in OSC recovery (3812)
6097 - fix debug_size parameters (3864)
6098 - fix mds_postrecov to initialize import for llog ctxt (3121)
6099 - replace config semaphore with spinlock (3306)
6100 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6101 - don't allow enqueue to complete on a destroyed export (3822)
6102 - down write_lock before checking llog header bitmap (3825)
6103 - recover from lock replay timeout (3764)
6104 - up llog sem before sending rpc (3652)
6105 - reduce ns lock hold times when setting kms (3267)
6106 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6107 - fix NULL deref and obd_dev leak on setup error (3312)
6108 - replace some LBUG about llog ops with error handling (3841)
6109 - don't match INVALID dentries from d_lookup and spin (3784)
6110 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6111 - fix invalid assertion in ptlrpc_set_wait (3880)
6113 - add libwrap support for the TCP acceptor (3996)
6114 - add /proc/sys/portals/routes for non-root route listing (3994)
6115 - allow setting MDS UUID in .xml (2580)
6116 - print the stack of a process that LBUGs (4228)
6118 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6121 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6122 - make sure to unset replay flag from failed open requests (3440)
6123 - if default stripe count is 0, use OST count for inode size (3636)
6124 - update parent mtime/ctime on client for create/unlink (2611)
6125 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6126 - free recovery state on server during a forced cleanup (3571)
6127 - unregister_reply for resent reqs (3063)
6128 - loop back devices mounting and status check on 2.6 (3563)
6129 - fix resource-creation race that can provoke i_size == 0 (3513)
6130 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6131 - more debugging for page-accounting assertion (3746)
6132 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6133 - avoid deadlock after precreation failure (3758)
6134 - fix race and lock order deadlock in orphan handling (3450, 3750)
6135 - add validity checks when grabbing inodes from l_ast_data (3599)
6137 - add /proc/.../recovery_status to obdfilter (3428)
6138 - lightweight CDEBUG infrastructure, debug daemon (3668)
6139 - change default OSC RPC parameters to be better on small clusters
6140 - turn off OST read cache for files smaller than 32MB
6141 - install man pages and include them in rpms (3100)
6142 - add new init script for (un)mounting lustre filesystems (2593)
6143 - run chkconfig in %post for init scripts (3701)
6144 - drop scimac NAL (unmaintained)
6146 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6149 - clean kiobufs before and after use (3485)
6150 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6151 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6152 - change init script to fail more clearly if not run as root (1528)
6153 - allow clients to reconnect during replay (1742)
6154 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6155 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6156 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6157 - fix cond_resched (3554)
6158 - only evict unfinished clients after recovery (3515)
6159 - allow bulk resend, prevent data loss (3570)
6160 - dynamic ptlrpc request buffer allocation (2102)
6161 - don't allow unlinking open directory if it isn't empty (2904)
6162 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6163 - remove extraneous obd dereference causing LASSERT failure (3334)
6164 - don't use get_cycles() when creating temp. files on the mds (3156)
6165 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6166 - handle EEXIST for set-stripe, set proper directory name (3336)
6168 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6169 - fix ksocknal_fmb_callback() error messages (2918)
6171 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6174 - don't copy lvb into (possibly NULL) reply on error (2983)
6175 - don't deref dentry after dput, don't free lvb on error (2922)
6176 - use the kms to determine writeback rpc length (2947)
6177 - increment oti_logcookies when osc is inactive (2948)
6178 - update client's i_blocks count via lvb messages (2543)
6179 - handle intent open/close of special files properly (1557)
6180 - mount MDS with errors=remount-ro, like obdfilter (2009)
6181 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6182 - don't use cancelling-locks' kms values (2947)
6183 - use highest lock extent for kms, not last one (2925)
6184 - don't dereference ERR_PTR() dentry in error handling path (3107)
6185 - fix thread race in portals_debug_dumplog() (3122)
6186 - create lprocfs device entries at setup instead of at attach (1519)
6187 - common AST error handler, don't evict client on completion race (3145)
6188 - zero nameidata in detach_mnt in 2.6 (3118)
6189 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6190 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6191 - initialize RPC timeout timer earlier for 2.6 (3219)
6192 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6193 - print nal/nid for unknown nid (3258)
6194 - additional checks for oscc recovery before doing precreate (3284)
6195 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6196 - don't crash in mdc_close for bad permissions on open (3285)
6197 - zero i_rdev for non-device files (3147)
6198 - clear page->private before handing to FS, better assertion (3119)
6199 - tune the read pipeline (3236)
6200 - fix incorrect decref of invalidated dentry (2350)
6201 - provide read-ahead stats and refine rpc in flight stats (3328)
6202 - don't hold journal transaction open across create RPC (3313)
6203 - update atime on MDS at close time (3265)
6204 - close LDAP connection when recovering to avoid server load (3315)
6205 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6206 - don't leak open file on MDS after open resend (3325)
6207 - serialize filter_precreate and filter_destroy_precreated (3329)
6208 - loop device shouldn't call sync_dev() for nul device (3092)
6209 - clear page cache after eviction (2766)
6210 - resynchronize MDS->OST in background (2824)
6211 - refuse to mount the same filesystem twice on same mountpoint (3394)
6212 - allow llmount to create routes for mounting behind routers (3320)
6213 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6214 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6215 - fix rare problem with rename on htree directories (3417)
6217 - allow default OST striping configuration per directory (1414)
6218 - fix compilation for qswnal for 2.6 kernels (3125)
6219 - increase maximum number of MDS request buffers for large systems
6220 - change liblustreapi to be useful for external progs like lfsck (3098)
6221 - increase local configuration timeout for slow disks (3353)
6222 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6224 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6227 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6228 - don't overwrite extent policy data in reply if lock was blocked (2901)
6229 - drop filter export grants atomically with removal from device (2663)
6230 - del obd_self_export from work_list in class_disconnect_exports (2908)
6231 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6232 - swab reply message in mdc_close, other PPC fixes (2464)
6233 - fix destroying of named logs (2325)
6234 - overwrite old logs when running lconf --write_conf (2264)
6235 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6236 - fix race in target_handle_connect (2898)
6237 - mds_reint_create() should take same inode create lock (2926)
6238 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6239 - don't close files for self_export to avoid uninitialized obd (2936)
6240 - allow MDS with the same name as client node (2939)
6241 - hold dentry reference for closed log files for unlink (2325)
6242 - reserve space for all logs during transactions (2059)
6243 - don't evict page beyond end of stripe extent (2925)
6244 - don't oops on a deleted current working directory (2399)
6245 - handle hard links to targets without a parent properly (2517)
6246 - don't dereference NULL lock when racing during eviction (2867)
6247 - don't grow lock extents when lots of conflicting locks (2919)
6249 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6252 - account for cache space usage on clients to avoid data loss (974)
6253 - lfsck support in lustre kernel code (2349)
6254 - reduce journal credits needed for BRW writes (2370)
6255 - orphan handling to avoid losing space on client/server crashes
6256 - ptlrpcd can be blocked, stopping ALL progress (2477)
6257 - use lock value blocks to assist in proper KMS, faster stat (1021)
6258 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6259 - recovery for initial connections (2355)
6260 - fixes for mds_cleanup_orphans (1934)
6261 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6262 - block all file creations until orphan recovery completes (1901)
6263 - client remove rq_connection from request struct (2423)
6264 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6265 - recovery timer race (2670)
6266 - mdc_close recovey bug (2532)
6267 - ptlrpc cleanup bug (2710)
6268 - mds timeout on local locks (2588)
6269 - namespace lock held during RPCs (2431)
6270 - handle interrupted sync write properly (2503)
6271 - don't try to handle a message that hasn't been replied to (2699)
6272 - client assert failure during cleanup after abort recovery (2701)
6273 - leak mdc device after failed mount (2712)
6274 - ptlrpc_check_set allows timedout requests to complete (2714)
6275 - wait for inflight reqs when ptlrpcd finishes (2710)
6276 - make sure unregistered services are removed from the srv_list
6277 - reset bulk XID's when resending them (caught by 1138 test)
6278 - unregister_bulk after timeout
6279 - fix lconf error (2694)
6280 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6281 - readahead locks pages, leaves pending causing memory pressure (2673)
6282 - increase OST request buffers to 4096 on large machines (2729)
6283 - fix up permission of existing directories in simple_mkdir (2661)
6284 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6285 - don't assign transno to errored transactions (2742)
6286 - don't delete objects on OST if given a bogus objid from MDS (2751)
6287 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6288 - if rq_no_resend, then timeout request after recovery (2432)
6289 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6290 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6291 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6292 - MDS recovery shouldn't delete objects that it has given out (2730)
6293 - if enqueue arrives after completion, don't clobber LVB (2819)
6294 - don't unlock pages twice when trigger_group_io returns error (2814)
6295 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6296 - don't write pages to disk if there was an error (1450)
6297 - don't ping imports that have recovery disabled (2676)
6298 - take buffered bytes into account when balancing socknal conn (2817)
6299 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6300 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6301 - remove little-endian swabbing of llog records (1987)
6302 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6303 - retry reposting request buffers if they fail (1191)
6304 - grow extent at grant time to avoid granting a revoked lock (2809)
6305 - lock revoke doesn't evict page if covered by a second lock (2765)
6306 - disable VM readahead to avoid reading outside lock extents (2805)
6308 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6309 - updated kernel patches for hp-2.4.20 kernel (2681)
6311 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6314 - fix truncated write corruption (2366)
6315 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6317 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6318 - ensure len(uuid) < 37 in lmc (1171)
6319 - fix ia64 OOPS in llog_test (2255)
6320 - zero end of page at obdfilter for partial page writes (2648)
6321 - don't leave stale dentries around after renames (bug 2428)
6322 - fix timeouts when evicting a client with a single lock held (2642)
6323 - set deadline for the initial HELLO message to drain (2634)
6324 - print out dotted-quad IP addresses in the socknal (2302)
6326 - additional debugging for MDS client eviction problem (2443)
6327 - fix mkfsoptions support for osts (2603, 2604)
6329 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6332 - add series for the vanilla 2.6.0 kernel
6333 - add series for the vanilla 2.4.24 kernel
6334 - add series for a cray x86/64 UL kernel drop
6335 - fix xattr patches for the vanilla 2.4.19 series
6337 - generate true UUIDs in lmc (1171)
6338 - have portals stack dumping break in UML (2466)
6339 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6340 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6341 - wake callback waiting threads on client eviction (2460)
6342 - Add --ptldebug and --subsystem to lmc (1719)
6343 - update assertion to allow safe interrupt allocation
6344 - set rq_no_resend for cancel requests (2432)
6345 - recalculate ptlrpcd timeout after resend (2494)
6346 - call vfs_rmdir when removing pending directories (2368)
6347 - fix renaming a file to itself (2429)
6348 - lmc creates a default one-stripe lov (2454)
6349 - expand procfs space to handle large clusters (2326)
6350 - increase UML stack to avoid overflow
6351 - update lconf's list of debug and subsystem masks
6352 - fix lfs find --obd (2510)
6353 - /proc tunable for disabling filter read caching (2591)
6354 - stop rpm packages from altering slapd.conf (2301)
6355 - disable nagle in the socknal under 0conf (2578)
6356 - choose mds inode size based on stripe count (2572)
6357 - fix kernel-source rpm problems (2516)
6359 - add --disable-doc to avoid pdf generation (2421)
6360 - update documentation, tests, type-os, comments
6361 - avoid format warnings on ia64
6362 - remove the TOE NAL
6363 - tiny code cleanups by removing unused fields
6365 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6368 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6369 - remove the most blatant lies from BUILDING file (2371)
6370 - change default debug level to reasonable production setting
6371 - reduce client side cache size to reduce cache flush time
6372 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6373 - make TCP zerocopy and pinger support enabled by default (2476)
6374 - sync writes completed after process exits caused crashes (2319)
6375 - maintain correct mount count on the MDS (2356)
6376 - backout 1557, because 2316 wasn't really fixed
6377 - better file I/O statistics gathering in /proc
6378 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6379 - another kernel patch to fix zero-copy TCP function export
6380 - don't take duplicate lock when processing re-sent getattr (2420)
6381 - lctl uses obd_self_export instead of creating new conn (2353)
6382 - MDS/OST recovery case which requires object creation asserted (2425)
6383 - move lfs from /usr/sbin to /usr/bin in packages
6384 - fix race between mds_client_add and mds_client_free (2417)
6385 - use kmalloc instead of slabs in portals (2430)
6386 - don't create duplicate records when a failover MDS is present (2442)
6387 - remove unnecessary mount age check (2332)
6388 - don't remove directory inodes from locks prematurely (2451)
6389 - don't break if MDS service name is the same as hostname (2103)
6390 - fix races in client write RPC generation when cache full (2482)
6392 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6395 - remove now-unused request->rq_obd (278)
6396 - if an allocation fails, print out how much memory we've used (1933)
6397 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6398 - add the "configurable stack size" patch to most series files (1256)
6399 - ability to write large log records, for 100+ OST configs (2306)
6400 - fix NULL deref when filter_prep fails (2314)
6401 - fix operator precedence error in filter_sync
6402 - dynamic allocation of socknal TX descriptors (2315)
6403 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6404 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6405 - ensure that CWARN messages in Portals always get to the syslog
6406 - __init/__exit are not for prototype decls (ldlm_init/exit)
6407 - x86-64 compile warning fixes
6408 - fix gateway LMC keyword conflict (2318)
6409 - fix MDS lock inversions in getattr/reint paths (1844)
6410 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6411 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6412 - DLM race condition prevented some lock evictions (2328)
6413 - ENOMEM detection and retry on socknal sends (2230)
6414 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6415 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6416 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6417 - better error messages when a client is rejected during recovery (1505)
6418 - avoid cancelling locks which were never granted, after failure (2330)
6419 - fix i_sem/journal inversion in mds_client_add (2333)
6420 - fix truncate/getattr lock cycle deadlock (2334)
6421 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6422 - fix two rare exit paths which could leak an l_lock() ref (2321)
6423 - fencepost error in MDS/OST orphan recovery (2226)
6424 - make log record alignment 8 bytes (1988)
6425 - lstripe now fails when requested offset > ost_count (2237)
6426 - ensure that all kernel series have a complete list.h (1607)
6427 - fix crashes in special-file operations (2316)
6428 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6430 - allow configurable automake binary, for testing new versions
6431 - small update to the lfs documentation
6433 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6435 * fix negative export reference count in fsfilt_sync (2312)
6437 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6438 * release candidate 0.9.1
6440 - orphans are moved into the PENDING directory for possible recovery
6441 - replayed opens now open by fid for orphan/rename safety (1042)
6442 - last close of an orphan inode generates a transno (683)
6443 - chdir() and mount() now pin the directory entry (1020)
6444 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6445 - discard very old requests without processing them (1502)
6446 - remove some common, well-understood CERRORs (1505)
6447 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6448 - clear "grant" flags in OST replies until OST grant code lands (1644)
6449 - fix read performance by not clobbering i_blksize on client (1598)
6450 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6451 - make LRU size a /proc tunable, clears locks when reduced (707)
6452 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6453 - server locks take references on exports now (1558)
6454 - build fixes for 2.4.20-rh trees (1663)
6455 - return an error from lov_create if all OSCs are inactive (1751)
6456 - fix import levels when a reconnect happens without a timeout (1597)
6457 - exit early from mds_open if we get a lookup error (1749)
6458 - partial page read at EOF wouldn't wait for disk before sending (1642)
6459 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6460 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6461 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6462 - write/truncate lock inversion (1639)
6463 - Don't auto-load obdclass, portals modules during cleanup (1495)
6464 - fix timestamps from jumping to "now" (1763)
6465 - extra journal assertions (1648)
6466 - add an extra multiunlink test (1771)
6467 - fix read_record/write_record API (1776)
6468 - fix leak of offset_extent, possible incorrect i_size later (1772)
6469 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6470 - Debugging for the kqswnal_get_idle_tx problems (1820)
6471 - Allow recovery to be attempted multiple times (1536)
6472 - Write out MDS last_rcvd file after it is first created (1600)
6473 - Fix tx_descriptor leak in failed transmit situations (1827)
6474 - ext3 journaling fixes for assertion failure after IO error (1871)
6475 - class_export_put() on freed export after completion AST error (1896)
6476 - Fix revalidate looping in VFS (1322)
6477 - Don't access a freed export during MDS_REINT timeout (1521)
6478 - Add open-unlink recovery support on the MDS (1673,1764)
6479 - Return an error if no MDS data was read from last_rcvd (1946)
6480 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6481 - Don't setup a disk device that doesn't match exported UUID (317)
6482 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6483 - avoid committing NULL handle in force close
6484 - local.sh is now a one-stripe LOV configuration
6485 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6486 - fix ext3 htree duplicate directory entry corruption (1516)
6487 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6488 - update to new LOV EA format (2097)
6489 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6491 2003-06-15 Phil Schwan <phil@clusterfs.com>
6494 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6495 - per-import/export recovery handling (958, 931, 959)
6496 - multiple last-rcvd slots, for serving multiple FSes (949)
6497 - connections are again shared between multiple imp/exports (963, 964)
6498 - "umount -f" would hang if any requests needed to be sent (393, 978)
6499 - avoid pinning large req buffer by copying for queued messages (989)
6500 - add "uuid" to "lctl device" command to help upcalls (991)
6501 - "open" RPCs with transnos would confuse recovery counters (1037)
6502 - do proper endian conversion of all wire messages (288, 340, 891)
6503 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6504 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6505 - fix ptlrpc_connection leak in target_handle_connect (1174)
6506 - fix import refcounting bug in OST and MDS cleanup (1134)
6507 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6508 - fix very unlikely obd_types race condition (501)
6509 - remove osc_open hack for echo_client (1187)
6510 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6511 - a failure in read_inode2 leads to deadlock (1139)
6512 - cancel ack-locks as soon as transaction is committed (1072)
6513 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6514 - make sure to commitrw after any preprw to avoid deadlock (1162)
6515 - failing to execute a file in a lustre FS would lock inode (1203)
6516 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6517 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6518 - obd_uuid2tgt would walk off the end of the list (1255)
6519 - on IA64 the osc would give portals incorrect bulk size (1258)
6520 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6521 - fix lock inversion caused by new llite matching code (1282)
6522 - limit the number of dirty pages on a client to 10MB (1286)
6523 - timed out locks were not being corrected cancelled (1289)
6524 - fix O_DIRECT above 4GB on IA-32 (1292)
6525 * major user-visible changes
6526 - fail out/fail over policy now controlled by the upcall (993)
6528 - add OBD_PING to check server availability and failure (954)
6529 - lustre messages are now sent in sending host order (288, 340, 891)
6530 - add eadatalen to MDS getattr reply (340)
6531 - OST read replies may contain second buffer, with per-page status (593)
6533 2003-03-11 Phil Schwan <phil@clusterfs.com>
6536 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6537 - failed open()s could cause deadlock; fixed (867, 869)
6538 - stop cancelling OST locks when files are closed (481)
6539 - overlapping XID spaces caused network corruption (851, 853)
6540 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6541 - setattr_raw added, to do single-RPC, server-side setattrs
6542 - lmc/lconf syntax change for OST UUIDs
6543 - fix crashy race condition between ptlrpc_free_req and osc_close
6544 - don't use request in mdc_enqueue if we hit a timeout (889)
6545 - don't set the inode i_size for regular files from the MDS (896)
6546 - handle out of order completion AST (842)
6547 - don't LBUG if a lock request times out after receiving AST (913)
6548 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6549 - if a bad lock AST arrives, send an error instead of dropping entirely
6550 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6551 - fix leak in bulk IO when only partially completed (899, 900, 926)
6552 - fix O_DIRECT for ia64 (55)
6553 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6554 - C-z after timeout could hang a process forever; fixed (977)
6556 - client-side I/O cache (678, 924, 929, 941, 970)
6558 - READPAGE and SETATTRs which don't take server-side locks get
6561 2003-02-11 Phil Schwan <phil@clusterfs.com>
6564 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6565 - Fix fsfilt_extN_readpage() to read a full page of directory
6566 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6567 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6568 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6569 - Most intent operations were removed, in favour of a new RPC mode
6570 that does a single RPC to the server and bypasses most of the VFS
6571 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6572 - Aggressively cancel local locks on DLM servers
6573 - mds_reint_unlink sends EA to the client if it's the last nlink.
6574 client uses that EA to unlink OST objects.
6575 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6576 - recursive symlinks were fixed (439)
6577 - fixed NULL deref in DEBUG_REQ
6578 - filter_update_lastobjid no longer calls sync, which annoyed extN
6579 - fixed multi-client small-writes to a single file problem (445)
6580 - fixed mtime updates during file writes (607)
6581 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6582 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6583 - fixed Linux OST size reporting problem (444, 656)
6584 - OST now updates object mtime with writes or setattr (607, 619)
6585 - client verifies file size before zeroing page past EOF (445)
6586 - OST now writes last allocated objid to disk with allocation (108)
6587 - LOV on echo now works (409)
6589 - mds_reint_unlink sends a new buffer, with the EA included. this
6590 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6591 set if a regular file was being unlinked, and it was the last link
6592 - use PtlGet from the target for bulk writes (315)
6593 - OST now updates object mtime with writes or setattr (607, 619)
6594 - LDLM now has a grant-time callback to revalidate locked items, if
6596 - Many MDS operations were reorganized to combat race conditions
6598 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6599 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6601 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6604 - Fully reactivate OST imports after reconnection (512, others)
6605 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6606 - More graceful error handling for truncating on dead OST (515)
6607 - Don't error out unless we're actually accessing dead stripes (474)
6608 - Fix garbage sizes when stripes are missing (410)
6609 - LRU counters were broken, causing constant lock purge (433, 432)
6610 - garbage on read from stripes with failed OSTs (441)
6611 - mark OSCs as active before reconnecting during recovery (438)
6612 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6613 - lfind did not preserve OST order in output (443)
6614 - symlinks cause hung clients, incorrect data (439)
6615 - stop dereferencing request after dropping refcount (457)
6616 - don't LASSERT(spin_is_locked) on non-SMP (455)
6617 - fixes for many rename() bugs
6618 - fstat didn't correctly synchronize attributes (399)
6619 - server must handle lock cancellation during blocking AST prep (487)
6620 - bulk descriptors were free()d too soon (511)
6621 - fix paths in lconf, which would load incorrect modules (451, 507)
6622 - fix confusing lconf 'host not found' error message (386)
6623 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6624 - fix race condition in mdc_blocking_ast() for inode access (526)
6625 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6626 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6627 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6628 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6629 - fix lstripe in several ways (kernel side) (527)
6630 - fix request leak in ldlm_cli_enqueue (262)
6631 - incorrect OSC was marked inactive after OST failure
6632 - call mds_fs_cleanup before unmounting filesystem (524)
6633 - fix races between taking ns_lock and ldlm_lock_change_resource
6634 - fix races updating LOV export open file list
6635 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6636 - fix recovery NULL deref in ldlm_cli_cancel_unused
6637 - fix some DLM races by using new hash table for lock handles (419)
6638 - permit the client to specify desired inodes, at replay
6639 - duplicate requests when we queue them for replay reintegration
6640 - fix last_rcvd offset calculation
6641 - sync after each recovered transaction, so we always make progress
6642 - never, not always, ERESTART requests without transnos
6643 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6645 - skip replay if the MDS says that the client is already connected
6646 - don't check for a recovery-enabled export to match lctl's UUID
6647 - don't INC_USE_COUNT for phantom exports
6648 - don't crash when cleaning up phantom exports (567)
6649 - don't double-finish or set replay data for errored mdc_open requests
6650 - abort requests when they time out, so we don't get old replies
6651 - send/receive replies for AST messages again
6652 - if the client says that it doesn't have the lock, cancel it on the
6654 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6655 mark it as destroyed, it will all work out when decref is called
6656 - fix module use counts (22, 581)
6658 - ASTs now expect a reply (server cancels lock on error reply)
6660 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6663 - fix many simultaneous client startup (392)
6664 - fix dentry->d_it clobbering
6665 - credentials weren't being shipped for readdir/getattr operations
6666 - remove invalid assertions triggered during some concurrent MD
6668 - proper Lustre versions added (336, 389)
6669 - fix memory leak for create error case (398)
6670 - fix LOV locking bug that would get cli/srv out of sync
6671 - fix echo client over LOV (409)
6672 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6673 - fix double-O_EXCL intent crash (424)
6674 - avoid sending multiple lock CANCELs (352)
6676 - MDS can do multi-client recovery (modulo bugs in new code)
6678 - many updates, edits, cleanups
6680 2002-11-18 Phil Schwan <phil@clusterfs.com>
6683 - fix null d_it dereference (346)
6684 - fix full OST/dbench hang (333)
6685 - fix permission problem with file removal (286)
6686 - fix removal of OSCs from LOV when they fail
6687 - fix NULL deref during bulk timeout (214)
6688 - fix problems related to multiple filesystems on one MDS (241)
6689 - fixed serious subtle metadata locking bugs
6690 - free locks on clients when inodes are removed due to memory
6692 - fix inode pointer in lock data (285)
6693 - partial support for multiple MDS on a single host (241)
6694 - data locks weren't cancelled at clear_inode time (290, 311)
6695 - intent locks could lead to unbounded lock growth (205)
6696 - added a maximum lock count, an LRU list, and a flusher
6697 - fix multiple rename (365)
6698 - properly abstracted the echo client
6699 - OSC locked 1 byte too many; fixed
6700 - rewrote brw callback code:
6701 - fixed recovery bugs related to LOVs (306)
6702 - fixed too-many-pages-in-one-write crash (191)
6703 - fixed (again) crash in sync_io_timeout (214)
6704 - probably fixed callback-related race (385)
6706 - Add capability to MDS protocol
6707 - LDLM cancellations and callbacks on different portals
6709 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6712 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6713 - defer open object destruction to close time (601981, 138)
6714 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6715 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6716 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6717 - don't call obd_setattr in ll_file_release for destroyed objects
6718 * protocol change to lustre_msg: move |version| and add |flags|
6719 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6720 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6721 * added replay of create, unlink, link and rename operations during
6722 MDS failover; recovery should be much more robust now
6723 * remove failed OSCs from LOVs (only lov_create uses this so far)
6724 * the lustre-HOWTO was brought (more) up to date (582544)
6726 2002-10-23 Phil Schwan <phil@clusterfs.com>
6729 - in-use dentries weren't being reused properly (617851)
6730 - prevent multiple LDLM setup (599178)
6731 - fix LOV size calculations for truncate (617853)
6732 - fix client handling of MDS intent errors (POSIX)
6733 - fix permission bug in lovstripe.c test (624321)
6734 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6735 - truncate past end of file could corrupt data
6736 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6737 - a race in recovery could return ETIMEDOUT to apps (623947)
6738 - building outside the source directory was fixed
6739 * the lustre-HOWTO was brought (more) up to date (582544)
6740 * major progress was made on recovery functionality
6742 2002-10-10 Phil Schwan <phil@clusterfs.com>
6745 - recovery deadlock fix
6746 - rm -rf causes LBUG fix (617817)
6747 - file open by multiple tasks fix (618962)
6748 - directory permissions bugs (602707 and 620007)
6749 - journal_stop fixed with locking (611313)
6750 - O_APPEND failures resolved (618273, perhaps 614459)
6751 - lconf PATH fix (619770)
6752 - IA64 build fix (621450)
6753 - RPC buffer sizes scale with amount of memory
6755 2002-10-01 Phil Schwan <phil@clusterfs.com>
6758 - locks would be cancelled without throwing away data pages,
6759 resulting in inconsistent data (605627)
6760 - inode attributes were not always being refreshed (605627, 612449)
6761 - lconf now continues to cleanup after lctl reports an error
6762 - MDS now enforces user permissions (602707)
6763 - lprocfs cleanup fixed, but not yet enabled (614157)
6764 - fixed infinite server hang, should a client not respond to an AST
6765 - avoid going into recovery if user calls readlink() with a buffer
6766 that's too small (613941)
6767 - AST RPCs no longer require replies (614867) -- this may be changed
6768 - don't crash server if client sends an IOV that's too big (611336)
6769 - fixed lock conversion deadlock (611892)
6770 - fixed the following of symlinks (614622)
6771 * recovery: the server can remove locks from a client that dies, other
6772 clients can make progress
6773 * more extN patch fixes
6774 * compile-time configurable ptlrpc buffer allocations
6776 - collaborative read cache document
6777 - Lustre Lite Performance CDR document-in-progress
6779 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6782 - fix typo in patch-2.4.18
6784 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6787 - clear ptlrpc request each time in handle_incoming_request()
6788 - unlink of files now destroys the object on the OST
6790 2002-09-19 Peter Braam <braam@clusterfs.com>
6792 * add hard link support
6793 * change obdfile creation method
6794 * kernel patch changed
6796 2002-09-19 Peter Braam <braam@clusterfs.com>
6799 - stack overflow bug in extN fixed
6801 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6803 * documentation updates
6804 - add man pages for config tools
6805 - update tests/README to describe testing with new config tools
6806 - finish metadata API descriptions
6807 * bug fixes and cleanups
6808 - statfs workaround for 16TB limit
6809 - LOV stripe allocation improved, can stripe on subset of OSTs
6810 - LOV file size/IO offset was wrong for files > 4GB in size
6811 - object EA data was being dropped, caused files to be unreadable
6812 - memory overflow with non-LOV OST caused memory corruption
6813 - fixed regression tests to work with new config tools, obdfilter
6814 - fixed bug when directory size became larger than 1 block
6815 - fixed bug (for single client case) when PWD was deleted
6816 - invalidate local directory pages when doing intent-based ops
6817 - avoid LDLM oops when lock callback contained bad data
6819 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6821 * documentation updates
6822 * bug fixes and cleanups
6823 - configuration tools
6826 - 64-bit compile warnings
6827 - 64-bit internal statfs data
6829 * test_brw on persistent OST devices
6831 * lprocfs (disabled)
6833 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6835 * documentation updates
6836 * bug fixes and cleanups
6837 * configuration tools
6839 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6842 * many small fixes to 0_5_4
6843 * io/network handling
6844 * thinkos in MDS operations
6846 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6849 * crucial basic fixes to 0.5.3
6850 * IOR, Iozone work over Elan
6853 2002-08-07 Phil Schwan <phil@clusterfs.com>
6854 * version 0_5_3, our first alpha
6855 * we use the new Portals iovs
6856 * documentation updates
6857 * bug fixes and cleanups
6858 * small changes in the DLM wire protocol
6860 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6861 * version 0_5_1 with some initial stability,
6862 * locking on MD and file I/O.
6863 * documentation updates
6864 * several bug fixes since 0.5.0
6865 * small changes in wire protocol
6867 2002-07-18 Phil Schwan <phil@clusterfs.com>
6869 * delivered as Lustre Light Alpha
6870 * fixed a crash after handling invalid MDS requests
6871 * fixed directory pages for architectures with non-4k pages sizes
6873 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6874 * release version v0_4_4
6875 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6877 * Fixes a number of interruption problems with OST operations.
6878 * Update documentation for portals header changes
6879 * Move all wire protocol structs/defines to lustre_idl.h
6880 * Fixes symlink length bug.
6881 * Add tcpdump to repository.
6883 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6884 * release version v0_4_3
6885 * Fixes statfs for inodes on extN.
6886 * Fixes bug in runtests which would delete /etc/hosts.
6887 * Use 64-bit object IDs wherever possible (not into VFS though)
6888 Remove ost_get_info, which is unused by lustre, and out of date.
6890 2002-07-03 Peter Braam <braam@clusterfs.com>
6891 * release version v0_4_2 Fixes a lookup error (type not passed)
6892 * move forward to head of Portals
6893 * move forward to latest Lustre kernel
6895 2002-06-25 Peter Braam <braam@clusterfs.com>
6896 * release version v0_4_1. Hopefully stable on single node use.